Generated configuration field reference

Generated from runtime source and JSON Schema. Do not edit manually. See Configuration for narrative guidance and examples.

• Generator: pnpm docs:config
• Drift check: pnpm check:config-reference

Host top-level fields

The authoritative contract is packages/im/runtime/src/host-config-schema.json, consumed by the Runtime at packages/host/gateway/src/config/configuration-projection.ts.

Path	Type	Required	Default	Description
http	object	no	—	HTTP, Console, REST/RPC/SSE, and Webhook Host.
http.readiness	object	no	—	Required runtime components for /pub/ready; no external network probes.
http.readiness.database	boolean	no	true	Require an initialized Database Host (not a live database query).
http.readiness.endpoints	array<object>	no	—	Required Endpoint admission slots, selected by exact plugin owner and stable slot name.
http.readiness.endpoints[].owner	string	yes	—	—
http.readiness.endpoints[].name	string	yes	—	—
http.readiness.agents	array<string>	no	—	Required configured Agent bindings; does not verify model credentials or provider reachability.
http.host	string | null	no	—	监听地址。
http.port	integer | string | null	no	—	Listening port: an integer from 1 to 65535, or a ${VAR} environment reference that expands to one (0 lets the system pick a free port). When unset the gateway listens on 8068.
http.token	string | null	no	—	Bootstrap token: signs in to the setup wizard until the first Console owner exists, then stops working (ADR 0049).
http.corsOrigins	array | null	no	—	—
http.trustProxy	boolean | integer | array<string> | string	no	—	Reverse proxies whose X-Forwarded-For / Forwarded header names the client (login rate limit, audit IP, loopback-only sockets): false (default) uses the TCP peer; true trusts every hop; a number trusts that many hops; a list trusts peers matching IPs / CIDRs or loopback, private, linklocal.
http.base	string | null	no	—	API 路径前缀。
http.tls	false | null | object	no	—	—
http.listeners	array<object>	no	—	—
http.listeners[].enabled	boolean | string	no	—	是否启用此监听器。
http.listeners[].host	string	no	—	监听地址。
http.listeners[].port	integer | string	no	—	Listening port: an integer from 1 to 65535, or a ${VAR} environment reference that expands to one.
http.listeners[].tls	false | null | object	no	—	—
console	object	no	—	Console Host features.
console.terminal	object	no	—	Console terminal (host shell / SSH) for owners holding ops.terminal; off by default.
console.terminal.enabled	boolean	no	false	Register the terminal routes.
console.terminal.shell	string	no	—	Local shell; defaults to $SHELL (powershell.exe on Windows).
console.terminal.cwd	string	no	"."	Working directory, relative to the project root.
console.terminal.idleTimeout	integer	no	900	Seconds without input before the session closes (0 = never).
console.terminal.maxSessions	integer	no	4	Concurrent sessions across all users.
console.terminal.ssh	array<object>	no	—	SSH targets the Console may open.
console.terminal.ssh[].name	string	yes	—	Target id the Console selects (not "local").
console.terminal.ssh[].label	string	no	—	Display name.
console.terminal.ssh[].host	string	yes	—	SSH host.
console.terminal.ssh[].port	integer	no	22	SSH port.
console.terminal.ssh[].username	string	yes	—	SSH user.
console.terminal.ssh[].privateKeyPath	string	no	—	Private key file (relative to the project root, ~ allowed); exclusive with privateKeyEnv.
console.terminal.ssh[].privateKeyEnv	string	no	—	Environment variable holding the private key.
console.terminal.ssh[].passwordEnv	string	no	—	Environment variable holding the password.
console.terminal.ssh[].passphraseEnv	string	no	—	Environment variable holding the key passphrase.
console.terminal.ssh[].hostFingerprint	string	no	—	Required host key fingerprint (SHA256:…), verified strictly; get it with ssh-keyscan -t ed25519 &lt;host&gt; | ssh-keygen -lf - on a trusted network.
console.terminal.ssh[].insecureAcceptAnyHostKey	boolean	no	—	Accept any host key without verification (unsafe, open to man-in-the-middle); only instead of hostFingerprint. The Console marks the target unsafe.
console.loginAssist	object	no	—	Login assistance shown in the Console (ADR 0058).
console.loginAssist.captchaHelperUrl	string	no	—	Hosted captcha helper page on its own origin; defaults to https://zhin.js.org/login-assist/tcaptcha.html.
gateway	object	no	—	Management gateway (zhin start): instance supervision and dependency transactions.
gateway.readyTimeout	integer	no	60	Seconds the instance has to report ready after a start, a restart or a dependency transaction before the start counts as failed (the transaction then rolls back). Read again before every start, restart and transaction; no gateway restart needed.
database	object	no	—	Database Host and dialect connection options.
database.dialect	string	no	—	数据库方言。
database.filename	string	no	—	SQLite 文件路径。
database.mode	string	no	—	SQLite 兼容模式字段。
database.host	string	no	—	数据库主机。
database.port	number | string	no	—	网络数据库端口。
database.user	string	no	—	MySQL / PostgreSQL 用户名。
database.username	string	no	—	Redis 用户名。
database.password	string	no	—	连接密码。
database.database	string | number	no	—	MySQL / PostgreSQL 数据库名，或 Redis 数据库索引。
database.uri	string	no	—	MySQL 连接 URI（可能含凭据）。
database.connectionString	string	no	—	PostgreSQL 连接串（可能含凭据）。
database.url	string	no	—	MongoDB / Redis 连接 URL（可能含凭据）。
database.dbName	string	no	—	MongoDB 数据库名。
database.pool	object | null	no	—	—
database.pool.min	number | string | null	no	—	最少连接数。
database.pool.max	number | string | null	no	—	最多连接数。
database.pool.acquireTimeoutMillis	number | string | null	no	—	获取连接超时毫秒。
database.pool.idleTimeoutMillis	number | string | null	no	—	空闲连接超时毫秒。
database.socket	object	no	—	—
database.socket.host	string	no	—	Redis 主机。
database.socket.port	number | string	no	—	Redis 端口。
database.socket.path	string	no	—	Unix socket 路径。
database.socket.connectTimeout	number | string	no	—	连接超时毫秒。
database.socket.tls	boolean | string	no	—	启用 TLS。
database.socket.keepAlive	number | string	no	—	TCP keep-alive 毫秒。
database.socket.noDelay	boolean | string	no	—	TCP noDelay。
database.ssl	boolean | string | object	no	—	—
database.charset	string	no	—	MySQL 字符集。
database.timezone	string	no	—	MySQL 时区。
database.connectTimeout	number | string	no	—	MySQL 连接超时毫秒。
database.connectionTimeoutMillis	number | string	no	—	PostgreSQL 连接超时毫秒。
database.query_timeout	number | string	no	—	PostgreSQL 查询超时毫秒。
database.statement_timeout	number | string	no	—	PostgreSQL statement 超时毫秒。
database.application_name	string	no	—	PostgreSQL 应用名称。
database.maxPoolSize	number | string	no	—	MongoDB 最大连接池大小。
database.minPoolSize	number | string	no	—	MongoDB 最小连接池大小。
database.serverSelectionTimeoutMS	number | string	no	—	MongoDB server 选择超时毫秒。
database.connectTimeoutMS	number | string	no	—	MongoDB 连接超时毫秒。
database.socketTimeoutMS	number | string	no	—	MongoDB socket 超时毫秒。
database.authSource	string	no	—	MongoDB 认证数据库。
database.replicaSet	string	no	—	MongoDB 副本集。
database.auth	object	no	—	—
database.auth.username	string	no	—	MongoDB 用户名。
database.auth.password	string	no	—	MongoDB 密码。
database.disableOfflineQueue	boolean | string	no	—	Redis 禁止离线队列。
database.pingInterval	number | string	no	—	Redis 心跳间隔毫秒。
database.console	object	no	—	PostgreSQL / MySQL：Console SQL 控制台使用的受限账号（ADR 0059；zhin db console-role 创建）。未配置时原始 SQL 关闭。
database.console.user	string	no	—	受限账号用户名，通常写 ${ZHIN_DB_CONSOLE_USER}。
database.console.password	string	no	—	受限账号密码，写 ${ZHIN_DB_CONSOLE_PASSWORD}，值只放 .env。
ai	object	no	—	Providers, Agents, sessions, memory, tools, and execution security.
ai.venues	array<object>	no	—	Shared multi-Bot IM rooms (ADR 0048 §3.2).
ai.venues[].id	string	yes	—	—
ai.venues[].label	string	no	—	—
ai.venues[].conversation	object	yes	—	—
ai.venues[].conversation.platform	string	yes	—	—
ai.venues[].conversation.sceneType	string: "private", "group", "channel"	no	—	—
ai.venues[].conversation.sceneId	string	yes	—	—
ai.venues[].conversation.threadId	string	no	—	—
ai.venues[].members	array<object>	yes	—	—
ai.venues[].members[].agent	string	yes	—	—
ai.venues[].members[].botEndpoint	string	yes	—	—
ai.venues[].members[].aliases	array<string>	no	—	—
ai.venues[].entry	string	no	—	—
ai.venues[].visibility	string: "visible", "summary", "backstage"	no	—	—
ai.externalPeers	array<object>	no	—	Declared third-party Bot accounts (ADR 0043/0048).
ai.externalPeers[].platform	string	yes	—	—
ai.externalPeers[].userId	string	yes	—	—
ai.externalPeers[].label	string	no	—	—
ai.agent	object	no	—	Agent execution, queueing, tool, and model policies.
ai.agent.mailbox	object	no	—	IM inbox policy of the agent Mailbox (ADR 0048 H3; renamed from inboundQueue, which now fails startup).
ai.agent.mailbox.groupMode	string: "supersede", "fifo"	no	—	supersede (default): a sender's new message replaces their own unanswered turn; fifo: every message waits in arrival order. Other members' messages always queue.
ai.agent.mailbox.ttlMs	integer	no	180000	A queued turn that waited longer than this is dropped with a short notice. 0 disables expiry.
ai.agent.mailbox.maxQueuedTurns	integer	no	5	Turns allowed to wait behind the running turn of one conversation; further messages get a short busy reply.
ai.agent.mailbox.merge	object	no	—	Group merge window (ADR 0038): mentions from several members within a short window get one combined reply. Private chats, the master, commands, stop words and messages with media, quotes or forwards never merge.
ai.agent.mailbox.merge.enabled	boolean	no	true	Turn the group merge window on (group chats only). Set false to answer every member one by one as before.
ai.agent.mailbox.merge.quietMs	integer	no	1500	The window is answered once no new message joined it for this long.
ai.agent.mailbox.merge.maxWaitMs	integer	no	6000	The window is answered at the latest this long after its first message.
ai.agent.mailbox.merge.maxSenders	integer	no	5	Senders one merged reply covers; further senders wait for the next window.
ai.agent.mailbox.live	object	no	—	Danmaku mode (ADR 0038): live-stream adapters that declare liveDanmaku collect danmaku in a window and reply once, with read-only tools and no approval cards.
ai.agent.mailbox.live.mode	string: "pick-one", "summary"	no	"pick-one"	pick-one answers one danmaku of the window; summary answers the window as a whole once.
ai.agent.mailbox.live.windowMs	integer	no	10000	How long a danmaku window collects messages from its first danmaku.
ai.agent.mailbox.live.minIntervalMs	integer	no	15000	Minimum time between two danmaku replies in one live room.
ai.agent.mailbox.coalesceWindowMs	number | string	no	—	—
ai.agent.execSecurity	string: "deny", "allowlist", "full"	no	—	Shell command security boundary.
ai.agent.execPreset	string: "readonly", "network", "development", "custom"	no	—	Command allowlist preset used outside full mode.
ai.agent.network	object	no	—	Network access for IM turns (web_fetch / web_search and networked commands). Off by default; enabled allows HTTPS egress, optionally limited to allowedDomains. When enabled is unset, execPreset: network still enables it for compatibility.
ai.agent.network.enabled	boolean	no	false	Allow HTTPS egress for IM turns.
ai.agent.network.allowedDomains	array<string>	no	—	When non-empty, only these domains (and their subdomains) are reachable.
ai.agent.masterShell	string: "off", "isolated"	no	"off"	Shell access for the master in IM private chats. off: no bash in IM turns. isolated: master DMs run bash inside a Docker container (workspace writable, no network, development allowlist, approval for anything else); fails closed when Docker is unavailable.
ai.agent.masterShellApprovalMode	string: "ask", "auto", "inherit"	no	"ask"	How approval-gated calls are approved in the master's IM private chat when masterShell is isolated. ask: show the master an approval prompt; auto: automatic reviewer (a reviewer rejection is escalated to the master); inherit: use execApprovalMode.
ai.agent.toolBudget	object	no	—	Interactive runtime turns and their synchronous children have no fixed step or total wall-clock cap. They share configured token quotas. Step/time caps and continuation settings apply to bounded and legacy executions.
ai.agent.toolBudget.extendBy	integer	no	10	Steps added per extension when progressing; 0 disables.
ai.agent.toolBudget.progressWindow	integer	no	5	Progress must have happened within this many recent steps.
ai.agent.toolBudget.maxTotalIterations	integer	no	40	Hard step cap per turn segment.
ai.agent.toolBudget.maxSubagentTotalIterations	integer	no	30	Hard step cap for an in-turn sub-agent.
ai.agent.toolBudget.maxWallClockMs	integer	no	600000	Wall-clock cap for the whole turn in ms; 0 disables.
ai.agent.toolBudget.maxTotalTokens	integer	no	0	Token cap for the whole turn; 0 disables.
ai.agent.toolBudget.failureStreakLimit	integer	no	3	Consecutive tool failures that stop extensions.
ai.agent.toolBudget.repeatLimit	integer	no	2	Consecutive identical repeated calls that stop extensions.
ai.agent.toolBudget.autoContinue	string: "off", "master_dm", "on"	no	"master_dm"	Auto-continue a progressing turn that hit its budget: off, only in the master's private chat, or always. Continuations reuse the same turn (same policy and approvals).
ai.agent.toolBudget.maxContinuations	integer	no	2	Maximum automatic continuations per turn.
ai.agent.toolBudget.progressUpdates	string: "off", "private", "all"	no	"private"	off disables progress. Continuous interactive turns report in their current conversation; private/all select scope for bounded turns.
ai.agent.toolBudget.progressMinIntervalMs	integer	no	45000	Minimum time between progress updates in ms.
ai.agent.toolBudget.progressMaxUpdates	integer	no	3	0 disables progress. Positive values cap bounded turns only; continuous interactive turns use the configured interval without a count cap.
ai.agent.injectionGuard	object	no	—	Prompt-injection defences. After lower-trust content (tool results, web pages, files, sub-agent reports, background messages) enters a turn, high-risk actions (shell, messaging other conversations, config changes, scheduling) need approval; replies are checked for system-prompt, secret and other-session memory leaks.
ai.agent.injectionGuard.confirmHighRisk	string: "external_content", "off"	no	"external_content"	external_content: high-risk actions after external content always ask for approval (denied in unattended turns). off: disabled.
ai.agent.injectionGuard.outboundCheck	string: "redact", "off"	no	"redact"	redact: mask leaked system-prompt lines, secrets and other sessions' memory in replies and outbound tool messages. off: disabled.
ai.agent.injectionGuard.trustedOutputTools	array<string>	no	[]	Tools whose results do not count as external content.
ai.agent.toolHardGate	object	no	—	Dangerous-tool hard gate (ADR 0036). Always on and independent of execApprovalMode: tools that manage members, publish content, move money or change the bot's identity (a core list, plus tools that declare risk: dangerous), and any call that writes the host's own configuration, dependencies or git hooks (zhin.config., .env, package.json, lockfiles, node_modules/, .git/, …) never run without a fresh human approval or a Console-approved card, even under bypass. A denial fails closed and is not retried in the same turn. Settings can only make it stricter.
ai.agent.toolHardGate.dangerousTools	array<string>	no	[]	Extra tools to treat as dangerous (local or qualified names). Cannot mark a core dangerous tool safe.
ai.agent.toolHardGate.allowFunds	boolean	no	false	Allow funds tools (red packets, transfers, payments). Off by default; when on they still need approval.
ai.agent.toolHardGate.approvalTimeoutMs	integer	no	300000	How long a dangerous call waits for approval before it is denied (ms).
ai.agent.toolHardGate.maxPendingPerScene	integer	no	3	Approval cards one group (scene) may have pending at once (ADR 0038). Beyond it a dangerous call is denied without a new card; asking again for an already pending card does not count.
ai.agent.toolHardGate.imApproval	object	no	—	IM approval channel for the hard gate (#1013). Off by default. When on, each new approval card is also sent to allowlisted approvers, who settle it with approve &lt;id&gt; / deny &lt;id&gt; in a private chat (or the dedicated approver group). It shares the Console approval queue: whichever answers first wins. Messages in business groups never settle a card; there is no session-wide or permanent approval.
ai.agent.toolHardGate.imApproval.enabled	boolean	no	false	Turn the IM approval channel on.
ai.agent.toolHardGate.imApproval.approverIds	array<string> | object	no	[]	Approver platform user ids: a flat list applies to every platform; an object maps a platform (adapter name) to its user ids. Anyone not listed cannot settle a card over IM.
ai.agent.toolHardGate.imApproval.delivery	string: "dm", "approver_group"	no	"dm"	dm sends each card to every approver privately; approver_group posts it to approverGroupId, which must not be the conversation that triggered the call.
ai.agent.toolHardGate.imApproval.approverGroupId	string	no	—	Dedicated approver group id, required for delivery: approver_group. Without it the IM channel stays unavailable and only the Console can approve.
ai.agent.reactions	object	no	—	Emoji-reaction pacing per session. The model may propose a reaction on any turn; the host applies one only if at least cooldownMs passed since the last applied reaction in that session and a random roll passes probability. Counters are kept in memory.
ai.agent.reactions.cooldownMs	integer	no	600000	Minimum gap between two applied reactions in one session (ms).
ai.agent.reactions.probability	number	no	0.35	Chance that an allowed reaction is applied; 0 disables reactions.
ai.agent.usageQuota	object	no	—	Per-user and per-group token quotas and a per-group rate limit. All limits are off (0) by default; the owner is exempt by default. Counters are kept in memory.
ai.agent.usageQuota.windowMs	integer	no	86400000	Quota window in ms.
ai.agent.usageQuota.perUserTokens	integer	no	0	Tokens one user may use per window; 0 = unlimited.
ai.agent.usageQuota.perGroupTokens	integer	no	0	Tokens one group may use per window; 0 = unlimited.
ai.agent.usageQuota.perGroupRequestsPerMinute	integer	no	0	Turns one group may start per minute; 0 = unlimited.
ai.agent.usageQuota.exemptOwner	boolean	no	true	The owner is never limited.
ai.agent.execApprovalMode	string: "ask", "bypass", "auto"	no	—	Approval policy for main Agent commands.
ai.agent.approval	object	no	—	IM approval options.
ai.agent.approval.adminSelfApprove	boolean	no	false	Let a zhin admin (held at endpoint level or above) approve, once, in their own private chat, the operations a policy marks admin-approvable (tools tagged admin_approvable such as schedule_add, session memory) when the bot owner cannot answer there. Default false: only the bot owner approves. Never covers sensitive files, shell commands, global/platform memory, human-only or dangerous tools, or groups without an owner.
ai.agent.toolExecution	string: "parallel", "sequential", "tiered"	no	—	How tool calls in one model step are scheduled.
ai.agent.modelSizeHint	string: "", "small", "medium", "large"	no	—	Optional model-size hint; an empty string clears the hint.
ai.agent.promptCacheRetention	string: "in_memory", "24h"	no	—	Provider prompt-cache retention policy.
ai.agent.steeringMode	string: "one-at-a-time", "all"	no	—	Process steering messages one at a time or drain all pending messages together.
ai.agent.followUpMode	string: "one-at-a-time", "all"	no	—	Process follow-up messages one at a time or drain all pending messages together.
ai.agent.outputSchema	boolean | string: "segments" | object	no	—	Final-output mode: defaults to segments for IM conversations; false for text, true or segments for ordered canonical segments, or a custom JSON Schema object. Detached and non-IM turns do not inherit the IM preset.
ai.agent.schedule	object	no	—	Unattended Schedule execution policy.
ai.agent.schedule.security	object	no	—	Schedule command security policy.
ai.agent.schedule.security.execPreset	string: "readonly", "network"	no	—	Schedule Jobs may use only the read-only or network preset.
ai.agent.schedule.security.allowedDomains	array<string>	no	—	—
ai.agent.schedule.budget	object	no	—	—
ai.agent.schedule.budget.maxTokens	number | string	no	—	—
ai.agent.schedule.budget.maxToolCalls	number | string	no	—	—
ai.agent.schedule.budget.timeoutMs	number | string	no	—	—
ai.agent.execAllowlist	array<string>	no	—	allowlist 模式下允许的命令（支持正则字符串，如 "^ls "、"^cat "），与 preset 合并
ai.agent.phaseTrace	boolean | string	no	—	观测主回合阶段日志（或通过 ZHIN_AGENT_PHASE_TRACE=1 开启）
ai.agent.modelHarness	object	no	—	按模型 / provider 模式覆盖 model harness
ai.agent.modelHarness.services	object	no	—	Exact runtime Service id; native-model entries override that Service default.
ai.agent.modelHarness.services.<key>.defaults	object	no	—	—
ai.agent.modelHarness.services.<key>.defaults.maxIterations	number | string	no	—	—
ai.agent.modelHarness.services.<key>.defaults.explorationBudget	number | string	no	—	连续只读探索调用达到该次数时注入提示；0 关闭
ai.agent.modelHarness.services.<key>.defaults.explorationGuidance	boolean | string	no	—	在系统提示中加入"先规划、限定目标文件"的探索纪律
ai.agent.modelHarness.services.<key>.defaults.iterationExtension	number | string	no	—	每次有进展时延长的步数（覆盖 toolBudget.extendBy）
ai.agent.modelHarness.services.<key>.defaults.maxTotalIterations	number | string	no	—	单段回合的硬步数上限（覆盖 toolBudget.maxTotalIterations）
ai.agent.modelHarness.services.<key>.defaults.emptyReplyRetries	number | string	no	—	空回复（无文本且无工具调用）时额外重试次数；默认 0
ai.agent.modelHarness.services.<key>.defaults.requestTimeoutMs	number | string	no	—	单次模型请求超时（毫秒）；0 关闭
ai.agent.modelHarness.services.<key>.models	object	no	—	—
ai.agent.modelHarness.services.<key>.models.<key>.maxIterations	number | string	no	—	—
ai.agent.modelHarness.services.<key>.models.<key>.explorationBudget	number | string	no	—	连续只读探索调用达到该次数时注入提示；0 关闭
ai.agent.modelHarness.services.<key>.models.<key>.explorationGuidance	boolean | string	no	—	在系统提示中加入"先规划、限定目标文件"的探索纪律
ai.agent.modelHarness.services.<key>.models.<key>.iterationExtension	number | string	no	—	每次有进展时延长的步数（覆盖 toolBudget.extendBy）
ai.agent.modelHarness.services.<key>.models.<key>.maxTotalIterations	number | string	no	—	单段回合的硬步数上限（覆盖 toolBudget.maxTotalIterations）
ai.agent.modelHarness.services.<key>.models.<key>.emptyReplyRetries	number | string	no	—	空回复（无文本且无工具调用）时额外重试次数；默认 0
ai.agent.modelHarness.services.<key>.models.<key>.requestTimeoutMs	number | string	no	—	单次模型请求超时（毫秒）；0 关闭
ai.agent.providerResilience	object	no	—	固定模型韧性：重试 / 退避 / 超时（见 ProviderResilienceConfig）
ai.agent.providerResilience.maxRetries	number | string	no	—	Retries per model after the first attempt. Default 2. 0 disables retry.
ai.agent.providerResilience.baseDelayMs	number | string	no	—	First backoff delay in ms; doubles each retry. Default 1000.
ai.agent.providerResilience.maxDelayMs	number | string	no	—	Upper bound for one backoff delay in ms. Default 10000.
ai.agent.providerResilience.maxRetryAfterMs	number | string	no	—	A Retry-After longer than this stops retry. Default 30000.
ai.agent.providerResilience.requestTimeoutMs	number | string	no	—	Per-attempt timeout in ms (whole streamed completion). Default 120000. 0 disables.
ai.agent.providerResilience.firstResponseTimeoutMs	number | string	no	—	Per-attempt wait in ms for the provider's first response (response headers or the first stream event). Default 30000. 0 disables. A service that does not answer in time is not re-asked on the same model: the attempt fails as no_response.
ai.agent.providerResilience.emptyReplyRetries	number | string	no	—	Extra attempts when the model returns neither text nor tool calls (or a tool call without a name). Default 0: the reply is passed through unchanged. Quirk presets (see model harness) raise this for models known to emit spurious empty replies.
ai.agent.deferredTools	object	no	—	工具 schema 按需加载
ai.agent.deferredTools.maxLoadedPerSession	number | string	no	—	—
ai.agent.deferredTools.discoverTopK	number | string	no	—	—
ai.agent.deferredTools.alwaysLoadedTools	array<string>	no	—	额外常驻工具；在当前场景档（群聊、私聊、主人私聊）的常驻集合基础上追加
ai.agent.deferredTools.alwaysLoadedMode	string	no	—	extend（默认）追加到内置常驻集合；replace 完全替换
ai.agent.deferredTools.mcpServers	object	no	—	—
ai.agent.deferredTools.mcpServers.<key>.alwaysLoaded	array<string>	no	—	—
ai.agent.persona	string	no	—	—
ai.agent.maxIterations	number | string	no	—	—
ai.agent.timeout	number | string	no	—	—
ai.agent.preExecTimeout	number | string	no	—	—
ai.agent.maxTools	number | string	no	—	—
ai.agent.contextTailMessageLimit	number | string	no	—	—
ai.agent.rateLimit	object	no	—	—
ai.agent.rateLimit.maxRequestsPerMinute	number | string	no	—	每分钟最大请求数（默认 20）
ai.agent.rateLimit.cooldownSeconds	number | string	no	—	冷却时间（秒），超限后需等待（默认 10）
ai.agent.toneAwareness	boolean | string	no	—	—
ai.agent.contextTokens	number | string	no	—	—
ai.agent.maxHistoryShare	number | string	no	—	—
ai.agent.compaction	object	no	—	—
ai.agent.compaction.enabled	boolean | string	no	—	—
ai.agent.compaction.auto	boolean | string	no	—	—
ai.agent.compaction.keepRecentTokens	number | string	no	—	—
ai.agent.compaction.minKeepCount	number | string	no	—	—
ai.agent.compaction.thresholdRatio	number | string	no	—	Share of the model context window (system prompt and tools included) at which auto-compaction runs. Default 0.7 (#730).
ai.agent.disabledTools	array<string>	no	—	—
ai.agent.allowedTools	array<string>	no	—	—
ai.agent.maxSubagentIterations	number | string	no	—	—
ai.agent.skillInstructionMaxChars	number | string	no	—	—
ai.agent.promptTrace	boolean | string	no	—	—
ai.agent.promptTraceVerbose	boolean | string	no	—	—
ai.agent.recordRequestBodies	boolean	no	—	—
ai.agent.promptCache	boolean | string	no	—	—
ai.agent.promptCacheKeyPrefix	string	no	—	—
ai.agent.gitStatus	boolean | string	no	—	§1 Runtime 注入单行 git 状态摘要（非 git 仓库自动跳过）
ai.agent.contextPaths	array<string>	no	—	追加的上下文文件路径（支持 ~ 与相对路径；默认另加载 ~/.config/zhin/AGENTS.md 等全局路径）
ai.agent.systemPromptMaxChars	number | string	no	—	系统提示词总字符上限；按 retention/order 让出预算，required 内容放不下则失败。
ai.agent.thinkingPreview	boolean | string	no	—	Activity feedback 展示 LLM 实际 thinking 内容（截断），而非静态 "思考中..."。
ai.agent.thinkingPreviewMaxLength	number | string	no	—	thinkingPreview 展示的最大字符数。
ai.enabled	boolean | string	no	—	—
ai.agents	object | null	no	—	Agent authority and display metadata. Model selection belongs to runtime Profile purposes.
ai.agents.<key>.mcpServers	array | null	no	—	—
ai.agents.<key>.nickname	string	no	—	Fallback nickname when the Endpoint does not provide account profile data.
ai.sessions	object | null	no	—	—
ai.sessions.maxHistory	number | string	no	—	最大历史消息数（数据库模式默认200，内存模式默认100）
ai.sessions.expireMs	number | string	no	—	会话过期时间（毫秒，数据库模式默认7天，内存模式默认24小时）
ai.sessions.useDatabase	boolean | string	no	—	是否使用数据库持久化存储（默认 true）
ai.sessions.coldStartMaxMessages	number | string	no	—	无摘要时冷启动最多条数（默认 50）
ai.sessions.coldStartMaxAgeMs	number | string	no	—	冷启动时间窗（毫秒，默认 24h）
ai.sessions.sessionIdleArchiveMs	number | string	no	—	空闲后自动归档 active 会话（毫秒，默认 7 天；0=关闭）
ai.context	object | null	no	—	—
ai.context.enabled	boolean | string	no	—	是否启用消息记录（默认 true）
ai.context.maxRecentMessages	number | string	no	—	历史消息条数上限（默认 100）；与 historyTokens 取先到者
ai.context.historyTokens	number | string | object	no	—	每轮发给模型的历史 token 预算：数字对所有场景生效，也可写 { group, private, masterPrivate }；默认群聊 8000、普通私聊 12000、主人私聊 24000；0 = 不按 token 限制
ai.context.historyTokens.group	number | string	no	—	群聊 / 频道（默认 8000）
ai.context.historyTokens.private	number | string	no	—	普通私聊（默认 12000）
ai.context.historyTokens.masterPrivate	number | string	no	—	主人私聊（默认 24000）
ai.context.summaryThreshold	number | string	no	—	触发总结的消息数量阈值（默认 50）
ai.context.keepAfterSummary	number | string	no	—	总结后保留的消息数量（默认 10）
ai.context.maxContextTokens	number | string	no	—	上下文最大 token 估算（默认 4000）
ai.context.summaryPrompt	string	no	—	自定义总结提示词
ai.multimodal	object | null	no	—	入站/出站多模态（base64 契约；平台 adapter 负责编解码与发送）
ai.multimodal.enabled	boolean | string	no	—	—
ai.multimodal.maxFileBytes	number | string	no	—	—
ai.multimodal.inboundDir	string	no	—	—
ai.multimodal.outboundDir	string	no	—	—
ai.multimodal.image	object	no	—	—
ai.multimodal.image.maxDimension	number | string	no	—	—
ai.multimodal.image.preferNativeVision	boolean | string	no	—	—
ai.multimodal.audio	object	no	—	—
ai.multimodal.audio.strategy	string	no	—	—
ai.multimodal.video	object	no	—	—
ai.multimodal.video.strategy	string	no	—	—
ai.multimodal.video.mcpServer	string	no	—	—
ai.multimodal.video.maxFrames	number | string	no	—	—
ai.multimodal.outbound	object	no	—	—
ai.multimodal.outbound.splitMessages	string	no	—	—
ai.multimodal.perception	object | null	no	—	Perception pre-pass: when the turn's model cannot take an image / video / PDF, this media-capable model describes it and the description is injected as untrusted text ([图片描述] …). Unset = off (placeholders as before).
ai.multimodal.perception.kinds	array<string>	no	—	Kinds to describe. Default ['image']; file covers PDFs only.
ai.multimodal.perception.maxChars	number | string	no	—	Description length cap in characters. Default 800.
ai.multimodal.perception.timeoutMs	number | string	no	—	Per-attempt timeout in ms. Default 30000.
ai.multimodal.perception.maxItemsPerTurn	number | string	no	—	Media described per turn; the rest keep the placeholder. Default 4.
ai.memory	object | null	no	—	三层 Markdown 文件记忆（global / platform / session）。 默认启用；设 enabled: false 可关闭注入。
ai.memory.enabled	boolean | string	no	—	—
ai.memory.semantic	object | null	no	—	L4 语义记忆：memory_entries 表 + memory_search/upsert 工具
ai.memory.semantic.enabled	boolean | string	no	—	—
ai.memory.semantic.autoConsolidate	boolean | string	no	—	—
ai.memory.semantic.recall	object | null	no	—	Per-turn recall (#734): BM25 top-k entries relevant to the current message are added to the turn context. Default on when semantic memory is enabled; topK 5, maxTokens 400.
ai.memory.semantic.recall.enabled	boolean | string	no	—	—
ai.memory.semantic.recall.topK	number | string | null	no	—	—
ai.memory.semantic.recall.maxTokens	number | string | null	no	—	—
ai.memory.budgets	object	no	—	—
ai.memory.budgets.session	number | string	no	—	—
ai.memory.budgets.platform	number | string	no	—	—
ai.memory.budgets.global	number | string	no	—	—
ai.memory.budgets.daily	number | string	no	—	—
ai.knowledge	object	no	—	本地知识库（knowledge_search 工具数据源）。 放置 .md / .txt 文件到 baseDir，Agent 可通过 knowledge_search 工具检索。
ai.knowledge.baseDir	string	no	—	项目根目录内的知识库目录路径。配置该段时必须显式声明。
ai.githubMcp	object	no	—	PAT for adapter-github auto-registered server-github MCP (overrides env when set).
ai.githubMcp.token	string	no	—	—
ai.githubMcp.enabled	boolean | string	no	—	为 true 时注册 server-github MCP（PAT，人身份）；默认 false，Bot 写操作用 adapter github_* 工具
ai.mcpServers	array | null	no	—	External MCP servers for the Agent MCP client (global, all bots). Listed servers are exposed to ZhinAgent tool pool when connected.
ai.access	object | null	no	—	AI 访问控制（平台 AIGC 合规；Endpoint 可用 aiAccess 覆盖）
ai.access.mode	string	no	—	open | closed | whitelist，默认 open
ai.access.users	array<string>	no	—	允许触发 AI 的 sender.id
ai.access.groups	array<string>	no	—	允许触发 AI 的 group/channel id
ai.access.denyMessage	string	no	—	私聊拒绝时的回复文案
ai.trigger	object | null	no	—	AI 触发配置
ai.trigger.enabled	boolean | string	no	—	是否启用（默认 true）
ai.trigger.prefixes	array<string>	no	—	触发前缀列表（默认 ['#', 'AI:']）
ai.trigger.respondToAt	boolean | string	no	—	是否响应 @ 机器人（默认 true）
ai.trigger.respondToPrivate	boolean | string	no	—	是否响应私聊（默认 true）
ai.trigger.keywords	array<string>	no	—	触发关键词（可选）
ai.trigger.ignorePrefixes	array<string>	no	—	忽略的前缀（命令前缀，避免与命令冲突，默认 ['/', '!', '！']）
ai.trigger.timeout	number | string	no	—	超时时间（毫秒，默认 60000）
ai.trigger.thinkingMessage	string	no	—	思考中提示语（可选，设置后会在处理前发送）
ai.trigger.errorTemplate	string	no	—	入口层错误提示模板，{error} 为错误信息（可选；不配置时用聊天口吻说"刚卡了一下"，主人额外看到原因）
ai.trigger.ambient	object	no	—	群聊旁听：群里没人叫机器人时，由判断模型决定是否接话。被 @、引用和私聊不经过这里，总会交给 Agent（Agent 也可以选择不回复）。
ai.trigger.ambient.conversations	array<string>	no	—	开启旁听的群，格式 platform:groupId；* 表示所有群。默认为空（关闭）
ai.trigger.ambient.quietMs	number | string	no	—	群安静多久后才判断一次（毫秒，默认 4000）
ai.trigger.ambient.cooldownMs	number | string	no	—	两次主动插话之间的最短间隔（毫秒，默认 120000）；对话进行中不受限制
ai.trigger.ambient.maxPerHour	number | string	no	—	每个群每小时最多主动插话次数（默认 3）；对话进行中不计入
ai.trigger.ambient.conversationWindowMs	number | string	no	—	机器人说话后，这段时间内算作对话进行中：群友接它的话只由判断模型决定，不受冷却和每小时次数限制（毫秒，默认 180000）
ai.trigger.ambient.judgeIntervalMs	number | string	no	—	没在对话时，同一个群至少隔这么久才判断一次（毫秒，默认 30000），节省判断模型的额度；窗口内的消息顺延到窗口结束再判，不会丢
ai.trigger.ambient.minConfidence	number | string	no	—	判断模型置信度门槛 0..1（默认 0.7）
ai.trigger.ambient.mode	string: "active", "shadow"	no	—	shadow 只记录判断结果、不回复（默认 active）
ai.trigger.sceneEvents	object	no	—	Scene awareness: high-priority social events (bot muted/unmuted, join request, member joined/left, notice updated, bot poked) wake one turn on their own without waiting for the next human message; low-priority events (reactions, card/title changes) never wake alone and only appear in the next turn's [Scene events] summary. When the bot cannot speak, only the session profile is updated and the model is not called.
ai.trigger.sceneEvents.conversations	array<string>	no	—	Groups/channels that social events may wake, as platform:sceneId; * for every group. Defaults to ai.trigger.ambient.conversations (no wake when neither is set)
ai.trigger.sceneEvents.coalesceMs	number | string	no	—	Coalesce window per scene in ms, clamped to 3000–5000 (default 4000)
ai.trigger.sceneEvents.minIntervalMs	number | string	no	—	Minimum gap between two wakes in one scene in ms (default 15000); early batches are deferred, not dropped
ai.usage	object | null	no	—	结构化用量存储（SQLite、MySQL、PostgreSQL）；独立于日志级别，不保存正文或工具参数。
ai.usage.retentionDays	integer | string	no	30	用量明细保留天数，正整数，默认 30 天；按回合结束时间清理，子调用与所属回合一起删除。
ai.audit	object | null	no	—	决策审计 ai_decision_events（ADR 0036）：每次进不进模型、说不说话、调不调工具都记一行，在 Console「机器人状态 → 决策回放」查看与导出；不保存正文、提示词或密钥。
ai.audit.retentionDays	integer | string	no	14	决策审计保留天数，正整数，默认 14 天；启动时与之后每小时按事件时间清理过期行。
ai.imageStorage	object	no	—	Optional public R2 storage for generated images. Bucket access and lifecycle are managed by the operator.
ai.imageStorage.accountId	string	yes	—	—
ai.imageStorage.bucket	string	yes	—	—
ai.imageStorage.accessKeyId	string	yes	—	—
ai.imageStorage.secretAccessKey	string	yes	—	—
ai.imageStorage.publicBaseUrl	string	yes	—	Public HTTPS base URL; not the S3 upload endpoint.
ai.imageStorage.prefix	string	yes	—	Object directory ending in /; must be covered by the bucket lifecycle.
ai.personalPrincipals	object	no	—	Administrator mapping from authenticated Console principal ids to exact IM owners; never inferred from client input.
menuKeyword	string | false	no	—	Built-in command menu keyword. Defaults to 菜单; false or an empty string disables the menu.
commandSuggestion	boolean | object	no	—	Reply "no such command, did you mean …" instead of handing a near-miss command (private chat, @mention or prefixed message) to the AI. false disables it.
menuStyle	object	no	—	Command menu appearance when an HTML renderer is available; plain text fallback remains available.
menuStyle.title	string	no	—	Menu title.
menuStyle.subtitle	string	no	—	Menu subtitle.
menuStyle.accentColor	string	no	—	Six-digit hex accent color; invalid values use the renderer default.
menuStyle.backgroundImage	string	no	—	HTTP(S) image or PNG/JPEG/WebP data URL; unsupported values use the renderer default.
mcp	object	no	—	Expose Bot tools through an MCP Server.
mcp.enabled	boolean | string	no	—	—
mcp.path	string | null	no	—	—
mcp.token	string | null	no	—	—
mcp.allowUnauthenticatedLocalhost	boolean | string	no	—	—
a2a	object	no	—	A2A Agent Card, remote execution, and cross-process Mailbox transport.
a2a.enabled	boolean | string	no	—	—
a2a.path	string | null	no	—	—
a2a.token	string | null	no	—	—
a2a.publicUrl	string | null	no	—	—
a2a.mailbox	object | null	no	—	Cross-process A2A Mailbox transport for signed Mail + Authority + Budget envelopes (ADR 0048 §3.6).
a2a.mailbox.enabled	boolean | string	no	—	—
a2a.mailbox.path	string	no	—	Inbound HTTP route path (defaults to /a2a/mailbox).
a2a.mailbox.secret	string	no	—	—
a2a.mailbox.maxBodyBytes	number | string	no	—	—
a2a.mailbox.peers	array<object>	no	—	—
a2a.mailbox.peers[].actor	string	yes	—	—
a2a.mailbox.peers[].url	string	yes	—	—
a2a.mailbox.peers[].token	string	no	—	—
a2a.mailbox.peers[].secret	string	no	—	—
a2a.mailbox.peers[].timeoutMs	number | string	no	—	—
speech	object	no	—	Speech-to-text and text-to-speech Host.
speech.stt	object | null	no	—	—
speech.stt.enabled	boolean | string	no	—	—
speech.stt.provider	string	no	—	—
speech.stt.model	string	no	—	—
speech.stt.host	string	no	—	—
speech.stt.apiKey	string	no	—	—
speech.tts	object | null	no	—	—
speech.tts.enabled	boolean | string	no	—	—
speech.tts.provider	string	no	—	—
speech.tts.voice	string	no	—	—
speech.tts.model	string	no	—	—
speech.tts.host	string	no	—	—
speech.tts.apiKey	string	no	—	—
speech.tts.rate	string	no	—	edge-tts
speech.tts.pitch	string	no	—	—
speech.tts.edgeTtsCommand	string	no	—	—
speech.tts.region	string	no	—	azure
speech.tts.subscriptionKey	string	no	—	—
speech.tts.baseUrl	string	no	—	custom OpenAI-compatible
speech.tts.headers	object	no	—	—
speech.tts.speed	number | string	no	—	—
htmlRenderer	object	no	—	HTML and image rendering options backed by Shotium.
htmlRenderer.width	number	no	—	Default layout width in CSS pixels.
htmlRenderer.defaultWidth	number	no	—	Legacy alias of width.
htmlRenderer.viewport	object	no	—	Default viewport for Shotium rendering.
htmlRenderer.viewport.width	number	no	—	Viewport width in CSS pixels.
htmlRenderer.viewport.height	number	no	—	Viewport height in CSS pixels.
htmlRenderer.backgroundColor	string	no	—	Fragment background color.
htmlRenderer.defaultBackgroundColor	string	no	—	Legacy alias of backgroundColor.
htmlRenderer.scale	number	no	—	Device scale factor.
htmlRenderer.type	string: "png", "jpeg", "webp"	no	—	Default raster output format.
htmlRenderer.quality	number	no	—	Quality for jpeg/webp output.
htmlRenderer.timeout	number	no	—	Navigation timeout in milliseconds.
htmlRenderer.waitUntil	string: "load", "networkidle"	no	—	Navigation wait strategy.
htmlRenderer.fontFamily	string	no	—	Default font-family for fragments.
htmlRenderer.maxImageHeight	number	no	—	Split overly tall PNGs above this height; 0 disables splitting.
htmlRenderer.sliceCompression	number	no	—	Compression level used when splitting PNGs.
htmlRenderer.allowFileAccess	boolean	no	—	Allow local file subresources.
htmlRenderer.takeOverHtmlSegments	boolean	no	—	Take over html/markdown rich-segment image rendering.
htmlRenderer.cacheDir	string	no	—	HTTP cache directory; use off to disable.
htmlRenderer.cacheMaxBytes	number	no	—	Maximum HTTP cache size in bytes.
htmlRenderer.userAgent	string	no	—	Custom user agent string.
htmlRenderer.idleTimeoutMs	number	no	—	Daemon idle-exit timeout in milliseconds.
htmlRenderer.logStats	boolean	no	—	Log render timing statistics.
htmlRenderer.mode	string: "inprocess", "daemon"	no	—	Whether Shotium runs in-process or as a daemon.
assistant	object	no	—	Scheduled jobs, event ingress, and failure notifications.
assistant.enabled	boolean | string	no	—	启用统一 JobStore（默认 false，Stable 行为不变）
assistant.defaults	object | null	no	—	—
assistant.defaults.notify	object | object | object | object | null	no	—	未显式指定 notify 时的默认投递目标（M3）
assistant.defaults.notifyOnFailure	boolean | string	no	—	Job 失败时是否通知（默认 false）
assistant.profile	object | null	no	—	—
assistant.profile.enabled	boolean | string	no	—	启用 Profile 覆盖 Bootstrap（默认 false）
assistant.profile.file	string | null	no	—	相对项目根路径，默认 assistant.profile.yml
assistant.home	object | null	no	—	—
assistant.home.enabled	boolean | string	no	—	—
assistant.home.restUrl	string	no	—	HA REST 基址，如 http://homeassistant.local:8123
assistant.home.restToken	string	no	—	HA 长期访问令牌
assistant.home.mcpServer	string	no	—	可选：ai.mcpServers 中的名称（REST 为主，MCP 仅校验）
assistant.home.aliases	object | null	no	—	中文/友好名 → entity_id
assistant.home.policy	object | null	no	—	—
assistant.home.policy.requireMaster	boolean | string	no	—	仅 master 可调用 home_*（默认 true）
assistant.home.policy.confirmServices	array | null	no	—	写操作需 Owner 审批的 HA domain（默认 lock、alarm_control_panel）
assistant.home.policy.allowedServiceDomains	array | null	no	—	home_call_service 允许的 domain 白名单（默认 light/climate/scene/cover/script）
assistant.home.watch	array | null	no	—	要通过 HA WebSocket 订阅状态变化的设备别名列表
assistant.home.debounceMs	number | string | null	no	—	同实体状态变化推送的防抖间隔（毫秒，默认 5000）
assistant.events	object | null	no	—	—
assistant.events.enabled	boolean | string	no	—	启用 POST /api/assistant/events（需 assistant.enabled）
assistant.events.token	string	no	—	可选专用 Bearer；未设时复用 http.token（路由仍在 /api 下受全局鉴权保护）。
assistant.events.allowedSources	array<string>	no	—	允许的来源白名单；空数组表示不限制
assistant.events.rateLimitPerMinute	number | string | null	no	—	每来源每分钟最大请求数（默认 60）
log_level	string | number	no	—	Runtime log level.
plugin	object	no	—	Root Plugin configuration; replaced by its project schema during composition.
locale	string: "zh-CN", "en"	no	—	Language used when no personal preference is set; Simplified Chinese when omitted. Language and time zone are independent.
features	object	no	—	Feature configuration beyond the plugin blocks (ADR 0055): in-chat self-service. Restart to apply.
features.selfService	object	no	—	In-chat feature / 功能 and role / 角色 commands for chat role admins (ADR 0055 / 0056). Off by default. Platform group owners / admins are not zhin roles: grant them admin explicitly.
features.selfService.enabled	boolean	no	false	Register the in-chat feature and role commands.

Standalone AI import proposal

ai.plan.yml is for explicit review, apply and export. It is not startup configuration and never automatically replaces the runtime graph.

packages/im/ai/provider-plan.schema.json

Path	Type	Required	Default	Description
version	1	yes	—	—
services	object	no	—	—
services.<key>.driver	string | object	yes	—	—
services.<key>.label	string	no	—	—
services.<key>.options	object	no	—	—
services.<key>.models	object	no	—	—
services.<key>.models.<key>.id	string	yes	—	—
services.<key>.models.<key>.kinds	array<string: "language", "decision", "image">	yes	—	—
services.<key>.models.<key>.protocol	string	yes	—	—
services.<key>.models.<key>.input	array<string: "text", "image", "audio", "video", "file">	no	—	—
services.<key>.models.<key>.streaming	boolean	no	—	—
services.<key>.models.<key>.toolCalling	boolean	no	—	—
services.<key>.models.<key>.thinking	boolean	no	—	—
services.<key>.models.<key>.contextWindow	integer	no	—	—
services.<key>.models.<key>.maxOutputTokens	integer	no	—	—
accounts	object	no	—	—
accounts.<key>.service	string	yes	—	—
accounts.<key>.label	string	no	—	—
accounts.<key>.auth	object | object | object	yes	—	—
profiles	object	no	—	—
profiles.<key>.account	string	yes	—	—
profiles.<key>.model	string	yes	—	—
profiles.<key>.label	string	no	—	—
profiles.<key>.parameters	object	no	—	—
purposes	object	no	—	—
purposes.chat	string | object | null	no	—	—
purposes.summary	string | object | null	no	—	—
purposes.ambient	string | object | null	no	—	—
purposes.image	string | object | null	no	—	—
purposes.perception	object | null	no	—	—
purposes.decisions	object	no	—	—
purposes.decisions.skills	object | null	no	—	—
purposes.decisions.tools	object | null	no	—	—
purposes.decisions.memory	object | null	no	—	—
purposes.decisions.agents	object | null	no	—	—
purposes.decisions.approval	object | null	no	—	—
purposes.decisions.ambient	object | null	no	—	—
purposes.agents	object	no	—	—
purposes.adapters	object	no	—	—
purposes.endpoints	object	no	—	—
purposes.scenes	object	no	—	—
remove	object	no	—	—
remove.services	array<string>	no	—	—
remove.accounts	array<string>	no	—	—
remove.profiles	array<string>	no	—	—

Plugin instance fields

These fields use the Runtime schema reader for trusted plugins’ config: Schema.object(...) declarations or schema.json; isolated plugins must use schema.json. Top-level <name> is the default instanceKey; plugins.<name> only stores the package reference.

bilibili

plugins/adapters/bilibili/schema.json

Path	Type	Required	Default	Description
bilibili.apiBaseUrl	string	no	"https://api.bilibili.com"	—
bilibili.pollIntervalSec	number	no	60	Polling interval in seconds, at least 60; adds 0–20% random jitter.
bilibili.sendMode	string: "review", "off"	no	"review"	Outbound mode: review (default) queues AI replies as drafts; a master approves each send with the bili-draft send command. off disables outbound messages. Automatic direct replies are not supported.
bilibili.minSendIntervalSec	number	no	30	Minimum interval between approved sends in seconds (at least 10). Sending too soon is rejected, not queued.
bilibili.notifications	array<string: "reply", "at">	no	[]	Additional notifications to poll: reply (replies to me) and at (mentions). Requires cookies; video comments, plus dynamic or article comments when dynamics or articles are configured. Disabled by default.
bilibili.notificationScope	string: "whitelist", "all"	no	"whitelist"	Notification scope: whitelist (default) accepts only configured videos, dynamics and articles; all accepts notifications from any video, plus any dynamic or article comment when dynamics or articles are configured.
bilibili.dataDir	string	no	"data/bilibili"	Local data directory. Pending drafts and QR-login credentials are stored under <dataDir>/<endpoint id>/ (relative to the working directory) and survive restarts. Keep it out of the repository.
bilibili.notify	object	no	—	Chat conversation to notify when a new draft arrives or the login expires (usually your private chat with the bot). Without it, drafts only appear on the Console Comment review page and in logs.
bilibili.notify.adapter	string	yes	—	Adapter name, such as icqq or telegram.
bilibili.notify.endpoint	string	yes	—	Endpoint id of that adapter.
bilibili.notify.kind	string: "private", "group", "channel"	no	"private"	Conversation type.
bilibili.notify.id	string | number	yes	—	Conversation id: your account for a private chat, or the group id.
bilibili.videos	array<string | number>	no	—	Allowed videos, using BV or av IDs. Only comments on these videos are polled; no site-wide search.
bilibili.dynamics	array<string>	no	—	Allowed dynamics (posts), as dynamic IDs quoted as strings (the number in t.bilibili.com/<id> or bilibili.com/opus/<id>). Each is resolved to its comment area at startup (image post type=11 / text post type=17); dynamics that cannot be resolved are skipped. No site-wide search.
bilibili.articles	array<string | number>	no	—	Allowed articles, as cv IDs (cv123 or 123); comment area type=12. No site-wide search.
bilibili.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
bilibili.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
bilibili.endpoints[].id	string	yes	—	Local endpoint name, not a Bilibili account ID.
bilibili.endpoints[].sessdata	string	no	—	SESSDATA cookie, acting as your account. Reference ${BILIBILI_SESSDATA}; never commit it to the repository.
bilibili.endpoints[].biliJct	string	no	—	bili_jct cookie (CSRF token). Reference ${BILIBILI_BILI_JCT}.
bilibili.endpoints[].videos	array<string | number>	no	—	Allowed videos, using BV or av IDs. Only comments on these videos are polled; no site-wide search.
bilibili.endpoints[].dynamics	array<string>	no	—	Allowed dynamics (posts), as dynamic IDs quoted as strings (the number in t.bilibili.com/<id> or bilibili.com/opus/<id>). Each is resolved to its comment area at startup (image post type=11 / text post type=17); dynamics that cannot be resolved are skipped. No site-wide search.
bilibili.endpoints[].articles	array<string | number>	no	—	Allowed articles, as cv IDs (cv123 or 123); comment area type=12. No site-wide search.
bilibili.endpoints[].pollIntervalSec	number	no	60	Polling interval in seconds, at least 60; adds 0–20% random jitter.
bilibili.endpoints[].sendMode	string: "review", "off"	no	"review"	Outbound mode: review (default) queues AI replies as drafts; a master approves each send with the bili-draft send command. off disables outbound messages. Automatic direct replies are not supported.
bilibili.endpoints[].minSendIntervalSec	number	no	30	Minimum interval between approved sends in seconds (at least 10). Sending too soon is rejected, not queued.
bilibili.endpoints[].notifications	array<string: "reply", "at">	no	[]	Additional notifications to poll: reply (replies to me) and at (mentions). Requires cookies; video comments, plus dynamic or article comments when dynamics or articles are configured. Disabled by default.
bilibili.endpoints[].notificationScope	string: "whitelist", "all"	no	"whitelist"	Notification scope: whitelist (default) accepts only configured videos, dynamics and articles; all accepts notifications from any video, plus any dynamic or article comment when dynamics or articles are configured.
bilibili.endpoints[].dataDir	string	no	"data/bilibili"	Local data directory. Pending drafts and QR-login credentials are stored under <dataDir>/<endpoint id>/ (relative to the working directory) and survive restarts. Keep it out of the repository.
bilibili.endpoints[].notify	object	no	—	Chat conversation to notify when a new draft arrives or the login expires (usually your private chat with the bot). Without it, drafts only appear on the Console Comment review page and in logs.
bilibili.endpoints[].notify.adapter	string	yes	—	Adapter name, such as icqq or telegram.
bilibili.endpoints[].notify.endpoint	string	yes	—	Endpoint id of that adapter.
bilibili.endpoints[].notify.kind	string: "private", "group", "channel"	no	"private"	Conversation type.
bilibili.endpoints[].notify.id	string | number	yes	—	Conversation id: your account for a private chat, or the group id.
bilibili.endpoints[].apiBaseUrl	string	no	"https://api.bilibili.com"	—
bilibili.endpoints[].commandPrefix	string	no	—	—
bilibili.endpoints[].nickname	string | null	no	—	Fallback account nickname
bilibili.endpoints[].profile	object	no	—	Fallback account profile
bilibili.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
bilibili.endpoints[].profile.age	integer | null	no	—	—
bilibili.endpoints[].profile.birthday	string | null	no	—	—
bilibili.endpoints[].profile.signature	string | null	no	—	—
bilibili.endpoints[].profile.avatar	string | null	no	—	—
bilibili.endpoints[].profile.interests	array | null	no	—	—
bilibili.endpoints[].profile.bio	string | null	no	—	—
bilibili.nickname	string | null	no	—	Fallback account nickname
bilibili.profile	object	no	—	Fallback account profile
bilibili.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
bilibili.profile.age	integer | null	no	—	—
bilibili.profile.birthday	string | null	no	—	—
bilibili.profile.signature	string | null	no	—	—
bilibili.profile.avatar	string | null	no	—	—
bilibili.profile.interests	array | null	no	—	—
bilibili.profile.bio	string | null	no	—	—

bilibili-live

plugins/adapters/bilibili-live/schema.json

Path	Type	Required	Default	Description
bilibili-live.roomId	string | number	no	—	Live room id (short or full); use your own room only.
bilibili-live.sendMode	string: "direct", "off"	no	"direct"	Outbound mode: direct (default, needs login) posts the reply as danmaku; off only listens.
bilibili-live.maxLength	number	no	20	Maximum characters per danmaku (default 20; higher live levels allow more).
bilibili-live.maxChunks	number	no	2	Maximum danmaku per reply (default 2); the rest is truncated.
bilibili-live.minSendIntervalSec	number	no	5	Minimum seconds between two replies, at least 2; a reply sent sooner is rejected, not queued.
bilibili-live.apiBaseUrl	string	no	"https://api.bilibili.com"	—
bilibili-live.liveApiBaseUrl	string	no	"https://api.live.bilibili.com"	—
bilibili-live.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
bilibili-live.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
bilibili-live.endpoints[].id	string	yes	—	Local endpoint name, not a Bilibili account ID.
bilibili-live.endpoints[].sessdata	string	no	—	SESSDATA cookie, acting as your account. Reference ${BILIBILI_SESSDATA}; never commit it. Omit to listen anonymously.
bilibili-live.endpoints[].biliJct	string	no	—	bili_jct cookie (csrf token) needed to send danmaku. Reference ${BILIBILI_BILI_JCT}.
bilibili-live.endpoints[].buvid3	string	no	—	Optional buvid3 cookie; requested from the platform at startup when omitted.
bilibili-live.endpoints[].roomId	string | number	no	—	Live room id (short or full); use your own room only.
bilibili-live.endpoints[].sendMode	string: "direct", "off"	no	"direct"	Outbound mode: direct (default, needs login) posts the reply as danmaku; off only listens.
bilibili-live.endpoints[].maxLength	number	no	20	Maximum characters per danmaku (default 20; higher live levels allow more).
bilibili-live.endpoints[].maxChunks	number	no	2	Maximum danmaku per reply (default 2); the rest is truncated.
bilibili-live.endpoints[].minSendIntervalSec	number	no	5	Minimum seconds between two replies, at least 2; a reply sent sooner is rejected, not queued.
bilibili-live.endpoints[].apiBaseUrl	string	no	"https://api.bilibili.com"	—
bilibili-live.endpoints[].liveApiBaseUrl	string	no	"https://api.live.bilibili.com"	—
bilibili-live.endpoints[].commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
bilibili-live.endpoints[].nickname	string | null	no	—	Fallback account nickname
bilibili-live.endpoints[].profile	object	no	—	Fallback account profile
bilibili-live.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
bilibili-live.endpoints[].profile.age	integer | null	no	—	—
bilibili-live.endpoints[].profile.birthday	string | null	no	—	—
bilibili-live.endpoints[].profile.signature	string | null	no	—	—
bilibili-live.endpoints[].profile.avatar	string | null	no	—	—
bilibili-live.endpoints[].profile.interests	array | null	no	—	—
bilibili-live.endpoints[].profile.bio	string | null	no	—	—
bilibili-live.nickname	string | null	no	—	Fallback account nickname
bilibili-live.profile	object	no	—	Fallback account profile
bilibili-live.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
bilibili-live.profile.age	integer | null	no	—	—
bilibili-live.profile.birthday	string | null	no	—	—
bilibili-live.profile.signature	string | null	no	—	—
bilibili-live.profile.avatar	string | null	no	—	—
bilibili-live.profile.interests	array | null	no	—	—
bilibili-live.profile.bio	string | null	no	—	—

dingtalk

plugins/adapters/dingtalk/schema.json

Path	Type	Required	Default	Description
dingtalk.apiBaseUrl	string	no	"https://oapi.dingtalk.com"	—
dingtalk.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
dingtalk.endpoints[].appKey	string	yes	—	Dingtalk app key
dingtalk.endpoints[].appSecret	string	yes	—	Dingtalk app secret
dingtalk.endpoints[].webhookPath	string	no	—	Dingtalk webhook path
dingtalk.endpoints[].robotCode	string	no	—	Dingtalk robot code
dingtalk.endpoints[].apiBaseUrl	string	no	—	—
dingtalk.endpoints[].commandPrefix	string	no	—	—
dingtalk.endpoints[].id	string	yes	—	Dingtalk bot name
dingtalk.endpoints[].mode	string: "webhook", "stream"	no	"webhook"	—
dingtalk.endpoints[].cardTemplateId	string	no	—	Stream interactive card template ID
dingtalk.endpoints[].cardButtonCount	integer	no	2	Fixed button count in the card template
dingtalk.endpoints[].streamProxy	object	no	—	Fault-injection tests only: pin TCP to 127.0.0.1 while preserving the actual WSS TLS identity.
dingtalk.endpoints[].streamProxy.port	integer	yes	—	—
dingtalk.endpoints[].streamProxy.serverName	string	yes	—	—
dingtalk.endpoints[].nickname	string | null	no	—	Fallback account nickname
dingtalk.endpoints[].profile	object	no	—	Fallback account profile
dingtalk.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
dingtalk.endpoints[].profile.age	integer | null	no	—	—
dingtalk.endpoints[].profile.birthday	string | null	no	—	—
dingtalk.endpoints[].profile.signature	string | null	no	—	—
dingtalk.endpoints[].profile.avatar	string | null	no	—	—
dingtalk.endpoints[].profile.interests	array | null	no	—	—
dingtalk.endpoints[].profile.bio	string | null	no	—	—
dingtalk.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
dingtalk.mode	string: "webhook", "stream"	no	"webhook"	—
dingtalk.cardTemplateId	string	no	—	Stream interactive card template ID
dingtalk.cardButtonCount	integer	no	2	Fixed button count in the card template
dingtalk.nickname	string | null	no	—	Fallback account nickname
dingtalk.profile	object	no	—	Fallback account profile
dingtalk.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
dingtalk.profile.age	integer | null	no	—	—
dingtalk.profile.birthday	string | null	no	—	—
dingtalk.profile.signature	string | null	no	—	—
dingtalk.profile.avatar	string | null	no	—	—
dingtalk.profile.interests	array | null	no	—	—
dingtalk.profile.bio	string | null	no	—	—

discord

plugins/adapters/discord/schema.json

Path	Type	Required	Default	Description
discord.connection	string: "gateway", "interactions"	no	"gateway"	Gateway WebSocket (default). interactions uses httpHostToken POST + Ed25519 verify.
discord.intents	array<number>	no	—	—
discord.enableSlashCommands	boolean	no	false	—
discord.globalCommands	boolean	no	false	—
discord.defaultActivity	object	no	—	—
discord.defaultActivity.name	string	yes	—	—
discord.defaultActivity.type	string: "PLAYING", "STREAMING", "LISTENING", "WATCHING", "COMPETING"	yes	—	—
discord.defaultActivity.url	string	no	—	—
discord.slashCommands	array<object>	no	—	—
discord.applicationId	string	no	—	Required when connection is interactions.
discord.publicKey	string	no	—	Required when connection is interactions (hex Ed25519 public key).
discord.interactionsPath	string	no	"/discord/interactions"	POST path on httpHostToken when connection is interactions.
discord.endpoints	array<object>	yes	—	Multiple accounts: each entry defines an endpoint. id and token are required; other fields override instance defaults.
discord.endpoints[].token	string	yes	—	Discord bot token
discord.endpoints[].connection	string: "gateway", "interactions"	no	—	—
discord.endpoints[].intents	array<number>	no	—	—
discord.endpoints[].enableSlashCommands	boolean	no	—	—
discord.endpoints[].globalCommands	boolean	no	—	—
discord.endpoints[].defaultActivity	object	no	—	—
discord.endpoints[].defaultActivity.name	string	yes	—	—
discord.endpoints[].defaultActivity.type	string: "PLAYING", "STREAMING", "LISTENING", "WATCHING", "COMPETING"	yes	—	—
discord.endpoints[].defaultActivity.url	string	no	—	—
discord.endpoints[].slashCommands	array<object>	no	—	—
discord.endpoints[].applicationId	string	no	—	—
discord.endpoints[].publicKey	string	no	—	—
discord.endpoints[].interactionsPath	string	no	—	—
discord.endpoints[].commandPrefix	string	no	—	—
discord.endpoints[].id	string	yes	—	Discord bot name
discord.endpoints[].gatewayFaultProxyUrl	string	no	—	Isolated acceptance only: loopback Gateway fault proxy; REST unchanged.
discord.endpoints[].restApiProxy	object	no	—	REST-only fault proxy for Gateway tests: pin discord.com:443 TCP to loopback while preserving SNI and certificate validation, independently of the Gateway proxy.
discord.endpoints[].restApiProxy.port	integer	yes	—	—
discord.endpoints[].nickname	string | null	no	—	Fallback account nickname
discord.endpoints[].profile	object	no	—	Fallback account profile
discord.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
discord.endpoints[].profile.age	integer | null	no	—	—
discord.endpoints[].profile.birthday	string | null	no	—	—
discord.endpoints[].profile.signature	string | null	no	—	—
discord.endpoints[].profile.avatar	string | null	no	—	—
discord.endpoints[].profile.interests	array | null	no	—	—
discord.endpoints[].profile.bio	string | null	no	—	—
discord.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
discord.gatewayFaultProxyUrl	string	no	—	Isolated acceptance only: loopback Gateway fault proxy; REST unchanged.
discord.nickname	string | null	no	—	Fallback account nickname
discord.profile	object	no	—	Fallback account profile
discord.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
discord.profile.age	integer | null	no	—	—
discord.profile.birthday	string | null	no	—	—
discord.profile.signature	string | null	no	—	—
discord.profile.avatar	string | null	no	—	—
discord.profile.interests	array | null	no	—	—
discord.profile.bio	string | null	no	—	—

douyin-live

plugins/adapters/douyin-live/schema.json

Path	Type	Required	Default	Description
douyin-live.sendMode	string: "off", "direct"	no	"off"	Outbound mode: off (default) only listens; direct is not implemented yet and returns not_implemented.
douyin-live.maxLength	number	no	50	Maximum characters per danmaku (default 50; reserved for future direct sending).
douyin-live.maxChunks	number	no	2	Maximum danmaku per reply (default 2).
douyin-live.minSendIntervalSec	number	no	5	Minimum seconds between two replies, at least 2.
douyin-live.signerUrl	string	no	—	Optional external signer (GET ?stub=<X-MS-STUB> returns the signature). The package ships no platform signing script; without it the socket connects unsigned and the platform may refuse.
douyin-live.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
douyin-live.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
douyin-live.endpoints[].id	string	yes	—	Local endpoint name, not a Douyin account ID.
douyin-live.endpoints[].cookie	string	no	—	Full browser cookie string. Reference ${DOUYIN_LIVE_COOKIE}; never commit it. Omit to request a guest ttwid at startup.
douyin-live.endpoints[].roomId	string | integer	yes	—	Live room web_rid (the number after live.douyin.com/); use your own room only.
douyin-live.endpoints[].sendMode	string: "off", "direct"	no	"off"	Outbound mode: off (default) only listens; direct is not implemented yet and returns not_implemented.
douyin-live.endpoints[].maxLength	number	no	50	Maximum characters per danmaku (default 50; reserved for future direct sending).
douyin-live.endpoints[].maxChunks	number	no	2	Maximum danmaku per reply (default 2).
douyin-live.endpoints[].minSendIntervalSec	number	no	5	Minimum seconds between two replies, at least 2.
douyin-live.endpoints[].signerUrl	string	no	—	Optional external signer (GET ?stub=<X-MS-STUB> returns the signature). The package ships no platform signing script; without it the socket connects unsigned and the platform may refuse.
douyin-live.endpoints[].commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
douyin-live.endpoints[].nickname	string | null	no	—	Fallback account nickname
douyin-live.endpoints[].profile	object	no	—	Fallback account profile
douyin-live.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
douyin-live.endpoints[].profile.age	integer | null	no	—	—
douyin-live.endpoints[].profile.birthday	string | null	no	—	—
douyin-live.endpoints[].profile.signature	string | null	no	—	—
douyin-live.endpoints[].profile.avatar	string | null	no	—	—
douyin-live.endpoints[].profile.interests	array | null	no	—	—
douyin-live.endpoints[].profile.bio	string | null	no	—	—
douyin-live.nickname	string | null	no	—	Fallback account nickname
douyin-live.profile	object	no	—	Fallback account profile
douyin-live.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
douyin-live.profile.age	integer | null	no	—	—
douyin-live.profile.birthday	string | null	no	—	—
douyin-live.profile.signature	string | null	no	—	—
douyin-live.profile.avatar	string | null	no	—	—
douyin-live.profile.interests	array | null	no	—	—
douyin-live.profile.bio	string | null	no	—	—

email

plugins/adapters/email/schema.json

Path	Type	Required	Default	Description
email.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
email.endpoints[].smtp	object	yes	—	—
email.endpoints[].smtp.host	string	yes	—	—
email.endpoints[].smtp.port	number	yes	—	—
email.endpoints[].smtp.secure	boolean	yes	—	—
email.endpoints[].smtp.auth	object	yes	—	—
email.endpoints[].smtp.auth.user	string	yes	—	—
email.endpoints[].smtp.auth.pass	string	yes	—	—
email.endpoints[].smtp.serverName	string	no	—	TLS SNI and certificate hostname verification. Preserve the original service hostname when using a proxy; does not change certificate validation policy.
email.endpoints[].imap	object	yes	—	—
email.endpoints[].imap.host	string	yes	—	—
email.endpoints[].imap.port	number	yes	—	—
email.endpoints[].imap.tls	boolean	yes	—	—
email.endpoints[].imap.user	string	yes	—	—
email.endpoints[].imap.password	string	yes	—	—
email.endpoints[].imap.checkInterval	number	no	60000	—
email.endpoints[].imap.reconnectInterval	number	no	5000	—
email.endpoints[].imap.mailbox	string	no	"INBOX"	—
email.endpoints[].imap.markSeen	boolean	no	true	—
email.endpoints[].imap.serverName	string	no	—	TLS SNI and certificate hostname verification. Preserve the original service hostname when using a proxy; does not change certificate validation policy.
email.endpoints[].attachments	object	no	—	—
email.endpoints[].attachments.enabled	boolean	no	false	—
email.endpoints[].attachments.downloadPath	string	no	—	—
email.endpoints[].attachments.maxFileSize	number	no	—	—
email.endpoints[].attachments.allowedTypes	array<string>	no	—	—
email.endpoints[].commandPrefix	string	no	—	—
email.endpoints[].id	string	yes	—	Email bot name
email.endpoints[].nickname	string | null	no	—	Fallback account nickname
email.endpoints[].profile	object	no	—	Fallback account profile
email.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
email.endpoints[].profile.age	integer | null	no	—	—
email.endpoints[].profile.birthday	string | null	no	—	—
email.endpoints[].profile.signature	string | null	no	—	—
email.endpoints[].profile.avatar	string | null	no	—	—
email.endpoints[].profile.interests	array | null	no	—	—
email.endpoints[].profile.bio	string | null	no	—	—
email.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
email.nickname	string | null	no	—	Fallback account nickname
email.profile	object	no	—	Fallback account profile
email.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
email.profile.age	integer | null	no	—	—
email.profile.birthday	string | null	no	—	—
email.profile.signature	string | null	no	—	—
email.profile.avatar	string | null	no	—	—
email.profile.interests	array | null	no	—	—
email.profile.bio	string | null	no	—	—

github

plugins/adapters/github/schema.json

Path	Type	Required	Default	Description
github.host	string	no	—	GitHub Enterprise hostname (default github.com)
github.webhook_path	string	no	"/github/webhook"	—
github.auto_reply_repos	array<string>	no	—	Repos whose Issue/PR comments auto-trigger AI without @bot
github.bot_login	string	no	—	Override App bot login (default {slug}[bot])
github.workspace_root	string	no	—	Managed git workspace root
github.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance.
github.endpoints[].app_id	string | number	yes	—	GitHub App ID
github.endpoints[].private_key	string	yes	—	GitHub App private key (PEM content or file path)
github.endpoints[].webhook_secret	string	no	—	Webhook HMAC secret; enables httpHostToken POST route
github.endpoints[].id	string	yes	—	GitHub App bot name
github.endpoints[].host	string	no	—	GitHub Enterprise hostname
github.endpoints[].webhook_path	string	no	—	Webhook route path
github.endpoints[].auto_reply_repos	array<string>	no	—	Repos whose comments auto-trigger AI
github.endpoints[].bot_login	string	no	—	Override App bot login
github.endpoints[].workspace_root	string	no	—	Managed git workspace root
github.endpoints[].commandPrefix	string	no	—	Command prefix for this endpoint.
github.endpoints[].nickname	string | null	no	—	Fallback account nickname
github.endpoints[].profile	object	no	—	Fallback account profile
github.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
github.endpoints[].profile.age	integer | null	no	—	—
github.endpoints[].profile.birthday	string | null	no	—	—
github.endpoints[].profile.signature	string | null	no	—	—
github.endpoints[].profile.avatar	string | null	no	—	—
github.endpoints[].profile.interests	array | null	no	—	—
github.endpoints[].profile.bio	string | null	no	—	—
github.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
github.nickname	string | null	no	—	Fallback account nickname
github.profile	object	no	—	Fallback account profile
github.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
github.profile.age	integer | null	no	—	—
github.profile.birthday	string | null	no	—	—
github.profile.signature	string | null	no	—	—
github.profile.avatar	string | null	no	—	—
github.profile.interests	array | null	no	—	—
github.profile.bio	string | null	no	—	—

icqq

plugins/adapters/icqq/schema.json

Path	Type	Required	Default	Description
icqq.password	string	no	—	QQ password (optional); scan a QR code to log in when omitted.
icqq.platform	number: 1, 2, 3, 4, 5, 6, 7	no	1	Login device platform: 1=Android, 2=aPad, 3=Watch, 4=iMac, 5=iPad, 6=Tim, 7=Custom.
icqq.ver	string	no	—	Protocol version, only applicable when the selected platform has multiple versions. Uses the latest when omitted.
icqq.signApiAddr	string	no	—	Signing server URL. When omitted, uses local signing automatically if @icqqjs/qqsign is installed.
icqq.dataDir	string	no	—	Data directory path; defaults to the data directory under the main module.
icqq.autoReconnect	boolean	no	true	Reconnect automatically after disconnecting (default true).
icqq.outboundMedia	string: "file", "base64"	no	"file"	Outbound media mode: file uses local file paths (default); base64 uses inline CQ base64:// data.
icqq.ignoreSelf	boolean	no	true	Filter the bot's own messages in groups and channels (default true).
icqq.resend	boolean	no	false	Attempt fragmented sending when rate-limited by the platform (default false).
icqq.cacheGroupMember	boolean	no	true	Cache group member lists (default true). Large numbers of groups (500+) may use approximately 100 MB or more of extra memory.
icqq.autoServer	boolean	no	true	Automatically select the optimal server (default true).
icqq.qqnt	boolean	no	true	Use the QQNT protocol (default true).
icqq.ntLogin	boolean	no	—	Use the NT login method.
icqq.logLevel	string: "trace", "debug", "info", "warn", "error", "fatal", "mark", "off"	no	"warn"	Level of the ICQQ SDK's own log lines (default warn). Its per-message lines with the message text (recv from / succeed to send) only appear at debug or below, escaped onto one line.
icqq.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
icqq.endpoints[].id	string	yes	—	QQ uin
icqq.endpoints[].password	string	no	—	QQ password for this endpoint.
icqq.endpoints[].platform	number: 1, 2, 3, 4, 5, 6, 7	no	—	Login device platform: 1=Android, 2=aPad, 3=Watch, 4=iMac, 5=iPad, 6=Tim, 7=Custom.
icqq.endpoints[].ver	string	no	—	Protocol version.
icqq.endpoints[].signApiAddr	string	no	—	Signing server URL.
icqq.endpoints[].dataDir	string	no	—	Data directory path.
icqq.endpoints[].autoReconnect	boolean	no	—	Reconnect automatically after disconnecting.
icqq.endpoints[].outboundMedia	string: "file", "base64"	no	—	Outbound media mode.
icqq.endpoints[].ignoreSelf	boolean	no	—	Filter the bot's own messages.
icqq.endpoints[].resend	boolean	no	—	Use fragmented sending when rate-limited by the platform.
icqq.endpoints[].cacheGroupMember	boolean	no	—	Cache group member lists.
icqq.endpoints[].autoServer	boolean	no	—	Automatically select the optimal server.
icqq.endpoints[].qqnt	boolean	no	—	Use the QQNT protocol.
icqq.endpoints[].ntLogin	boolean	no	—	Use the NT login method.
icqq.endpoints[].logLevel	string: "trace", "debug", "info", "warn", "error", "fatal", "mark", "off"	no	—	Level of the ICQQ SDK's own log lines.
icqq.endpoints[].commandPrefix	string	no	—	Command prefix for this endpoint.
icqq.endpoints[].notify	object	no	—	Send outage/login notices to the master through another adapter instance (adapter/endpoint/kind/id identify the actual conversation).
icqq.endpoints[].notify.adapter	string	yes	—	Adapter type used for the notice, such as qq or icqq.
icqq.endpoints[].notify.endpoint	string	yes	—	Endpoint id of that adapter.
icqq.endpoints[].notify.kind	string: "private", "group", "channel"	no	"private"	Conversation kind; defaults to private.
icqq.endpoints[].notify.id	string | number	yes	—	Master conversation id on that platform (openid for QQ official, QQ number for ICQQ).
icqq.endpoints[].notify.parent	object	no	—	可选的父频道/群容器
icqq.endpoints[].notify.parent.kind	string: "private", "group", "channel"	yes	—	—
icqq.endpoints[].notify.parent.id	string | number	yes	—	—
icqq.endpoints[].nickname	string | null	no	—	Fallback account nickname
icqq.endpoints[].profile	object	no	—	Fallback account profile
icqq.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
icqq.endpoints[].profile.age	integer | null	no	—	—
icqq.endpoints[].profile.birthday	string | null	no	—	—
icqq.endpoints[].profile.signature	string | null	no	—	—
icqq.endpoints[].profile.avatar	string | null	no	—	—
icqq.endpoints[].profile.interests	array | null	no	—	—
icqq.endpoints[].profile.bio	string | null	no	—	—
icqq.commandPrefix	string	no	""	Command prefix, empty by default. Each endpoints[i] can override it.
icqq.notify	object	no	—	Send outage/login notices to the master through another adapter instance (adapter/endpoint/kind/id identify the actual conversation).
icqq.notify.adapter	string	yes	—	Adapter type used for the notice, such as qq or icqq.
icqq.notify.endpoint	string	yes	—	Endpoint id of that adapter.
icqq.notify.kind	string: "private", "group", "channel"	no	"private"	Conversation kind; defaults to private.
icqq.notify.id	string | number	yes	—	Master conversation id on that platform (openid for QQ official, QQ number for ICQQ).
icqq.notify.parent	object	no	—	可选的父频道/群容器
icqq.notify.parent.kind	string: "private", "group", "channel"	yes	—	—
icqq.notify.parent.id	string | number	yes	—	—
icqq.nickname	string | null	no	—	Fallback account nickname
icqq.profile	object	no	—	Fallback account profile
icqq.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
icqq.profile.age	integer | null	no	—	—
icqq.profile.birthday	string | null	no	—	—
icqq.profile.signature	string | null	no	—	—
icqq.profile.avatar	string | null	no	—	—
icqq.profile.interests	array | null	no	—	—
icqq.profile.bio	string | null	no	—	—

juejin

plugins/adapters/juejin/schema.json

Path	Type	Required	Default	Description
juejin.apiBaseUrl	string	no	"https://api.juejin.cn"	—
juejin.pollIntervalSec	number	no	60	Polling interval in seconds, at least 60; adds 0–20% random jitter.
juejin.sendMode	string: "review", "off"	no	"review"	Outbound mode: review (default) queues AI replies as drafts; a master approves each send on the Console page or with juejin-draft send. off disables outbound messages. Automatic direct replies are not supported.
juejin.minSendIntervalSec	number	no	30	Minimum interval between approved sends in seconds (at least 10). Sending too soon is rejected, not queued.
juejin.dataDir	string	no	"data/juejin"	Local data directory. Pending drafts are stored in <dataDir>/<endpoint id>/drafts.json (mode 600, relative to the working directory) and survive restarts. Keep it out of the repository.
juejin.notify	object	no	—	Chat conversation to notify when a new draft arrives or the login expires (usually your private chat with the bot). Without it, drafts only appear on the Console Juejin comment review page and in logs.
juejin.notify.adapter	string	yes	—	Adapter name, such as icqq or telegram.
juejin.notify.endpoint	string	yes	—	Endpoint id of that adapter.
juejin.notify.kind	string: "private", "group", "channel"	no	"private"	Conversation type.
juejin.notify.id	string | number	yes	—	Conversation id: your account for a private chat, or the group id.
juejin.articles	array<string>	no	—	Allowed articles: article IDs (quoted strings) or https://juejin.cn/post/<id&gt; links. Only these articles are polled; there is no site-wide search. Empty means nothing is polled.
juejin.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
juejin.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
juejin.endpoints[].id	string	yes	—	Local endpoint name, not a Juejin account ID.
juejin.endpoints[].cookie	string	no	—	Full Juejin web login Cookie header (must include sessionid), acting as your account. Reference ${JUEJIN_COOKIE}; never commit it to the repository.
juejin.endpoints[].articles	array<string>	no	—	Allowed articles: article IDs (quoted strings) or https://juejin.cn/post/<id&gt; links. Only these articles are polled; there is no site-wide search. Empty means nothing is polled.
juejin.endpoints[].pollIntervalSec	number	no	60	Polling interval in seconds, at least 60; adds 0–20% random jitter.
juejin.endpoints[].sendMode	string: "review", "off"	no	"review"	Outbound mode: review (default) queues AI replies as drafts; a master approves each send on the Console page or with juejin-draft send. off disables outbound messages. Automatic direct replies are not supported.
juejin.endpoints[].minSendIntervalSec	number	no	30	Minimum interval between approved sends in seconds (at least 10). Sending too soon is rejected, not queued.
juejin.endpoints[].dataDir	string	no	"data/juejin"	Local data directory. Pending drafts are stored in <dataDir>/<endpoint id>/drafts.json (mode 600, relative to the working directory) and survive restarts. Keep it out of the repository.
juejin.endpoints[].notify	object	no	—	Chat conversation to notify when a new draft arrives or the login expires (usually your private chat with the bot). Without it, drafts only appear on the Console Juejin comment review page and in logs.
juejin.endpoints[].notify.adapter	string	yes	—	Adapter name, such as icqq or telegram.
juejin.endpoints[].notify.endpoint	string	yes	—	Endpoint id of that adapter.
juejin.endpoints[].notify.kind	string: "private", "group", "channel"	no	"private"	Conversation type.
juejin.endpoints[].notify.id	string | number	yes	—	Conversation id: your account for a private chat, or the group id.
juejin.endpoints[].apiBaseUrl	string	no	"https://api.juejin.cn"	—
juejin.endpoints[].commandPrefix	string	no	—	—
juejin.endpoints[].nickname	string | null	no	—	Fallback account nickname
juejin.endpoints[].profile	object	no	—	Fallback account profile
juejin.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
juejin.endpoints[].profile.age	integer | null	no	—	—
juejin.endpoints[].profile.birthday	string | null	no	—	—
juejin.endpoints[].profile.signature	string | null	no	—	—
juejin.endpoints[].profile.avatar	string | null	no	—	—
juejin.endpoints[].profile.interests	array | null	no	—	—
juejin.endpoints[].profile.bio	string | null	no	—	—
juejin.nickname	string | null	no	—	Fallback account nickname
juejin.profile	object	no	—	Fallback account profile
juejin.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
juejin.profile.age	integer | null	no	—	—
juejin.profile.birthday	string | null	no	—	—
juejin.profile.signature	string | null	no	—	—
juejin.profile.avatar	string | null	no	—	—
juejin.profile.interests	array | null	no	—	—
juejin.profile.bio	string | null	no	—	—

kook

plugins/adapters/kook/schema.json

Path	Type	Required	Default	Description
kook.connection	string: "websocket", "webhook"	no	"websocket"	WebSocket gateway (default). webhook requires httpHostToken, verify_token, and a public HTTPS callback URL.
kook.webhookPath	string	no	"/kook/webhook"	POST path registered via httpHostToken when connection is webhook.
kook.data_dir	string	no	—	—
kook.timeout	number	no	10000	—
kook.max_retry	number	no	3	—
kook.ignore	string: "bot", "self"	no	"bot"	—
kook.logLevel	string: "trace", "debug", "info", "warn", "error", "fatal", "mark", "off"	no	"info"	—
kook.endpoints	array<object>	yes	—	Multiple accounts: each entry defines an endpoint. id and token are required; other fields override instance defaults.
kook.endpoints[].token	string	yes	—	KOOK bot token
kook.endpoints[].connection	string: "websocket", "webhook"	no	—	—
kook.endpoints[].webhookPath	string	no	—	—
kook.endpoints[].verify_token	string	no	—	KOOK developer console verify token (required for webhook mode).
kook.endpoints[].encrypt_key	string	no	—	Optional Encrypt Key when message encryption is enabled in KOOK console.
kook.endpoints[].data_dir	string	no	—	—
kook.endpoints[].timeout	number	no	—	—
kook.endpoints[].max_retry	number	no	—	—
kook.endpoints[].ignore	string: "bot", "self"	no	—	—
kook.endpoints[].logLevel	string: "trace", "debug", "info", "warn", "error", "fatal", "mark", "off"	no	—	—
kook.endpoints[].commandPrefix	string	no	—	—
kook.endpoints[].id	string	yes	—	KOOK bot name
kook.endpoints[].streamProxy	object	no	—	Pin WSS fault-injection tests to loopback while preserving the actual TLS identity.
kook.endpoints[].streamProxy.port	integer	yes	—	—
kook.endpoints[].streamProxy.serverName	string	yes	—	—
kook.endpoints[].apiProxy	object	no	—	Optional fixed loopback TCP route for www.kookapp.cn JSON API only; verified TLS remains enabled.
kook.endpoints[].apiProxy.port	integer	yes	—	—
kook.endpoints[].nickname	string | null	no	—	Fallback account nickname
kook.endpoints[].profile	object	no	—	Fallback account profile
kook.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
kook.endpoints[].profile.age	integer | null	no	—	—
kook.endpoints[].profile.birthday	string | null	no	—	—
kook.endpoints[].profile.signature	string | null	no	—	—
kook.endpoints[].profile.avatar	string | null	no	—	—
kook.endpoints[].profile.interests	array | null	no	—	—
kook.endpoints[].profile.bio	string | null	no	—	—
kook.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
kook.nickname	string | null	no	—	Fallback account nickname
kook.profile	object	no	—	Fallback account profile
kook.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
kook.profile.age	integer | null	no	—	—
kook.profile.birthday	string | null	no	—	—
kook.profile.signature	string | null	no	—	—
kook.profile.avatar	string | null	no	—	—
kook.profile.interests	array | null	no	—	—
kook.profile.bio	string | null	no	—	—

lark

plugins/adapters/lark/schema.json

Path	Type	Required	Default	Description
lark.webhookPath	string	no	"/lark/webhook"	—
lark.apiBaseUrl	string	no	—	—
lark.isFeishu	boolean	no	true	—
lark.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
lark.endpoints[].appId	string	yes	—	Lark app ID
lark.endpoints[].appSecret	string	yes	—	Lark app secret
lark.endpoints[].encryptKey	string	no	—	Lark encrypt key
lark.endpoints[].verificationToken	string	no	—	Lark verification token
lark.endpoints[].webhookPath	string	no	—	—
lark.endpoints[].apiBaseUrl	string	no	—	—
lark.endpoints[].isFeishu	boolean	no	—	—
lark.endpoints[].commandPrefix	string	no	—	—
lark.endpoints[].id	string	yes	—	Lark bot name
lark.endpoints[].mode	string: "webhook", "websocket"	no	"webhook"	—
lark.endpoints[].streamProxy	object	no	—	WSS fault-injection tests: pin only TCP to loopback while preserving TLS identity.
lark.endpoints[].streamProxy.port	integer	yes	—	—
lark.endpoints[].streamProxy.serverName	string	yes	—	—
lark.endpoints[].webApiProxy	object	no	—	Disabled by default. Loopback TLS fault proxy for Feishu outbound APIs only; preserves the official Host, SNI and certificate validation. Independent of streamProxy.
lark.endpoints[].webApiProxy.port	integer	yes	—	—
lark.endpoints[].streaming	boolean	no	—	Use native streaming cards for progressive replies.
lark.endpoints[].streamingIntervalMs	integer	no	—	Reply stream update interval in milliseconds.
lark.endpoints[].nickname	string | null	no	—	Fallback account nickname
lark.endpoints[].profile	object	no	—	Fallback account profile
lark.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
lark.endpoints[].profile.age	integer | null	no	—	—
lark.endpoints[].profile.birthday	string | null	no	—	—
lark.endpoints[].profile.signature	string | null	no	—	—
lark.endpoints[].profile.avatar	string | null	no	—	—
lark.endpoints[].profile.interests	array | null	no	—	—
lark.endpoints[].profile.bio	string | null	no	—	—
lark.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
lark.mode	string: "webhook", "websocket"	no	"webhook"	—
lark.webApiProxy	object	no	—	Disabled by default. Loopback TLS fault proxy for Feishu outbound APIs only; preserves the official Host, SNI and certificate validation. Independent of streamProxy.
lark.webApiProxy.port	integer	yes	—	—
lark.streaming	boolean	no	true	Use native streaming cards for progressive replies.
lark.streamingIntervalMs	integer	no	1000	Reply stream update interval in milliseconds.
lark.nickname	string | null	no	—	Fallback account nickname
lark.profile	object	no	—	Fallback account profile
lark.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
lark.profile.age	integer | null	no	—	—
lark.profile.birthday	string | null	no	—	—
lark.profile.signature	string | null	no	—	—
lark.profile.avatar	string | null	no	—	—
lark.profile.interests	array | null	no	—	—
lark.profile.bio	string | null	no	—	—

line

plugins/adapters/line/schema.json

Path	Type	Required	Default	Description
line.webhookPath	string	no	"/line/webhook"	—
line.apiBaseUrl	string	no	"https://api.line.me"	—
line.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
line.endpoints[].channelSecret	string	yes	—	LINE channel secret
line.endpoints[].channelAccessToken	string	yes	—	LINE channel access token
line.endpoints[].webhookPath	string	no	—	—
line.endpoints[].apiBaseUrl	string	no	—	—
line.endpoints[].commandPrefix	string	no	—	—
line.endpoints[].id	string	yes	—	LINE bot name
line.endpoints[].nickname	string | null	no	—	Fallback account nickname
line.endpoints[].profile	object	no	—	Fallback account profile
line.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
line.endpoints[].profile.age	integer | null	no	—	—
line.endpoints[].profile.birthday	string | null	no	—	—
line.endpoints[].profile.signature	string | null	no	—	—
line.endpoints[].profile.avatar	string | null	no	—	—
line.endpoints[].profile.interests	array | null	no	—	—
line.endpoints[].profile.bio	string | null	no	—	—
line.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
line.nickname	string | null	no	—	Fallback account nickname
line.profile	object	no	—	Fallback account profile
line.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
line.profile.age	integer | null	no	—	—
line.profile.birthday	string | null	no	—	—
line.profile.signature	string | null	no	—	—
line.profile.avatar	string | null	no	—	—
line.profile.interests	array | null	no	—	—
line.profile.bio	string | null	no	—	—

milky

plugins/adapters/milky/schema.json

Path	Type	Required	Default	Description
milky.connection	string: "ws", "sse", "webhook", "wss"	no	"ws"	ws (default), sse (EventSource client), webhook or wss (webhook/wss via httpHostToken)
milky.reconnect_interval	number	no	5000	—
milky.heartbeat_interval	number	no	30000	—
milky.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
milky.endpoints[].connection	string: "ws", "sse", "webhook", "wss"	no	—	—
milky.endpoints[].baseUrl	string	yes	—	Milky HTTP API base URL (required); WS event path is derived as ws(s)😕/host/event
milky.endpoints[].path	string	no	—	Path for webhook / reverse-wss
milky.endpoints[].access_token	string	no	—	Milky access token
milky.endpoints[].reconnect_interval	number	no	—	—
milky.endpoints[].heartbeat_interval	number	no	—	—
milky.endpoints[].commandPrefix	string	no	—	—
milky.endpoints[].id	string	yes	—	Milky bot name
milky.endpoints[].nickname	string | null	no	—	Fallback account nickname
milky.endpoints[].profile	object	no	—	Fallback account profile
milky.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
milky.endpoints[].profile.age	integer | null	no	—	—
milky.endpoints[].profile.birthday	string | null	no	—	—
milky.endpoints[].profile.signature	string | null	no	—	—
milky.endpoints[].profile.avatar	string | null	no	—	—
milky.endpoints[].profile.interests	array | null	no	—	—
milky.endpoints[].profile.bio	string | null	no	—	—
milky.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
milky.nickname	string | null	no	—	Fallback account nickname
milky.profile	object	no	—	Fallback account profile
milky.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
milky.profile.age	integer | null	no	—	—
milky.profile.birthday	string | null	no	—	—
milky.profile.signature	string | null	no	—	—
milky.profile.avatar	string | null	no	—	—
milky.profile.interests	array | null	no	—	—
milky.profile.bio	string | null	no	—	—

napcat

plugins/adapters/napcat/schema.json

Path	Type	Required	Default	Description
napcat.connection	string: "ws", "wss", "http"	no	"ws"	ws (default), wss (reverse WS), or http (POST webhook + HTTP API outbound)
napcat.reconnect_interval	number	no	5000	—
napcat.heartbeat_interval	number	no	30000	—
napcat.poll_interval	number	no	30000	—
napcat.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
napcat.endpoints[].connection	string: "ws", "wss", "http"	no	—	—
napcat.endpoints[].url	string	no	—	NapCat WebSocket URL (required for connection: ws)
napcat.endpoints[].path	string	no	—	WS path for reverse-wss
napcat.endpoints[].http_url	string	no	—	HTTP API base URL (connection: http outbound)
napcat.endpoints[].post_path	string	no	—	HTTP POST event path (connection: http inbound)
napcat.endpoints[].access_token	string	no	—	NapCat access token
napcat.endpoints[].reconnect_interval	number	no	—	—
napcat.endpoints[].heartbeat_interval	number	no	—	—
napcat.endpoints[].poll_interval	number	no	—	—
napcat.endpoints[].commandPrefix	string	no	—	—
napcat.endpoints[].id	string	yes	—	NapCat bot name
napcat.endpoints[].nickname	string | null	no	—	Fallback account nickname
napcat.endpoints[].profile	object	no	—	Fallback account profile
napcat.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
napcat.endpoints[].profile.age	integer | null	no	—	—
napcat.endpoints[].profile.birthday	string | null	no	—	—
napcat.endpoints[].profile.signature	string | null	no	—	—
napcat.endpoints[].profile.avatar	string | null	no	—	—
napcat.endpoints[].profile.interests	array | null	no	—	—
napcat.endpoints[].profile.bio	string | null	no	—	—
napcat.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
napcat.nickname	string | null	no	—	Fallback account nickname
napcat.profile	object	no	—	Fallback account profile
napcat.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
napcat.profile.age	integer | null	no	—	—
napcat.profile.birthday	string | null	no	—	—
napcat.profile.signature	string | null	no	—	—
napcat.profile.avatar	string | null	no	—	—
napcat.profile.interests	array | null	no	—	—
napcat.profile.bio	string | null	no	—	—

onebot11

plugins/adapters/onebot11/schema.json

Path	Type	Required	Default	Description
onebot11.connection	string: "ws", "wss"	no	"ws"	connection: ws (forward WS client, default) or wss (reverse WS via httpHostToken)
onebot11.reconnect_interval	number	no	5000	—
onebot11.heartbeat_interval	number	no	30000	—
onebot11.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
onebot11.endpoints[].connection	string: "ws", "wss"	no	—	—
onebot11.endpoints[].url	string	no	—	OneBot implementation WebSocket URL (required for connection: ws)
onebot11.endpoints[].path	string	no	—	WS path for reverse-wss (connection: wss)
onebot11.endpoints[].access_token	string	no	—	OneBot access token
onebot11.endpoints[].reconnect_interval	number	no	—	—
onebot11.endpoints[].heartbeat_interval	number	no	—	—
onebot11.endpoints[].commandPrefix	string	no	—	—
onebot11.endpoints[].id	string	yes	—	OneBot11 bot name
onebot11.endpoints[].nickname	string | null	no	—	Fallback account nickname
onebot11.endpoints[].profile	object	no	—	Fallback account profile
onebot11.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
onebot11.endpoints[].profile.age	integer | null	no	—	—
onebot11.endpoints[].profile.birthday	string | null	no	—	—
onebot11.endpoints[].profile.signature	string | null	no	—	—
onebot11.endpoints[].profile.avatar	string | null	no	—	—
onebot11.endpoints[].profile.interests	array | null	no	—	—
onebot11.endpoints[].profile.bio	string | null	no	—	—
onebot11.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
onebot11.nickname	string | null	no	—	Fallback account nickname
onebot11.profile	object	no	—	Fallback account profile
onebot11.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
onebot11.profile.age	integer | null	no	—	—
onebot11.profile.birthday	string | null	no	—	—
onebot11.profile.signature	string | null	no	—	—
onebot11.profile.avatar	string | null	no	—	—
onebot11.profile.interests	array | null	no	—	—
onebot11.profile.bio	string | null	no	—	—

onebot12

plugins/adapters/onebot12/schema.json

Path	Type	Required	Default	Description
onebot12.connection	string: "ws", "webhook", "wss"	no	"ws"	ws (default), webhook (httpHostToken POST), or wss (reverse WS via httpHostToken)
onebot12.reconnect_interval	number	no	5000	—
onebot12.heartbeat_interval	number	no	30000	—
onebot12.endpoints	array<object>	yes	—	Each endpoint explicitly selects platform and userId. id is the local account name; other connection fields override instance defaults.
onebot12.endpoints[].connection	string: "ws", "webhook", "wss"	no	—	—
onebot12.endpoints[].url	string	no	—	OneBot implementation WebSocket URL (required for connection: ws)
onebot12.endpoints[].path	string	no	—	HTTP/WS path for webhook or reverse-wss
onebot12.endpoints[].api_url	string	no	—	HTTP action endpoint for webhook outbound (required for connection: webhook send)
onebot12.endpoints[].access_token	string	no	—	OneBot access token
onebot12.endpoints[].reconnect_interval	number	no	—	—
onebot12.endpoints[].heartbeat_interval	number	no	—	—
onebot12.endpoints[].commandPrefix	string	no	—	—
onebot12.endpoints[].id	string	yes	—	OneBot12 bot name
onebot12.endpoints[].platform	string	yes	—	Selected OneBot12 bot platform
onebot12.endpoints[].userId	string	yes	—	Selected OneBot12 bot user ID
onebot12.endpoints[].nickname	string | null	no	—	Fallback account nickname
onebot12.endpoints[].profile	object	no	—	Fallback account profile
onebot12.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
onebot12.endpoints[].profile.age	integer | null	no	—	—
onebot12.endpoints[].profile.birthday	string | null	no	—	—
onebot12.endpoints[].profile.signature	string | null	no	—	—
onebot12.endpoints[].profile.avatar	string | null	no	—	—
onebot12.endpoints[].profile.interests	array | null	no	—	—
onebot12.endpoints[].profile.bio	string | null	no	—	—
onebot12.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
onebot12.nickname	string | null	no	—	Fallback account nickname
onebot12.profile	object	no	—	Fallback account profile
onebot12.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
onebot12.profile.age	integer | null	no	—	—
onebot12.profile.birthday	string | null	no	—	—
onebot12.profile.signature	string | null	no	—	—
onebot12.profile.avatar	string | null	no	—	—
onebot12.profile.interests	array | null	no	—	—
onebot12.profile.bio	string | null	no	—	—

qq

plugins/adapters/qq/schema.json

Path	Type	Required	Default	Description
qq.mode	string: "websocket", "webhook", "middleware"	no	"websocket"	WebSocket gateway (default). webhook/middleware use httpHostToken POST.
qq.sandbox	boolean	no	false	—
qq.botKind	string: "public", "private"	no	"public"	Public/private bot type. Both include GROUP_AND_C2C_EVENT; channel messages use PUBLIC_GUILD_MESSAGES or GUILD_MESSAGES respectively. Used to derive intents when they are not explicitly configured.
qq.intents	array<string>	no	—	WebSocket Identify intents. When omitted, derived from botKind; public includes PUBLIC_GUILD_MESSAGES.
qq.accessTokenUrl	string	no	—	—
qq.gatewayUrl	string	no	—	—
qq.webhookPath	string	no	"/qq/webhook"	POST path on httpHostToken for webhook/middleware modes.
qq.endpoints	array<object>	yes	—	Multiple accounts: each entry defines an endpoint. id, appid and secret are required; other fields override instance defaults.
qq.endpoints[].appid	string	yes	—	QQ official bot app ID
qq.endpoints[].secret	string	yes	—	QQ official bot secret
qq.endpoints[].mode	string: "websocket", "webhook", "middleware"	no	—	—
qq.endpoints[].sandbox	boolean	no	—	—
qq.endpoints[].botKind	string: "public", "private"	no	—	Override top-level botKind (public/private) when deriving this endpoint's intents.
qq.endpoints[].intents	array<string>	no	—	Override top-level intents.
qq.endpoints[].accessTokenUrl	string	no	—	—
qq.endpoints[].gatewayUrl	string	no	—	—
qq.endpoints[].webhookPath	string	no	—	—
qq.endpoints[].commandPrefix	string	no	—	—
qq.endpoints[].id	string	yes	—	QQ bot name
qq.endpoints[].streamProxy	object	no	—	Optional fixed loopback TCP fault proxy for WSS only; original gateway TLS identity must match serverName.
qq.endpoints[].streamProxy.port	integer	yes	—	—
qq.endpoints[].streamProxy.serverName	string	yes	—	—
qq.endpoints[].notify	object	no	—	Send outage/login notices to the master through another adapter instance (adapter/endpoint/kind/id identify the actual conversation).
qq.endpoints[].notify.adapter	string	yes	—	Adapter type used for the notice, such as qq or icqq.
qq.endpoints[].notify.endpoint	string	yes	—	Endpoint id of that adapter.
qq.endpoints[].notify.kind	string: "private", "group", "channel"	no	"private"	Conversation kind; defaults to private.
qq.endpoints[].notify.id	string | number	yes	—	Master conversation id on that platform (openid for QQ official, QQ number for ICQQ).
qq.endpoints[].notify.parent	object	no	—	可选的父频道/群容器
qq.endpoints[].notify.parent.kind	string: "private", "group", "channel"	yes	—	—
qq.endpoints[].notify.parent.id	string | number	yes	—	—
qq.endpoints[].streaming	boolean	no	—	Publish supported replies while generating. Set false to send only complete replies.
qq.endpoints[].streamingIntervalMs	number	no	—	Minimum delay in milliseconds between reply updates; the first update is immediate.
qq.endpoints[].nickname	string | null	no	—	Fallback account nickname
qq.endpoints[].profile	object	no	—	Fallback account profile
qq.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
qq.endpoints[].profile.age	integer | null	no	—	—
qq.endpoints[].profile.birthday	string | null	no	—	—
qq.endpoints[].profile.signature	string | null	no	—	—
qq.endpoints[].profile.avatar	string | null	no	—	—
qq.endpoints[].profile.interests	array | null	no	—	—
qq.endpoints[].profile.bio	string | null	no	—	—
qq.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
qq.streamProxy	object	no	—	Optional fixed loopback TCP fault proxy for WSS only; original gateway TLS identity must match serverName.
qq.streamProxy.port	integer	yes	—	—
qq.streamProxy.serverName	string	yes	—	—
qq.notify	object	no	—	Send outage/login notices to the master through another adapter instance (adapter/endpoint/kind/id identify the actual conversation).
qq.notify.adapter	string	yes	—	Adapter type used for the notice, such as qq or icqq.
qq.notify.endpoint	string	yes	—	Endpoint id of that adapter.
qq.notify.kind	string: "private", "group", "channel"	no	"private"	Conversation kind; defaults to private.
qq.notify.id	string | number	yes	—	Master conversation id on that platform (openid for QQ official, QQ number for ICQQ).
qq.notify.parent	object	no	—	可选的父频道/群容器
qq.notify.parent.kind	string: "private", "group", "channel"	yes	—	—
qq.notify.parent.id	string | number	yes	—	—
qq.streaming	boolean	no	true	Publish supported replies while generating. Set false to send only complete replies.
qq.streamingIntervalMs	number	no	500	Minimum delay in milliseconds between reply updates; the first update is immediate.
qq.nickname	string | null	no	—	Fallback account nickname
qq.profile	object	no	—	Fallback account profile
qq.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
qq.profile.age	integer | null	no	—	—
qq.profile.birthday	string | null	no	—	—
qq.profile.signature	string | null	no	—	—
qq.profile.avatar	string | null	no	—	—
qq.profile.interests	array | null	no	—	—
qq.profile.bio	string | null	no	—	—

sandbox

plugins/adapters/sandbox/schema.json

Path	Type	Required	Default	Description
sandbox.streaming	boolean	no	true	Publish supported replies while generating. Set false to send only complete replies.
sandbox.streamingIntervalMs	number	no	500	Minimum delay in milliseconds between reply updates; the first update is immediate.
sandbox.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
sandbox.endpoints[].owner	string	no	—	Sandbox owner user ID
sandbox.endpoints[].streaming	boolean	no	—	Publish supported replies while generating. Set false to send only complete replies.
sandbox.endpoints[].streamingIntervalMs	number	no	—	Minimum delay in milliseconds between reply updates; the first update is immediate.
sandbox.endpoints[].commandPrefix	string	no	—	—
sandbox.endpoints[].id	string	yes	—	Sandbox bot name
sandbox.endpoints[].nickname	string | null	no	—	Fallback account nickname
sandbox.endpoints[].profile	object	no	—	Fallback account profile
sandbox.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
sandbox.endpoints[].profile.age	integer | null	no	—	—
sandbox.endpoints[].profile.birthday	string | null	no	—	—
sandbox.endpoints[].profile.signature	string | null	no	—	—
sandbox.endpoints[].profile.avatar	string | null	no	—	—
sandbox.endpoints[].profile.interests	array | null	no	—	—
sandbox.endpoints[].profile.bio	string | null	no	—	—
sandbox.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
sandbox.nickname	string | null	no	—	Fallback account nickname
sandbox.profile	object	no	—	Fallback account profile
sandbox.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
sandbox.profile.age	integer | null	no	—	—
sandbox.profile.birthday	string | null	no	—	—
sandbox.profile.signature	string | null	no	—	—
sandbox.profile.avatar	string | null	no	—	—
sandbox.profile.interests	array | null	no	—	—
sandbox.profile.bio	string | null	no	—	—

satori

plugins/adapters/satori/schema.json

Path	Type	Required	Default	Description
satori.connection	string: "ws", "webhook"	no	"ws"	ws (default) or webhook (httpHostToken POST route)
satori.heartbeat_interval	number	no	10000	WS PING interval in milliseconds
satori.endpoints	array<object>	yes	—	Multiple accounts: each entry selects an explicit platform and userId. id and baseUrl are required; remaining fields override instance defaults.
satori.endpoints[].baseUrl	string	yes	—	Satori SDK HTTP/WS base URL (e.g. http://127.0.0.1:5140)
satori.endpoints[].connection	string: "ws", "webhook"	no	—	—
satori.endpoints[].token	string	no	—	Bearer token for API and WS IDENTIFY
satori.endpoints[].path	string	no	—	Webhook POST path (connection: webhook)
satori.endpoints[].heartbeat_interval	number	no	—	—
satori.endpoints[].commandPrefix	string	no	—	—
satori.endpoints[].id	string	yes	—	Satori bot name
satori.endpoints[].platform	string	yes	—	Selected Satori login platform
satori.endpoints[].userId	string	yes	—	Selected Satori login user ID
satori.endpoints[].nickname	string | null	no	—	Fallback account nickname
satori.endpoints[].profile	object	no	—	Fallback account profile
satori.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
satori.endpoints[].profile.age	integer | null	no	—	—
satori.endpoints[].profile.birthday	string | null	no	—	—
satori.endpoints[].profile.signature	string | null	no	—	—
satori.endpoints[].profile.avatar	string | null	no	—	—
satori.endpoints[].profile.interests	array | null	no	—	—
satori.endpoints[].profile.bio	string | null	no	—	—
satori.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
satori.nickname	string | null	no	—	Fallback account nickname
satori.profile	object	no	—	Fallback account profile
satori.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
satori.profile.age	integer | null	no	—	—
satori.profile.birthday	string | null	no	—	—
satori.profile.signature	string | null	no	—	—
satori.profile.avatar	string | null	no	—	—
satori.profile.interests	array | null	no	—	—
satori.profile.bio	string | null	no	—	—

slack

plugins/adapters/slack/schema.json

Path	Type	Required	Default	Description
slack.socketMode	boolean	no	true	Prefer Socket Mode (default). Set false for HTTP Events via httpHostToken.
slack.webhookPath	string	no	"/slack/events"	—
slack.clientPingTimeout	number	no	15000	Socket Mode client ping timeout (ms)
slack.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
slack.endpoints[].token	string	yes	—	Bot User OAuth Token (xoxb-...)
slack.endpoints[].signingSecret	string	no	—	Required for HTTP Events API (socketMode: false)
slack.endpoints[].appToken	string	no	—	App-Level Token (xapp-...) for Socket Mode
slack.endpoints[].socketMode	boolean	no	—	—
slack.endpoints[].webhookPath	string	no	—	—
slack.endpoints[].clientPingTimeout	number	no	—	—
slack.endpoints[].commandPrefix	string	no	—	—
slack.endpoints[].id	string	yes	—	Slack bot name
slack.endpoints[].streamProxy	object	no	—	Socket fault-injection tests: pin WSS TCP to loopback while API requests retain direct standard TLS.
slack.endpoints[].streamProxy.port	integer	yes	—	—
slack.endpoints[].streamProxy.serverName	string	yes	—	—
slack.endpoints[].webApiProxy	object	no	—	Fault proxy for outbound JSON Web APIs only: pin slack.com:443 TCP to loopback while preserving SNI and certificate validation. Independent of the Socket proxy; does not apply to upload hosts.
slack.endpoints[].webApiProxy.port	integer	yes	—	—
slack.endpoints[].streaming	boolean	no	—	Use native streaming replies when the target supports them.
slack.endpoints[].streamingIntervalMs	integer	no	—	Reply stream update interval in milliseconds.
slack.endpoints[].nickname	string | null	no	—	Fallback account nickname
slack.endpoints[].profile	object	no	—	Fallback account profile
slack.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
slack.endpoints[].profile.age	integer | null	no	—	—
slack.endpoints[].profile.birthday	string | null	no	—	—
slack.endpoints[].profile.signature	string | null	no	—	—
slack.endpoints[].profile.avatar	string | null	no	—	—
slack.endpoints[].profile.interests	array | null	no	—	—
slack.endpoints[].profile.bio	string | null	no	—	—
slack.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
slack.streaming	boolean	no	true	Use native streaming replies when the target supports them.
slack.streamingIntervalMs	integer	no	1000	Reply stream update interval in milliseconds.
slack.nickname	string | null	no	—	Fallback account nickname
slack.profile	object	no	—	Fallback account profile
slack.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
slack.profile.age	integer | null	no	—	—
slack.profile.birthday	string | null	no	—	—
slack.profile.signature	string | null	no	—	—
slack.profile.avatar	string | null	no	—	—
slack.profile.interests	array | null	no	—	—
slack.profile.bio	string | null	no	—	—

telegram

plugins/adapters/telegram/schema.json

Path	Type	Required	Default	Description
telegram.polling	boolean	no	true	Long-poll getUpdates (default). Set false for webhook via httpHostToken.
telegram.webhook	object	no	—	Webhook settings when polling is false.
telegram.webhook.domain	string	no	—	—
telegram.webhook.path	string	no	"/telegram/webhook"	—
telegram.webhook.secretToken	string	no	—	Verified via X-Telegram-Bot-Api-Secret-Token.
telegram.allowedUpdates	array<string>	no	["message","callback_query"]	—
telegram.apiBaseUrl	string	no	"https://api.telegram.org"	—
telegram.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
telegram.endpoints[].token	string	yes	—	Telegram bot token
telegram.endpoints[].polling	boolean	no	—	—
telegram.endpoints[].webhook	object	no	—	—
telegram.endpoints[].webhook.domain	string	no	—	—
telegram.endpoints[].webhook.path	string	no	—	—
telegram.endpoints[].webhook.secretToken	string	no	—	—
telegram.endpoints[].allowedUpdates	array<string>	no	—	—
telegram.endpoints[].apiBaseUrl	string	no	—	—
telegram.endpoints[].commandPrefix	string	no	—	—
telegram.endpoints[].id	string	yes	—	Telegram bot name
telegram.endpoints[].streaming	boolean	no	—	Publish supported replies while generating. Set false to send only complete replies.
telegram.endpoints[].streamingIntervalMs	number	no	—	Minimum delay in milliseconds between reply updates; the first update is immediate.
telegram.endpoints[].nickname	string | null	no	—	Fallback account nickname
telegram.endpoints[].profile	object	no	—	Fallback account profile
telegram.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
telegram.endpoints[].profile.age	integer | null	no	—	—
telegram.endpoints[].profile.birthday	string | null	no	—	—
telegram.endpoints[].profile.signature	string | null	no	—	—
telegram.endpoints[].profile.avatar	string | null	no	—	—
telegram.endpoints[].profile.interests	array | null	no	—	—
telegram.endpoints[].profile.bio	string | null	no	—	—
telegram.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
telegram.streaming	boolean	no	true	Publish supported replies while generating. Set false to send only complete replies.
telegram.streamingIntervalMs	number	no	500	Minimum delay in milliseconds between reply updates; the first update is immediate.
telegram.nickname	string | null	no	—	Fallback account nickname
telegram.profile	object	no	—	Fallback account profile
telegram.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
telegram.profile.age	integer | null	no	—	—
telegram.profile.birthday	string | null	no	—	—
telegram.profile.signature	string | null	no	—	—
telegram.profile.avatar	string | null	no	—	—
telegram.profile.interests	array | null	no	—	—
telegram.profile.bio	string | null	no	—	—

wechat-mp

plugins/adapters/wechat-mp/schema.json

Path	Type	Required	Default	Description
wechat-mp.path	string	no	"/wechat/webhook"	—
wechat-mp.encrypt	boolean	no	false	—
wechat-mp.encryptMode	string: "plain", "compatible", "secure"	no	—	Default follows encrypt: 'plain' when encrypt=false, 'compatible' when encrypt=true
wechat-mp.replyMode	string: "passive", "customer_service"	no	"passive"	—
wechat-mp.passiveReplyTimeoutMs	number	no	4500	—
wechat-mp.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
wechat-mp.endpoints[].appId	string	yes	—	WeChat MP app ID
wechat-mp.endpoints[].appSecret	string	yes	—	WeChat MP app secret
wechat-mp.endpoints[].token	string	yes	—	WeChat MP callback token
wechat-mp.endpoints[].encodingAESKey	string	no	—	WeChat MP encoding AES key
wechat-mp.endpoints[].path	string	no	—	—
wechat-mp.endpoints[].encrypt	boolean	no	—	—
wechat-mp.endpoints[].encryptMode	string: "plain", "compatible", "secure"	no	—	—
wechat-mp.endpoints[].replyMode	string: "passive", "customer_service"	no	—	—
wechat-mp.endpoints[].passiveReplyTimeoutMs	number	no	—	—
wechat-mp.endpoints[].commandPrefix	string	no	—	—
wechat-mp.endpoints[].id	string	yes	—	WeChat MP bot name
wechat-mp.endpoints[].nickname	string | null	no	—	Fallback account nickname
wechat-mp.endpoints[].profile	object	no	—	Fallback account profile
wechat-mp.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
wechat-mp.endpoints[].profile.age	integer | null	no	—	—
wechat-mp.endpoints[].profile.birthday	string | null	no	—	—
wechat-mp.endpoints[].profile.signature	string | null	no	—	—
wechat-mp.endpoints[].profile.avatar	string | null	no	—	—
wechat-mp.endpoints[].profile.interests	array | null	no	—	—
wechat-mp.endpoints[].profile.bio	string | null	no	—	—
wechat-mp.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
wechat-mp.nickname	string | null	no	—	Fallback account nickname
wechat-mp.profile	object	no	—	Fallback account profile
wechat-mp.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
wechat-mp.profile.age	integer | null	no	—	—
wechat-mp.profile.birthday	string | null	no	—	—
wechat-mp.profile.signature	string | null	no	—	—
wechat-mp.profile.avatar	string | null	no	—	—
wechat-mp.profile.interests	array | null	no	—	—
wechat-mp.profile.bio	string | null	no	—	—

wecom

plugins/adapters/wecom/schema.json

Path	Type	Required	Default	Description
wecom.webhookPath	string	no	"/wecom/callback"	—
wecom.apiBaseUrl	string	no	"https://qyapi.weixin.qq.com"	—
wecom.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
wecom.endpoints[].corpId	string	yes	—	WeCom corp ID
wecom.endpoints[].agentSecret	string	yes	—	WeCom agent secret
wecom.endpoints[].token	string	yes	—	WeCom callback token
wecom.endpoints[].encodingAESKey	string	yes	—	WeCom encoding AES key
wecom.endpoints[].webhookPath	string	no	—	—
wecom.endpoints[].apiBaseUrl	string	no	—	—
wecom.endpoints[].commandPrefix	string	no	—	—
wecom.endpoints[].id	string	yes	—	WeCom bot name
wecom.endpoints[].agentId	string	yes	—	WeCom application AgentId (not the application secret)
wecom.endpoints[].nickname	string | null	no	—	Fallback account nickname
wecom.endpoints[].profile	object	no	—	Fallback account profile
wecom.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
wecom.endpoints[].profile.age	integer | null	no	—	—
wecom.endpoints[].profile.birthday	string | null	no	—	—
wecom.endpoints[].profile.signature	string | null	no	—	—
wecom.endpoints[].profile.avatar	string | null	no	—	—
wecom.endpoints[].profile.interests	array | null	no	—	—
wecom.endpoints[].profile.bio	string | null	no	—	—
wecom.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
wecom.nickname	string | null	no	—	Fallback account nickname
wecom.profile	object	no	—	Fallback account profile
wecom.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
wecom.profile.age	integer | null	no	—	—
wecom.profile.birthday	string | null	no	—	—
wecom.profile.signature	string | null	no	—	—
wecom.profile.avatar	string | null	no	—	—
wecom.profile.interests	array | null	no	—	—
wecom.profile.bio	string | null	no	—	—

weixin-ilink

plugins/adapters/weixin-ilink/schema.json

Path	Type	Required	Default	Description
weixin-ilink.botAgent	string	no	—	—
weixin-ilink.baseUrl	string	no	"https://ilinkai.weixin.qq.com"	—
weixin-ilink.cdnBaseUrl	string	no	"https://novac2c.cdn.weixin.qq.com/c2c"	—
weixin-ilink.longPollTimeoutMs	number	no	35000	—
weixin-ilink.dataDir	string	no	"data/weixin-ilink"	Endpoint state root for credentials, cursors, context tokens, and media
weixin-ilink.endpoints	array<object>	yes	—	Multiple accounts: mount multiple endpoints in one plugin instance. Each entry uses the top-level fields; id is required and other fields override defaults.
weixin-ilink.endpoints[].botToken	string	no	—	iLink bot token; omit to load endpoint state or start QR login
weixin-ilink.endpoints[].id	string	yes	—	Weixin iLink bot name
weixin-ilink.endpoints[].botAgent	string	no	—	—
weixin-ilink.endpoints[].baseUrl	string	no	—	—
weixin-ilink.endpoints[].cdnBaseUrl	string	no	—	—
weixin-ilink.endpoints[].longPollTimeoutMs	number	no	—	—
weixin-ilink.endpoints[].dataDir	string	no	—	Endpoint state root
weixin-ilink.endpoints[].commandPrefix	string	no	—	Command prefix for this endpoint.
weixin-ilink.endpoints[].nickname	string | null	no	—	Fallback account nickname
weixin-ilink.endpoints[].profile	object	no	—	Fallback account profile
weixin-ilink.endpoints[].profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
weixin-ilink.endpoints[].profile.age	integer | null	no	—	—
weixin-ilink.endpoints[].profile.birthday	string | null	no	—	—
weixin-ilink.endpoints[].profile.signature	string | null	no	—	—
weixin-ilink.endpoints[].profile.avatar	string | null	no	—	—
weixin-ilink.endpoints[].profile.interests	array | null	no	—	—
weixin-ilink.endpoints[].profile.bio	string | null	no	—	—
weixin-ilink.commandPrefix	string	no	""	Command prefix. The default empty string allows unprefixed commands; '/' requires a leading slash. Each endpoints[i] can override it.
weixin-ilink.nickname	string | null	no	—	Fallback account nickname
weixin-ilink.profile	object	no	—	Fallback account profile
weixin-ilink.profile.gender	string | null: "male", "female", "other", "unknown", null	no	—	—
weixin-ilink.profile.age	integer | null	no	—	—
weixin-ilink.profile.birthday	string | null	no	—	—
weixin-ilink.profile.signature	string | null	no	—	—
weixin-ilink.profile.avatar	string | null	no	—	—
weixin-ilink.profile.interests	array | null	no	—	—
weixin-ilink.profile.bio	string | null	no	—	—

process-monitor

plugins/features/process-monitor/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
process-monitor.enabled	boolean	no	true	Enable process monitoring.
process-monitor.notifyOnStart	boolean	no	true	Notify on first startup.
process-monitor.notifyOnRestart	boolean	no	true	Notify on normal restart.
process-monitor.notifyOnCrash	boolean	no	true	Notify after a crash restart.
process-monitor.notifyChannels	array<object>	no	[]	Notification channel (slice 1 supports webhook only).
process-monitor.notifyChannels[].type	string: "webhook"	yes	—	Only webhook is supported for now.
process-monitor.notifyChannels[].target	string	yes	—	Webhook URL.

blackjack

plugins/games/blackjack/schema.json

This Schema declares no fields.

dice-duel

plugins/games/dice-duel/schema.json

This Schema declares no fields.

dungeon-expedition

plugins/games/dungeon-expedition/schema.json

This Schema declares no fields.

guess-number

plugins/games/guess-number/schema.json

This Schema declares no fields.

hub

plugins/games/hub/schema.json

This Schema declares no fields.

idiom-chain

plugins/games/idiom-chain/schema.json

This Schema declares no fields.

roulette

plugins/games/roulette/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
roulette.chambers	integer	no	6	Number of chambers; one of them holds the live round.
roulette.maxPlayers	integer	no	6	Maximum players in one game.
roulette.turnTimeoutSeconds	integer	no	60	A player who does not pull the trigger within this many seconds loses.
roulette.lobbyTimeoutSeconds	integer	no	300	A lobby that has not started after this many seconds is dissolved.
roulette.muteSeconds	integer	no	60	Seconds the loser is muted in groups that turned on the mute penalty (off in every group by default).

rps

plugins/games/rps/schema.json

This Schema declares no fields.

text-adventure

plugins/games/text-adventure/schema.json

This Schema declares no fields.

tic-tac-toe

plugins/games/tic-tac-toe/schema.json

This Schema declares no fields.

word-riddle

plugins/games/word-riddle/schema.json

This Schema declares no fields.

activity-feedback

plugins/services/activity-feedback/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
activity-feedback.enabled	boolean	no	true	Enable activity feedback for this plugin instance.
activity-feedback.defaults	object	no	—	Default policy for all platforms and Endpoints.
activity-feedback.defaults.enabled	boolean	no	true	Enable this policy layer.
activity-feedback.defaults.phases	object	no	—	Feedback policy keyed by lifecycle phase.
activity-feedback.defaults.phases.queued	object	no	—	—
activity-feedback.defaults.phases.queued.private	object	no	—	—
activity-feedback.defaults.phases.queued.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.queued.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.queued.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.queued.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.queued.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.queued.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.queued.group	object	no	—	—
activity-feedback.defaults.phases.queued.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.queued.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.queued.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.queued.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.queued.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.queued.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.queued.channel	object	no	—	—
activity-feedback.defaults.phases.queued.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.queued.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.queued.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.queued.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.queued.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.queued.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.active	object	no	—	—
activity-feedback.defaults.phases.active.private	object	no	—	—
activity-feedback.defaults.phases.active.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.active.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.active.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.active.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.active.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.active.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.active.group	object	no	—	—
activity-feedback.defaults.phases.active.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.active.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.active.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.active.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.active.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.active.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.active.channel	object	no	—	—
activity-feedback.defaults.phases.active.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.active.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.active.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.active.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.active.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.active.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.thinking	object	no	—	—
activity-feedback.defaults.phases.thinking.private	object	no	—	—
activity-feedback.defaults.phases.thinking.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.thinking.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.thinking.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.thinking.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.thinking.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.thinking.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.thinking.group	object	no	—	—
activity-feedback.defaults.phases.thinking.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.thinking.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.thinking.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.thinking.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.thinking.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.thinking.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.thinking.channel	object	no	—	—
activity-feedback.defaults.phases.thinking.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.thinking.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.thinking.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.thinking.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.thinking.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.thinking.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_start	object	no	—	—
activity-feedback.defaults.phases.schedule_start.private	object	no	—	—
activity-feedback.defaults.phases.schedule_start.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_start.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_start.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_start.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_start.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_start.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_start.group	object	no	—	—
activity-feedback.defaults.phases.schedule_start.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_start.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_start.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_start.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_start.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_start.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_start.channel	object	no	—	—
activity-feedback.defaults.phases.schedule_start.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_start.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_start.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_start.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_start.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_start.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_finish	object	no	—	—
activity-feedback.defaults.phases.schedule_finish.private	object	no	—	—
activity-feedback.defaults.phases.schedule_finish.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_finish.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_finish.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_finish.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_finish.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_finish.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_finish.group	object	no	—	—
activity-feedback.defaults.phases.schedule_finish.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_finish.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_finish.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_finish.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_finish.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_finish.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_finish.channel	object	no	—	—
activity-feedback.defaults.phases.schedule_finish.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_finish.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_finish.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_finish.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_finish.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_finish.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_error	object	no	—	—
activity-feedback.defaults.phases.schedule_error.private	object	no	—	—
activity-feedback.defaults.phases.schedule_error.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_error.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_error.private.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_error.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_error.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_error.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_error.group	object	no	—	—
activity-feedback.defaults.phases.schedule_error.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_error.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_error.group.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_error.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_error.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_error.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.defaults.phases.schedule_error.channel	object	no	—	—
activity-feedback.defaults.phases.schedule_error.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.defaults.phases.schedule_error.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.defaults.phases.schedule_error.channel.message	string	no	—	Status text for type=message.
activity-feedback.defaults.phases.schedule_error.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.defaults.phases.schedule_error.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.defaults.phases.schedule_error.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms	object	no	—	Platform overrides keyed by platform name; values use the same shape as defaults.
activity-feedback.platforms.<platform>.enabled	boolean	no	true	Enable this policy layer.
activity-feedback.platforms.<platform>.phases	object	no	—	Feedback policy keyed by lifecycle phase.
activity-feedback.platforms.<platform>.phases.queued	object	no	—	—
activity-feedback.platforms.<platform>.phases.queued.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.queued.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.queued.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.queued.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.queued.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.queued.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.queued.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.queued.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.queued.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.queued.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.queued.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.queued.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.queued.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.queued.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.queued.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.queued.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.queued.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.queued.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.queued.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.queued.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.queued.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.active	object	no	—	—
activity-feedback.platforms.<platform>.phases.active.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.active.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.active.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.active.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.active.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.active.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.active.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.active.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.active.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.active.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.active.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.active.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.active.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.active.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.active.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.active.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.active.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.active.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.active.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.active.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.active.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.thinking	object	no	—	—
activity-feedback.platforms.<platform>.phases.thinking.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.thinking.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.thinking.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.thinking.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.thinking.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.thinking.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.thinking.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.thinking.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.thinking.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.thinking.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.thinking.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.thinking.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.thinking.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.thinking.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.thinking.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.thinking.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.thinking.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.thinking.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.thinking.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.thinking.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.thinking.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_start	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_start.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_start.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_start.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_start.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_start.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_start.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_start.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_start.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_start.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_start.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_start.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_start.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_start.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_start.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_start.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_start.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_start.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_start.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_start.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_start.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_start.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_finish	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_finish.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_finish.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_finish.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_finish.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_finish.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_finish.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_finish.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_finish.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_finish.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_finish.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_finish.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_finish.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_finish.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_finish.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_finish.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_error	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_error.private	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_error.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_error.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_error.private.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_error.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_error.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_error.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_error.group	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_error.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_error.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_error.group.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_error.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_error.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_error.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.platforms.<platform>.phases.schedule_error.channel	object	no	—	—
activity-feedback.platforms.<platform>.phases.schedule_error.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.platforms.<platform>.phases.schedule_error.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.platforms.<platform>.phases.schedule_error.channel.message	string	no	—	Status text for type=message.
activity-feedback.platforms.<platform>.phases.schedule_error.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.platforms.<platform>.phases.schedule_error.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.platforms.<platform>.phases.schedule_error.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints	object	no	—	Endpoint overrides keyed by platform:endpointKey; values use the same shape as defaults.
activity-feedback.endpoints.<platform:endpointKey>.enabled	boolean	no	true	Enable this policy layer.
activity-feedback.endpoints.<platform:endpointKey>.phases	object	no	—	Feedback policy keyed by lifecycle phase.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.queued.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.active	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.active.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.thinking.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_start.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_finish.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel	object	no	—	—
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.message	string	no	—	Status text for type=message.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.endpoints.<platform:endpointKey>.phases.schedule_error.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule	object	no	—	Schedule-only start, finish, and error feedback.
activity-feedback.schedule.phases	object	no	—	—
activity-feedback.schedule.phases.start	object	no	—	—
activity-feedback.schedule.phases.start.private	object	no	—	—
activity-feedback.schedule.phases.start.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.start.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.start.private.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.start.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.start.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.start.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.start.group	object	no	—	—
activity-feedback.schedule.phases.start.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.start.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.start.group.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.start.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.start.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.start.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.start.channel	object	no	—	—
activity-feedback.schedule.phases.start.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.start.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.start.channel.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.start.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.start.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.start.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.finish	object	no	—	—
activity-feedback.schedule.phases.finish.private	object	no	—	—
activity-feedback.schedule.phases.finish.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.finish.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.finish.private.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.finish.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.finish.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.finish.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.finish.group	object	no	—	—
activity-feedback.schedule.phases.finish.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.finish.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.finish.group.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.finish.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.finish.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.finish.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.finish.channel	object	no	—	—
activity-feedback.schedule.phases.finish.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.finish.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.finish.channel.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.finish.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.finish.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.finish.channel.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.error	object	no	—	—
activity-feedback.schedule.phases.error.private	object	no	—	—
activity-feedback.schedule.phases.error.private.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.error.private.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.error.private.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.error.private.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.error.private.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.error.private.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.error.group	object	no	—	—
activity-feedback.schedule.phases.error.group.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.error.group.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.error.group.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.error.group.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.error.group.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.error.group.platformConfig	object	no	—	Platform-specific options.
activity-feedback.schedule.phases.error.channel	object	no	—	—
activity-feedback.schedule.phases.error.channel.type	string: "reaction", "message", "typing", "none"	no	—	How the phase is presented.
activity-feedback.schedule.phases.error.channel.emoji	string	no	—	Reaction value for type=reaction.
activity-feedback.schedule.phases.error.channel.message	string	no	—	Status text for type=message.
activity-feedback.schedule.phases.error.channel.autoRemove	boolean	no	true	Remove the feedback after the phase stops.
activity-feedback.schedule.phases.error.channel.removeDelay	number	no	—	Delay before removal in milliseconds; negative values are normalized to zero at runtime.
activity-feedback.schedule.phases.error.channel.platformConfig	object	no	—	Platform-specific options.

typesafe

plugins/services/typesafe/schema.json

This Schema declares no fields.

60s

plugins/utils/60s/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
60s.apiBase	string	no	"https://60s.viki.moe"	Base URL of the 60s API service.

answers

plugins/utils/answers/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; every field can be overridden per adapter / account; conversation fields: dailyStable, quizTimeoutSeconds, enableDrivingQuiz; in-chat self-service: quizTimeoutSeconds, enableDrivingQuiz

Path	Type	Required	Default	Description
answers.dailyStable	boolean	no	false	Return the same Book of Answers response when the same user asks the same question on the same day.
answers.timezone	string	no	"Asia/Shanghai"	IANA timezone used by dailyStable to determine the day.
answers.quizTimeoutSeconds	number	no	120	Seconds to wait for a quiz answer; the question expires afterward.
answers.enableDrivingQuiz	boolean	no	true	Enable the original driving-theory quiz, for entertainment and reference only.

bilibili

plugins/utils/bilibili/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
bilibili.pollCron	string	no	"0 * * * * *"	Poll schedule (6-field cron, every 60 seconds by default).
bilibili.jitterMs	number	no	15000	Maximum random delay before each poll, in milliseconds.
bilibili.requestGapMs	number	no	1500	Gap between dynamic requests within one poll, in milliseconds.
bilibili.maxBackoffMinutes	number	no	30	Upper bound of exponential backoff after API failures or rate limiting, in minutes.
bilibili.maxPerChannel	number	no	20	Maximum UP subscriptions per conversation.
bilibili.timeout	number	no	10000	HTTP request timeout in milliseconds.
bilibili.live	boolean	no	true	Poll live-stream status.
bilibili.dynamic	boolean	no	true	Poll UP activity.
bilibili.cookie	string	no	""	Optional Bilibili cookie for stricter rate limits. Reference ${BILIBILI_COOKIE}; do not put credentials directly in configuration.

calendar

plugins/utils/calendar/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
calendar.provider	"none" | "caldav" | "google" | "microsoft"	no	"none"	Calendar backend. Tools report that the calendar is disabled when set to none.
calendar.timezone	string	no	"Asia/Manila"	IANA timezone for interpreting times without offsets and for display.
calendar.work_day_start	string	no	"09:00"	Daily start time used for free-time lookups (HH:MM).
calendar.work_day_end	string	no	"18:00"	Daily end time used for free-time lookups (HH:MM).
calendar.oauth_redirect_base_url	string	no	""	External base URL for the chat authorization callback (such as https://bot.example.com). Uses the local address of the zhin HTTP Host when empty.
calendar.oauth_callback_path	string	no	"/calendar/oauth/callback"	Authorization callback route on the zhin HTTP Host. Adjust it when a reverse proxy or tunnel forwards only some paths.
calendar.caldav	object	no	—	—
calendar.caldav.enabled	boolean	no	false	—
calendar.caldav.calendar_url	string	no	""	Calendar collection URL.
calendar.caldav.username	string	no	""	—
calendar.caldav.password	string	no	""	Prefer an environment-variable reference.
calendar.google	object	no	—	—
calendar.google.enabled	boolean	no	false	—
calendar.google.client_id	string	no	""	$
calendar.google.client_secret	string	no	""	$
calendar.google.refresh_token	string	no	""	${GOOGLE_CALENDAR_REFRESH_TOKEN}; obtain using the calendar authorization flow.
calendar.google.calendar_id	string	no	"primary"	Calendar ID, primary by default; can be set to another calendar ID.
calendar.google.token_file	string	no	""	Optional JSON path that persists the rotated refresh_token (such as data/calendar-google-token.json).
calendar.microsoft	object	no	—	—
calendar.microsoft.enabled	boolean	no	false	—
calendar.microsoft.client_id	string	no	""	$
calendar.microsoft.client_secret	string	no	""	Optional for public clients. $
calendar.microsoft.refresh_token	string	no	""	${MS_GRAPH_REFRESH_TOKEN}; obtain using the calendar authorization flow.
calendar.microsoft.tenant	string	no	"common"	Azure tenant, common by default.
calendar.microsoft.calendar_id	string	no	""	Empty, primary or me selects the default calendar; otherwise a Graph calendar ID.
calendar.microsoft.token_file	string	no	""	Optional JSON path that persists the rotated refresh_token (such as data/calendar-ms-token.json).

code-runner

plugins/utils/code-runner/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
code-runner.provider	"piston" | "glot"	no	"piston"	Backend protocol: piston (a self-hosted Piston /api/v2/execute) or glot (a glot.io-compatible /api/run).
code-runner.endpoint	string	no	""	Backend URL. For piston, the full execute URL (e.g. http://127.0.0.1:2000/api/v2/execute); for glot, the run prefix (e.g. https://glot.io/api/run). Empty means not configured. ${CODE_RUNNER_ENDPOINT} is supported.
code-runner.token	string	no	—	Optional backend token; prefer ${CODE_RUNNER_TOKEN}. Sent as the raw Authorization header for piston and as "Authorization: Token <token>" for glot.

content-moderation

plugins/utils/content-moderation/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
content-moderation.groupModeration	array<object>	no	[]	Exact scope for automatic advertising/abuse moderation. Disabled when unconfigured. Requires AI, atomic persistent state and adapter group-management support.
content-moderation.groupModeration[].adapter	string	yes	—	—
content-moderation.groupModeration[].endpoint	string	yes	—	—
content-moderation.groupModeration[].account	string	yes	—	—
content-moderation.groupModeration[].group	string	yes	—	—
content-moderation.enabled	boolean	no	true	Enable moderation.
content-moderation.onError	string: "open", "closed"	no	"open"	Default policy when a review source fails: open treats it as passed; closed treats it as critical.
content-moderation.maskChar	string	no	"*"	Text masking character.
content-moderation.replyTemplate	string	no	—	Reply text for the reply action. When omitted, use the message language; explicit text is preserved.
content-moderation.inbound	object	no	—	—
content-moderation.inbound.enabled	boolean	no	true	—
content-moderation.inbound.bypass	array<string>	no	["owner","platform:admin"]	Senders matching any of these permit expressions skip inbound review (ADR 0056 §5 built-in DSL, the same evaluator as commands and tools). Moderation is a platform function: the default exempts the zhin owner and platform group owners / admins (platform:admin); [] reviews everyone.
content-moderation.inbound.whitelist	object	no	—	—
content-moderation.inbound.whitelist.userIds	array<string>	no	[]	Messages from these users are not reviewed.
content-moderation.inbound.whitelist.conversationIds	array<string>	no	[]	Messages in these conversations are not reviewed.
content-moderation.outbound	object	no	—	—
content-moderation.outbound.enabled	boolean	no	true	—
content-moderation.outbound.bypass	boolean	no	false	Skip outbound review when true.
content-moderation.actions	object	no	—	Severity-to-action mapping (string or array); uses the stricter default mapping when omitted.
content-moderation.actions.pass	string: "allow", "log", "reply", "redact", "drop", "recall" | array<string: "allow", "log", "reply", "redact", "drop", "recall">	no	—	—
content-moderation.actions.low	string: "allow", "log", "reply", "redact", "drop", "recall" | array<string: "allow", "log", "reply", "redact", "drop", "recall">	no	—	—
content-moderation.actions.medium	string: "allow", "log", "reply", "redact", "drop", "recall" | array<string: "allow", "log", "reply", "redact", "drop", "recall">	no	—	—
content-moderation.actions.high	string: "allow", "log", "reply", "redact", "drop", "recall" | array<string: "allow", "log", "reply", "redact", "drop", "recall">	no	—	—
content-moderation.actions.critical	string: "allow", "log", "reply", "redact", "drop", "recall" | array<string: "allow", "log", "reply", "redact", "drop", "recall">	no	—	—
content-moderation.sources	array<object>	no	[]	Review sources run in order: local word lists or HTTP review services.
content-moderation.sources[].id	string	yes	—	—
content-moderation.sources[].type	string: "local", "http"	yes	—	local is a word list; http is an external review service.
content-moderation.sources[].enabled	boolean	no	true	—
content-moderation.sources[].onError	string: "open", "closed"	no	—	Policy when this source fails; uses the global setting when empty.
content-moderation.sources[].includeBuiltin	boolean	no	true	Include the built-in graded blocked-word dictionary.
content-moderation.sources[].words	array<string | object>	no	[]	Custom words: strings using defaultSeverity, or { word, severity } objects.
content-moderation.sources[].wordFiles	array<string>	no	[]	Dictionary file; line formats: word, severity:word, or word|severity.
content-moderation.sources[].defaultSeverity	string: "low", "medium", "high", "critical"	no	"high"	Default severity for custom words without an explicit severity.
content-moderation.sources[].severity	string: "low", "medium", "high", "critical"	no	"high"	Legacy alias for defaultSeverity.
content-moderation.sources[].url	string	no	—	URL of the HTTP review service.
content-moderation.sources[].headers	object	no	—	Extra headers sent to the review service.
content-moderation.sources[].timeoutMs	number	no	5000	Timeout of one review request.
content-moderation.sources[].forceUpload	boolean	no	false	Download images before uploading them instead of passing the URL.

earthquake

plugins/utils/earthquake/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
earthquake.sources	array<string: "cenc", "usgs">	no	["cenc","usgs"]	Enabled data sources in priority order. The first source supplies location text when merging the same earthquake; other sources cover a failed source.
earthquake.cencUrls	array<string>	no	["https://api.wolfx.jp/cenc_eqlist.json","https://ceic.ac.cn/data/data.json"]	CENC catalog URLs, tried in order (primary followed by mirrors). HTTPS only, up to 5 URLs.
earthquake.usgsFeed	string: "significant_hour", "significant_day", "4.5_hour", "4.5_day", "2.5_hour", "2.5_day", "1.0_hour", "1.0_day"	no	"2.5_day"	USGS GeoJSON summary feed name (minimum-magnitude_time-range).
earthquake.pollCron	string	no	"0 */2 * * * *"	Subscription polling schedule (6-field cron: second, minute, hour, day, month, weekday).
earthquake.timezone	string	no	"Asia/Shanghai"	IANA timezone used to display earthquake times.
earthquake.requestTimeoutMs	number	no	10000	HTTP request timeout in milliseconds.
earthquake.maxRetries	number	no	2	Retries for network errors, timeouts, HTTP 429 and 5xx, excluding the initial request.
earthquake.retryBaseDelayMs	number	no	1000	Initial exponential-backoff delay in milliseconds; doubles for each retry with random jitter.
earthquake.retryMaxDelayMs	number	no	30000	Maximum backoff or Retry-After wait in milliseconds. Longer server waits abandon this round and cool down the source.
earthquake.defaultMinMagnitude	number	no	5	Minimum magnitude for subscriptions that do not specify one.
earthquake.queryMinMagnitude	number	no	3	Minimum magnitude for earthquake queries that do not specify one.
earthquake.queryLimit	number	no	5	Default number of earthquake query results.
earthquake.maxEventAgeMinutes	number	no	60	Only notify about earthquakes from the last specified number of minutes, avoiding delayed or replayed old events.
earthquake.pushLimit	number	no	3	Maximum notification messages per conversation in one rate-limit window.
earthquake.pushWindowMinutes	number	no	10	Notification rate-limit window in minutes.
earthquake.maxEventsPerMessage	number	no	5	Maximum earthquakes listed in one notification; the remainder is summarized as a count.
earthquake.maxSubscriptions	number	no	100	Maximum subscriptions across this instance, at most one per conversation.

group-suite

plugins/utils/group-suite/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; global only (other fields can be overridden per adapter / account): keywordReply, teachAllowRegex, welcomeConversations, ephemeralKeywords; conversation fields: welcomeMessage, basePointsMin, basePointsMax, streakBonus, streakCap, rankSize, teachMaxPerGroup, teachCooldownMs, teachPageSize; in-chat self-service: welcomeMessage, rankSize, teachCooldownMs, teachPageSize

Path	Type	Required	Default	Description
group-suite.keywordReply	boolean	no	false	Reply automatically to messages that match the keyword list (global only).
group-suite.basePointsMin	number	no	10	Lower bound of the random base points for one check-in.
group-suite.basePointsMax	number	no	30	Upper bound of the random base points for one check-in.
group-suite.streakBonus	number	no	5	Extra points per consecutive check-in day.
group-suite.streakCap	number	no	50	Maximum extra points from a check-in streak.
group-suite.rankSize	number	no	10	Number of members shown on check-in and activity leaderboards.
group-suite.teachMaxPerGroup	number	no	200	Maximum question-and-answer rules stored per group.
group-suite.teachCooldownMs	number	no	3000	Minimum interval between two teach commands from the same member.
group-suite.teachAllowRegex	boolean	no	false	Allow members to teach regular-expression rules. Off by default: patterns run against every inbound message. Only structurally safe patterns are accepted and they are tried on messages of at most 200 characters.
group-suite.teachPageSize	number	no	10	Rules shown per page by teach-list.
group-suite.welcomeMessage	string	no	""	Group welcome text. Empty disables; {name}/{id} refer to the joined member.
group-suite.welcomeConversations	array<object>	no	—	Exact welcome conversations. Omitted welcomes all groups; [] welcomes none. Missing parent/thread requires their absence; no wildcards.
group-suite.welcomeConversations[].endpoint	object	yes	—	—
group-suite.welcomeConversations[].endpoint.adapter	string	yes	—	—
group-suite.welcomeConversations[].endpoint.id	string	yes	—	—
group-suite.welcomeConversations[].kind	string: "group"	yes	—	—
group-suite.welcomeConversations[].id	string	yes	—	—
group-suite.welcomeConversations[].parent	object	no	—	—
group-suite.welcomeConversations[].parent.kind	string: "private", "group", "channel"	yes	—	—
group-suite.welcomeConversations[].parent.id	string	yes	—	—
group-suite.welcomeConversations[].threadId	string	no	—	—
group-suite.ephemeralKeywords	boolean	no	false	Explicit volatile keywords. With false, automatic replies require Database Host at startup and keyword management requires its persistent state service when invoked.

ledger

plugins/utils/ledger/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; global only (other fields can be overridden per adapter / account): timezone, reportTime; conversation fields: reportWhenEmpty

Path	Type	Required	Default	Description
ledger.timezone	string	no	"Asia/Shanghai"	IANA timezone used to calculate today, this week (starting Monday), this month and daily-report dates.
ledger.reportTime	string	no	"22:00"	Daily ledger report time (HH:MM in timezone); sent only to users who explicitly subscribed to daily reports.
ledger.reportWhenEmpty	boolean	no	false	Send a daily report even when no transactions were recorded that day.

link-poster

plugins/utils/link-poster/schema.json

This Schema declares no fields.

lottery

plugins/utils/lottery/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
lottery.pickCount	number	no	5	Numbers in each recommended pick.
lottery.scheduleCron	string	no	"0 0 18 * * *"	6-field cron expression for scheduled recommendations.
lottery.historyLimit	number	no	500	Maximum past draws read for analysis.
lottery.scheduleEnabled	boolean	no	true	Generate and push recommendations on the schedule.
lottery.backtestEnabled	boolean	no	true	Backtest the model on past draws before recommending.
lottery.backtestWindow	number	no	50	Number of recent draws sampled by the backtest.
lottery.backtestRandomTrials	number	no	64	Simulations of the random baseline in a backtest.
lottery.backtestMinHistory	number	no	30	Skip the backtest with fewer past draws than this.
lottery.backtestAdaptive	boolean	no	true	Tune model weights draw by draw during the backtest.
lottery.weightPersistEnabled	boolean	no	true	Save model weights to the database after simulation or review.
lottery.weightHoldoutFallback	boolean	no	true	Use default weights when the recent holdout performs worse than random.
lottery.games	array<string>	no	["kl8","ssq","dlt","fc3d","pl3","pl5"]	Enabled lottery games (KL8, SSQ, DLT, FC3D, PL3, PL5).
lottery.pushTargets	array<object>	no	[]	Conversations that receive scheduled and draw reports.
lottery.pushTargets[].adapter	string	yes	—	—
lottery.pushTargets[].endpointKey	string	no	—	—
lottery.pushTargets[].channelType	string	no	"private"	—
lottery.pushTargets[].channelId	string	yes	—	—
lottery.kl8	object	no	—	—
lottery.kl8.pickCount	number	no	5	Numbers in each KL8 group.
lottery.kl8.recommendGroups	number	no	3	Number of KL8 groups per recommendation.
lottery.kl8.groupStrategies	array<string>	no	["balanced","hot","cold"]	Strategy for each group (balanced, hot, cold).

music

plugins/utils/music/schema.json

This Schema declares no fields.

nettools

plugins/utils/nettools/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
nettools.cooldownSeconds	integer	no	10	Per-user cooldown in seconds, shared by all nettools commands and tools (0 disables)
nettools.requestTimeoutMs	integer	no	8000	Per-attempt timeout for each external HTTP request, in milliseconds
nettools.maxRetries	integer	no	2	Retries for transient failures (network errors, timeouts, 429, 5xx) with exponential backoff and jitter
nettools.pingMaxNodes	integer	no	4	Number of check-host.net probe nodes used by ping
nettools.pingPollIntervalMs	integer	no	2000	Wait between ping result polls, in milliseconds
nettools.pingMaxPolls	integer	no	5	Maximum number of ping result polls before reporting partial results
nettools.dohProviders	array<string: "cloudflare", "alidns", "google">	no	["cloudflare","alidns","google"]	DNS-over-HTTPS providers in fallback order
nettools.priceListLimit	integer	no	8	Number of registrars shown in a domain price comparison

qrcode

plugins/utils/qrcode/schema.json

This Schema declares no fields.

reminder

plugins/utils/reminder/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
reminder.timezone	string	no	"UTC"	IANA timezone for natural reminder times and display. Absolute ISO times still require an explicit offset.
reminder.maxTasks	number	no	20	Maximum active personal chat reminders per instance, separate from administrator tasks. Lowering the limit does not delete existing tasks.
reminder.adminMaxTasks	number	no	20	Maximum active administrator reminders (scheduled, paused or unconfirmed) per instance. Separate from chat tasks; the two quotas are additive.
reminder.maxTasksPerSender	number	no	10	Maximum tasks per sender in the same conversation.
reminder.batchSize	number	no	5	Maximum tasks delivered in each scan.
reminder.missedGraceSeconds	number	no	15	Skip recurring runs delayed beyond this many seconds. Missed runs during downtime are not backfilled.

repeater

plugins/utils/repeater/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; every field can be overridden per adapter / account; conversation fields: threshold, cooldown; in-chat self-service: threshold

Path	Type	Required	Default	Description
repeater.threshold	number	no	3	Minimum distinct users needed to trigger repetition.
repeater.cooldown	number	no	30000	Cooldown for the same group, in milliseconds.
repeater.maxLength	number	no	200	Maximum message length.

rss

plugins/utils/rss/schema.json

Layered configuration (ADR 0050 / 0055) — Levels: global, adapter, endpoint, conversation; fields: global only; conversation fields: switch only

Path	Type	Required	Default	Description
rss.pollCron	string	no	"0 */5 * * * *"	Polling schedule (6-field cron expression).
rss.maxPerGroup	number	no	30	Maximum feeds per conversation.
rss.maxItems	number	no	5	Maximum articles listed in one notification; the rest are counted.
rss.timeout	number	no	15000	Timeout for fetching a feed.
rss.allowPrivateNetwork	boolean	no	false	Allow feeds on loopback/LAN addresses (for example a self-hosted RSSHub). Refused by default because feed URLs come from chat users.

short-url

plugins/utils/short-url/schema.json

This Schema declares no fields.