Files
Fabric.OpenclawPlugin/src/channel.ts
hzhang 20e55849eb fix(channel): add describeAccount so health-monitor sees real configured
openclaw's `channelManager.getRuntimeSnapshot()` — called every minute
by the channel-health-monitor — runs accounts through
`applyDescribedAccountFields(next, plugin.config.describeAccount?.(...))`.
When the callback is missing it defaults `configured: true`. Fabric
never defined it, so every health-monitor cycle:

  snapshot = { enabled: true, configured: true, running: false }

For fabric's synthetic 'default' account (returned by
`listFabricAccountIds` when `channels.fabric.accounts` is empty —
the prod shape, where per-agent api-keys live in
`~/.openclaw/fabric-identity.json` and the channel framework never
runs `startAccount` so `running` stays false):

  isManagedAccount({enabled:true, configured:true}) === true
  -> not-running -> 'stopped' -> restart every ~10 min, logging
  '[fabric:default] health-monitor: restarting (reason: stopped)'

The restart is a no-op (fabric's `gateway.startAccount` is absent so
`startChannelInternal` returns early), but the log is loud and
operators chasing real outages keep wasting time on it.

Mirror `isConfigured` from describeAccount so the snapshot
truthfully reports configured:false for any account without a
fabricApiKey. The fabric plugin still self-manages real agents via
`gateway_start` -> `FabricInbound.start()`; the framework just no
longer thinks 'default' is something it should restart.

Verified in sim (this patch alone, no debug instrumentation):
- gateway up 8+ minutes, 0 restart events
- pre-patch sim with same config restarted at 5min mark
- evaluateChannelHealth snapshot for both 'default' and 'recruiter'
  accountId reads configured:false (instrumented with temporary
  console.log in channel-health-policy, since reverted)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-26 16:48:53 +01:00

230 lines
9.6 KiB
TypeScript

// Fabric channel plugin object (third-party `createChatChannelPlugin` path).
//
// COMPAT NOTE (openclaw v2026.5.7 plugin SDK):
// We depend on these generic SDK shapes — re-verify on openclaw upgrade:
// - createChannelPluginBase requires `capabilities`
// - ChannelSetupAdapter: only `applyAccountConfig` is required
// - outbound attachedResults.sendText returns Omit<OutboundDeliveryResult,
// "channel"> (so `messageId: string` is required)
// Casts at the createChatChannelPlugin boundary are intentional and
// localized; keep them here so upgrades touch one file.
import {
createChatChannelPlugin,
createChannelPluginBase,
buildChannelOutboundSessionRoute,
type ChannelOutboundSessionRouteParams,
} from 'openclaw/plugin-sdk/core';
import { FabricClient } from './fabric-client.js';
import {
listFabricAccountIds,
resolveFabricAccount,
resolveDefaultFabricAccountId,
type ResolvedFabricAccount,
} from './accounts.js';
import { getChannelType } from './channel-meta.js';
/**
* Map a Fabric channel xType to an openclaw routing peer.kind / ChatType.
*
* Fabric distinguishes channels by xType ('dm' | 'triage' | 'group' |
* 'broadcast' | 'announce' | ...). Openclaw's session router only knows
* 'direct' | 'group' | 'channel'. We collapse:
* - 'dm' → 'direct' (1:1 conversation; agent always speaks)
* - rest → 'group' (multi-party; turn-engine gates speech)
*
* Sessions are keyed by peer.kind, so inbound and outbound MUST agree —
* otherwise the agent's outbound message lands in a different session
* than the inbound that triggered it and conversation state splits.
*
* Outbound has no live xType (the agent target is just a channelId), so
* it consults the channel-meta cache populated by inbound. Cache miss
* (channel never observed) falls back to 'group' — same as the pre-fix
* behavior, no regression on cold cache. The proactive-DM-first-message
* edge case (agent DMs a channel before any inbound) still lands as
* 'group' on that one outbound; the next inbound + outbound pair will
* agree on 'direct'.
*/
export type FabricPeerRouting = { peerKind: 'direct' | 'group'; chatType: 'direct' | 'group' };
export function fabricPeerRoutingForXType(xType: string | null | undefined): FabricPeerRouting {
if (xType === 'dm') return { peerKind: 'direct', chatType: 'direct' };
return { peerKind: 'group', chatType: 'group' };
}
export function fabricPeerRoutingForChannel(channelId: string): FabricPeerRouting {
return fabricPeerRoutingForXType(getChannelType(channelId));
}
type AnyCfg = { channels?: { fabric?: unknown }; [k: string]: unknown };
// ---- target grammar: fabric:<channelId> ----
export function stripFabricTargetPrefix(raw: string): string | undefined {
let s = (raw ?? '').trim();
if (!s) return undefined;
if (s.toLowerCase().startsWith('fabric:')) s = s.slice('fabric:'.length).trim();
if (s.toLowerCase().startsWith('channel:')) s = s.slice('channel:'.length).trim();
return s || undefined;
}
export function normalizeFabricTarget(raw: string): string | undefined {
const id = stripFabricTargetPrefix(raw);
return id ? `fabric:${id}`.toLowerCase() : undefined;
}
export function looksLikeFabricTargetId(raw: string): boolean {
const t = (raw ?? '').trim();
if (!t) return false;
if (/^fabric:/i.test(t)) return true;
return /^[a-z0-9-]{8,}$/i.test(t);
}
export function resolveFabricOutboundSessionRoute(params: ChannelOutboundSessionRouteParams) {
const id = stripFabricTargetPrefix(params.target);
if (!id) return null;
// Consult the channel-meta cache populated by inbound — DM channels
// need peer.kind='direct' so the outbound session key matches the
// inbound one. Cache miss falls back to 'group' (the pre-fix default,
// no regression on cold cache).
const { peerKind, chatType } = fabricPeerRoutingForChannel(id);
return buildChannelOutboundSessionRoute({
cfg: params.cfg,
agentId: params.agentId,
channel: 'fabric',
accountId: params.accountId,
peer: { kind: peerKind, id },
chatType,
from: `fabric:channel:${id}`,
to: `fabric:${id}`,
});
}
// Posts an agent's reply to Fabric. `to` is the Fabric channelId; `accountId`
// is the agentId (= Fabric identity). One auth concept: account apiKey ->
// agent/login -> guild token -> POST message.
async function sendToFabric(
cfg: AnyCfg,
accountId: string | null | undefined,
to: string,
text: string,
): Promise<{ messageId: string }> {
const channelId = stripFabricTargetPrefix(to) ?? to;
const acc = resolveFabricAccount(cfg as never, accountId);
if (!acc.fabricApiKey) throw new Error(`fabric account ${acc.accountId} has no fabricApiKey`);
const client = new FabricClient(acc.centerApiBase);
const session = await client.agentLogin(acc.fabricApiKey);
// find which guild owns this channel by probing each guild's channel list
for (const g of session.guilds) {
const gt = session.guildAccessTokens.find((t) => t.guildNodeId === g.nodeId)?.token;
if (!gt) continue;
const res = await fetch(`${g.endpoint}/api/channels?guildId=${encodeURIComponent(g.nodeId)}`, {
headers: { authorization: `Bearer ${gt}` },
});
const channels = res.ok ? ((await res.json()) as Array<{ id: string }>) : [];
if (channels.some((c) => c.id === channelId)) {
await client.postMessage(g.endpoint, gt, channelId, text, session.user.id);
return { messageId: `${channelId}:${Date.now()}` };
}
}
// fallback: first guild
const g = session.guilds[0];
const gt = session.guildAccessTokens.find((t) => t.guildNodeId === g?.nodeId)?.token;
if (g && gt) {
await client.postMessage(g.endpoint, gt, channelId, text, session.user.id);
return { messageId: `${channelId}:${Date.now()}` };
}
throw new Error('fabric: no guild available to deliver');
}
export const fabricChannelPlugin = createChatChannelPlugin<ResolvedFabricAccount>({
base: {
...createChannelPluginBase<ResolvedFabricAccount>({
id: 'fabric',
meta: { id: 'fabric', label: 'Fabric', blurb: 'Connect OpenClaw agents to a Fabric guild.' },
capabilities: {
chatTypes: ['channel', 'group', 'direct'],
reactions: false,
threads: false,
media: false,
nativeCommands: false,
// Fabric has no message-length limit and we never want a reply split
// into multiple messages -> no block streaming.
blockStreaming: false,
},
reload: { configPrefixes: ['channels.fabric'] },
config: {
listAccountIds: (cfg) => listFabricAccountIds(cfg as never),
resolveAccount: (cfg, accountId) => resolveFabricAccount(cfg as never, accountId),
defaultAccountId: (cfg) => resolveDefaultFabricAccountId(cfg as never),
isConfigured: (account: ResolvedFabricAccount) => Boolean(account.fabricApiKey),
// openclaw's channelManager.getRuntimeSnapshot() — called every minute
// by the channel-health-monitor — defaults `configured: true` when the
// plugin doesn't expose describeAccount (see applyDescribedAccountFields
// in server-channels). Without this, fabric's synthetic 'default'
// account (returned by listFabricAccountIds when channels.fabric.accounts
// is empty — the prod shape) gets snapshot {enabled:true, configured:true,
// running:false} → isManagedAccount=true → not-running → restart loop
// every ~10 min, logging `[fabric:default] health-monitor: restarting`.
// Mirror isConfigured here so the snapshot truthfully reports false for
// any account without a fabricApiKey.
describeAccount: (account: ResolvedFabricAccount) => ({
configured: Boolean(account.fabricApiKey),
}),
},
// Minimal setup adapter: Fabric is configured directly under
// channels.fabric.* (no interactive wizard). applyAccountConfig is the
// only required member.
setup: {
applyAccountConfig: ({ cfg }: { cfg: unknown }) => cfg as never,
} as never,
}),
messaging: {
normalizeTarget: normalizeFabricTarget,
resolveSessionTarget: ({ id }: { id: string }) => normalizeFabricTarget(`fabric:${id}`),
resolveOutboundSessionRoute: (params: ChannelOutboundSessionRouteParams) =>
resolveFabricOutboundSessionRoute(params),
targetResolver: { looksLikeId: looksLikeFabricTargetId, hint: '<channelId|fabric:ID>' },
},
} as never,
security: {
dm: {
channelKey: 'fabric',
resolvePolicy: (a: ResolvedFabricAccount) => a.dmPolicy,
resolveAllowFrom: (a: ResolvedFabricAccount) => a.allowFrom,
defaultPolicy: 'allowlist',
},
},
threading: { topLevelReplyToMode: 'channel' },
outbound: {
base: {
deliveryMode: 'direct',
// Fabric has no length limit: never chunk — always one message.
chunker: (text: string) => [text],
textChunkLimit: Number.MAX_SAFE_INTEGER,
},
attachedResults: {
channel: 'fabric',
sendText: async (ctx: {
accountId?: string | null;
to: string;
text: string;
cfg?: unknown;
config?: unknown;
}) => {
// openclaw passes config under cfg or config depending on path.
// Note: inbound agent replies go through inbound.ts `deliver`
// (where turn coalescing happens). This path is for any direct
// outbound sends and posts immediately.
const cfg = (ctx.cfg ?? ctx.config ?? {}) as AnyCfg;
try {
const r = await sendToFabric(cfg, ctx.accountId ?? null, ctx.to, ctx.text);
console.log(`[fabric] outbound.sendText -> ${ctx.to} ok`);
return r;
} catch (e) {
console.log(`[fabric] outbound.sendText FAILED to=${ctx.to}: ${String(e)}`);
throw e;
}
},
},
} as never,
});