* feat(context-engine): add ContextEngine interface and registry
Introduce the pluggable ContextEngine abstraction that allows external
plugins to register custom context management strategies.
- ContextEngine interface with lifecycle methods: bootstrap, ingest,
ingestBatch, afterTurn, assemble, compact, prepareSubagentSpawn,
onSubagentEnded, dispose
- Module-level singleton registry with registerContextEngine() and
resolveContextEngine() (config-driven slot selection)
- LegacyContextEngine: pass-through implementation wrapping existing
compaction behavior for 100% backward compatibility
- ensureContextEnginesInitialized() guard for safe one-time registration
- 19 tests covering contract, registry, resolution, and legacy parity
* feat(plugins): add context-engine slot and registerContextEngine API
Wire the ContextEngine abstraction into the plugin system so external
plugins can register context engines via the standard plugin API.
- Add 'context-engine' to PluginKind union type
- Add 'contextEngine' slot to PluginSlotsConfig (default: 'legacy')
- Wire registerContextEngine() through OpenClawPluginApi
- Export ContextEngine types from plugin-sdk for external consumers
- Restore proper slot-based resolution in registry
* feat(context-engine): wire ContextEngine into agent run lifecycle
Integrate the ContextEngine abstraction into the core agent run path:
- Resolve context engine once per run (reused across retries)
- Bootstrap: hydrate canonical store from session file on first run
- Assemble: route context assembly through pluggable engine
- Auto-compaction guard: disable built-in auto-compaction when
the engine declares ownsCompaction (prevents double-compaction)
- AfterTurn: post-turn lifecycle hook for ingest + background
compaction decisions
- Overflow compaction: route through contextEngine.compact()
- Dispose: clean up engine resources in finally block
- Notify context engine on subagent lifecycle events
Legacy engine: all lifecycle methods are pass-through/no-op, preserving
100% backward compatibility for users without a context engine plugin.
* feat(plugins): add scoped subagent methods and gateway request scope
Expose runtime.subagent.{run, waitForRun, getSession, deleteSession}
so external plugins can spawn sub-agent sessions without raw gateway
dispatch access.
Uses AsyncLocalStorage request-scope bridge to dispatch internally via
handleGatewayRequest with a synthetic operator client. Methods are only
available during gateway request handling.
- Symbol.for-backed global singleton for cross-module-reload safety
- Fallback gateway context for non-WS dispatch paths (Telegram/WhatsApp)
- Set gateway request scope for all handlers, not just plugin handlers
- 3 staleness tests for fallback context hardening
* feat(context-engine): route /compact and sessions.get through context engine
Wire the /compact command and sessions.get handler through the pluggable
ContextEngine interface.
- Thread tokenBudget and force parameters to context engine compact
- Route /compact through contextEngine.compact() when registered
- Wire sessions.get as runtime alias for plugin subagent dispatch
- Add .pebbles/ to .gitignore
* style: format with oxfmt 0.33.0
Fix duplicate import (ControlUiRootState in server.impl.ts) and
import ordering across all changed files.
* fix: update extension test mocks for context-engine types
Add missing subagent property to bluebubbles PluginRuntime mock.
Add missing registerContextEngine to lobster OpenClawPluginApi mock.
* fix(subagents): keep deferred delete cleanup retryable
* style: format run attempt for CI
* fix(rebase): remove duplicate embedded-run imports
* test: add missing gateway context mock export
* fix: pass resolved auth profile into afterTurn compaction
Ensure the embedded runner forwards resolved auth profile context into
legacy context-engine compaction params on the normal afterTurn path,
matching overflow compaction behavior. This allows downstream LCM
summarization to use the intended provider auth/profile consistently.
Also fix strict TS typing in external-link token dedupe and align an
attempt unit test reasoningLevel value with the current ReasoningLevel
enum.
Regeneration-Prompt: |
We were debugging context-engine compaction where downstream summary
calls were missing the right auth/profile context in normal afterTurn
flow, while overflow compaction already propagated it. Preserve current
behavior and keep changes additive: thread the resolved authProfileId
through run -> attempt -> legacy compaction param builder without
broad refactors.
Add tests that prove the auth profile is included in afterTurn legacy
params and that overflow compaction still passes it through run
attempts. Keep existing APIs stable, and only adjust small type issues
needed for strict compilation.
* fix: remove duplicate imports from rebase
* feat: add context-engine system prompt additions
* fix(rebase): dedupe attempt import declarations
* test: fix fetch mock typing in ollama autodiscovery
* fix(test): add registerContextEngine to diffs extension mock APIs
* test(windows): use path.delimiter in ios-team-id fixture PATH
* test(cron): add model formatting and precedence edge case tests
Covers:
- Provider/model string splitting (whitespace, nested paths, empty segments)
- Provider normalization (casing, aliases like bedrock→amazon-bedrock)
- Anthropic model alias normalization (opus-4.5→claude-opus-4-5)
- Precedence: job payload > session override > config default
- Sequential runs with different providers (CI flake regression pattern)
- forceNew session preserving stored model overrides
- Whitespace/empty model string edge cases
- Config model as string vs object format
* test(cron): fix model formatting test config types
* test(phone-control): add registerContextEngine to mock API
* fix: re-export ChannelKind from config-reload-plan
* fix: add subagent mock to plugin-runtime-mock test util
* docs: add changelog fragment for context engine PR #22201
* feat(feishu): add broadcast support for multi-agent group observation
When multiple agents share a Feishu group chat, only the @mentioned
agent receives the message. This prevents observer agents from building
session memory of group activity they weren't directly addressed in.
Adds broadcast support (reusing the same cfg.broadcast schema as
WhatsApp) so all configured agents receive every group message in their
session transcripts. Only the @mentioned agent responds on Feishu;
observer agents process silently via no-op dispatchers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): guard sequential broadcast dispatch against single-agent failure
Wrap each dispatchForAgent() call in the sequential loop with try/catch
so one agent's dispatch failure doesn't abort delivery to remaining agents.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): avoid duplicate messages in broadcast observer mode and normalize agent IDs
- Skip recordPendingHistoryEntryIfEnabled for broadcast groups when not
mentioned, since the message is dispatched directly to all agents.
Previously the message appeared twice in the agent prompt.
- Normalize agent IDs with toLowerCase() before membership checks so
config casing mismatches don't silently skip valid agents.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): set WasMentioned per-agent and normalize broadcast IDs
- buildCtxPayloadForAgent now takes a wasMentioned parameter so active
agents get WasMentioned=true and observers get false (P1 fix)
- Normalize broadcastAgents to lowercase at resolution time and
lowercase activeAgentId so all comparisons and session key generation
use canonical IDs regardless of config casing (P2 fix)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): canonicalize broadcast agent IDs with normalizeAgentId
* fix(feishu): match ReplyDispatcher sync return types for noop dispatcher
The upstream ReplyDispatcher changed sendToolResult/sendBlockReply/
sendFinalReply to synchronous (returning boolean). Update the broadcast
observer noop dispatcher to match.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): deduplicate broadcast agent IDs after normalization
Config entries like "Main" and "main" collapse to the same canonical ID
after normalizeAgentId but were dispatched multiple times. Use Set to
deduplicate after normalization.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): honor requireMention=false when selecting broadcast responder
When requireMention is false, the routed agent should be active (reply
on Feishu) even without an explicit @mention. Previously activeAgentId
was null whenever ctx.mentionedBot was false, so all agents got the
noop dispatcher and no reply was sent — silently breaking groups that
disabled mention gating.
Hoist requireMention out of the if(isGroup) block so it's accessible
in the dispatch code.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): cross-account broadcast dedup to prevent duplicate dispatches
In multi-account Feishu setups, the same message event is delivered to
every bot account in a group. Without cross-account dedup, each account
independently dispatches broadcast agents, causing 2×N dispatches instead
of N (where N = number of broadcast agents).
Two changes:
1. requireMention=true + bot not mentioned: return early instead of
falling through to broadcast. The mentioned bot's handler will
dispatch for all agents. Non-mentioned handlers record to history.
2. Add cross-account broadcast dedup using a shared 'broadcast' namespace
(tryRecordMessagePersistent). The first handler to reach the broadcast
block claims the message; subsequent accounts skip. This handles the
requireMention=false multi-account case.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): strip CommandAuthorized from broadcast observer contexts
Broadcast observer agents inherited CommandAuthorized from the sender,
causing slash commands (e.g. /reset) to silently execute on every observer
session. Now only the active agent retains CommandAuthorized; observers
have it stripped before dispatch.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): use actual mention state for broadcast WasMentioned
The active broadcast agent's WasMentioned was set to true whenever
requireMention=false, even when the bot was not actually @mentioned.
Now uses ctx.mentionedBot && agentId === activeAgentId, consistent
with the single-agent path which passes ctx.mentionedBot directly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(feishu): skip history buffer for broadcast accounts and log parallel failures
1. In requireMention groups with broadcast, non-mentioned accounts no
longer buffer pending history — the mentioned handler's broadcast
dispatch already writes turns into all agent sessions. Buffering
caused duplicate replay via buildPendingHistoryContextFromMap.
2. Parallel broadcast dispatch now inspects Promise.allSettled results
and logs rejected entries, matching the sequential path's per-agent
error logging.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Changelog: note Feishu multi-agent broadcast dispatch
* Changelog: restore author credit for Feishu broadcast entry
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Tak Hoffman <781889+Takhoffman@users.noreply.github.com>
