Fix OpenClaw plugin recall assembly

[EasionF]

Fix OpenClaw plugin recall assembly

Summary

This patch changes the OpenViking OpenClaw plugin recall path so context-engine mode no longer injects recalled memories through before_prompt_build / prependContext.

Instead, auto-recall is owned by assemble() and added as a short internal context message alongside archive/session context. The recall payload is also shortened and filtered before injection.

Problem

The previous plugin path could inject recalled memory as plain text during before_prompt_build. In string-content model pipelines this can surface raw recall context in the final prompt/answer, increase prompt pollution, and duplicate recall work even when the context engine owns assembly.

Changes

• Skip before_prompt_build recall injection whenever a context engine is registered.
• • Move recall lookup into context-engine assemble().
• • Inject only short, normalized recall facts into assembled context messages.
• • Prefer abstract recall by default when the option is not explicitly configured.
• • Filter instruction-like preference memories unless the query is asking for preferences.
• • Update focused unit tests for context-engine assembly and normal plugin flow.

Validation

Focused plugin tests passed:

npm test -- --run tests/ut/context-engine-assemble.test.ts tests/ut/plugin-normal-flow-real-

[CLAassistant]

All committers have signed the CLA.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🏅 Score: 85

🧪 PR contains tests

🔒 No security concerns identified

✅ No TODO sections

🔀 Multiple PR themes Sub-PR theme: Move recall into context-engine assemble() Relevant files: examples/openclaw-plugin/context-engine.ts examples/openclaw-plugin/tests/ut/context-engine-assemble.test.ts Sub-PR theme: Skip before_prompt_build recall when context engine is registered Relevant files: examples/openclaw-plugin/index.ts examples/openclaw-plugin/tests/ut/plugin-normal-flow-real-server.test.ts Sub-PR theme: Filter instructional preferences and adjust recallPreferAbstract default Relevant files: examples/openclaw-plugin/memory-ranking.ts examples/openclaw-plugin/config.ts

⚡ Recommended focus areas for review Unused scoreThreshold parameter in pickMemoriesForInjection The pickMemoriesForInjection function still accepts a scoreThreshold parameter but no longer uses it to filter memories, which may lead to lower-scoring memories being included unexpectedly. export function pickMemoriesForInjection( items: FindResultItem[], limit: number, queryText: string, scoreThreshold: number = 0, ): FindResultItem[] { if (items.length === 0 || limit <= 0) { return []; } const query = buildRecallQueryProfile(queryText); const filteredItems = query.wantsPreference ? items : items.filter((item) => !isInstructionLikePreference(item)); const candidateItems = filteredItems.length > 0 ? filteredItems : items; const sorted = [...candidateItems].sort((a, b) => rankForInjection(b, query) - rankForInjection(a, query)); Garbled comment (encoding issue) Line 1058 contains a garbled Chinese comment due to encoding mismatch. // 鍚堝苟涓や釜浣嶇疆鐨勭粨鏋滐紝鍘婚噸

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
General	Add timeout to API calls in assemble recall Add timeout logic to the client.find() and client.read() calls in the assemble recall path, similar to the AUTO_RECALL_TIMEOUT_MS pattern used in the plugin's before_prompt_build hook, to improve API resilience. examples/openclaw-plugin/context-engine.ts [796-806] +// Add a timeout helper at the top of the file or use a shared utility +function withTimeout<T>(promise: Promise<T>, ms: number, message: string): Promise<T> { + return Promise.race([ + promise, + new Promise<never>((_, reject) => setTimeout(() => reject(new Error(message)), ms)), + ]); +} + +// Then use it in buildAssembleRecallFacts: const [userSettled, agentSettled] = await Promise.allSettled([ - client.find(queryText, { - targetUri: "viking://user/memories", - limit: candidateLimit, - scoreThreshold: 0, - }, agentId), - client.find(queryText, { - targetUri: "viking://agent/memories", - limit: candidateLimit, - scoreThreshold: 0, - }, agentId), + withTimeout( + client.find(queryText, { targetUri: "viking://user/memories", limit: candidateLimit, scoreThreshold: 0 }, agentId), + AUTO_RECALL_TIMEOUT_MS, + "openviking: assemble user memories search timeout", + ), + withTimeout( + client.find(queryText, { targetUri: "viking://agent/memories", limit: candidateLimit, scoreThreshold: 0 }, agentId), + AUTO_RECALL_TIMEOUT_MS, + "openviking: assemble agent memories search timeout", + ), ]); Suggestion importance[1-10]: 6 __ Why: The suggestion adds timeout logic to client.find() calls in the assemble recall path, aligning with the existing timeout pattern in before_prompt_build and improving API resilience. It's a useful robustness improvement.	Low