feat: bridge QMD into Honcho memory_search/memory_get - #99
unithejerk wants to merge 14 commits into
Conversation
When memory.backend: 'qmd' is configured, the Honcho plugin's memory_search/memory_get tools only return Honcho session transcripts. This patch adds parallel QMD query execution so indexed local files (such as wiki docs) appear in results alongside session data. Changes: - import execSync from child_process - isQmdConfigured(), qmdSearchMode(), qmdCommand() helpers reading from state.api.config.memory (openclaw.json) - qmdSearch shells out to qmd <mode> <query> --json -n <limit> with 30s timeout, maps qmd's file/line fields to Honcho's path/startLine format - qmdGet shells out to qmd get <path> for qmd:// URIs - search() runs QMD in parallel with Honcho, merges results, clamps session scores to 0.5 so file results sort above them - readFile() handles qmd:// paths before falling back to Honcho - status() reports provider: 'honcho+qmd' when QMD is active
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
WalkthroughAdds optional QMD CLI integration to the Honcho memory runtime: concurrent QMD searches merged ahead of Honcho results with clamped Honcho scores; readFile handles ChangesQMD Memory Backend Integration
Docs, small behavioral changes, and tooling tweaks
Sequence Diagram(s)sequenceDiagram
participant Client
participant MemoryRuntime
participant HonchoSession
participant QMD_CLI
Client->>MemoryRuntime: search(query, maxResults)
MemoryRuntime->>HonchoSession: collectTranscriptMatches(query)
MemoryRuntime->>QMD_CLI: execFile("qmd", ["search", "--mode", ...])
QMD_CLI-->>MemoryRuntime: stdout JSON results
HonchoSession-->>MemoryRuntime: honchoResults
MemoryRuntime->>MemoryRuntime: clampHonchoScores(<=0.5), prepend QMD results, slice(limit)
MemoryRuntime-->>Client: mergedResults
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
ESLint skipped: no ESLint configuration detected in root package.json. To enable, add Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@runtime.ts`:
- Around line 167-179: The qmdGet function uses execSync with a concatenated
command which allows command injection via the path parameter; change qmdGet to
call execFileSync (or execFile) using qmdCommand() as the executable and pass
['get', path] (without shell-concatenation/JSON.stringify) as the args array,
keep the same options (encoding, timeout, stdio) and preserve the try/catch
behavior so it returns stdout or null; update any import/require if needed and
ensure this mirrors the fix applied elsewhere (e.g., the readFile/relPath use)
to avoid shell interpretation of the path.
- Around line 185-210: Remove the duplicate earlier declaration and redundant
init: delete the first "let requestedSessionKey = typeof opts.sessionKey ===
\"string\" && opts.sessionKey.length > 0 ? opts.sessionKey : activeSessionKey ??
null;" so only the later const requestedSessionKey remains, and remove the
earlier redundant await state.ensureInitialized() so state.ensureInitialized()
is called only once before resolving participantPeer (keep the single await near
where participantPeer is resolved); ensure all references use the remaining
requestedSessionKey and that state initialization still happens before
state.resolveSessionParticipantPeer/getParticipantPeer.
- Around line 143-165: The qmdSearch function currently builds a shell string
and calls execSync, which allows command injection via the query; replace the
execSync call with execFileSync and pass qmdCommand() as the executable plus an
arguments array containing qmdSearchMode(), JSON.stringify(query), "--json",
"-n", String(limit) (so the shell is not used), preserve encoding, timeout and
stdio options, then parse stdout as before and keep the same mapping logic and
error handling in qmdSearch.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
13 tests covering: - isQmdConfigured detection (4 config scenarios) - qmd search field mapping and command construction (4 scenarios) - score clamping in merged results (2 scenarios) - qmd:// path handling in readFile (2 scenarios) - fallback behavior when QMD fails (1 scenario)
There was a problem hiding this comment.
🧹 Nitpick comments (1)
runtime.ts (1)
185-188: ⚡ Quick winRemove duplicated
requestedSessionKeyassignment.The same expression is computed twice (Line 185 and Line 208). Keep a single
constassignment to avoid drift and unnecessary mutable state.Suggested cleanup
- let requestedSessionKey = - typeof opts.sessionKey === "string" && opts.sessionKey.length > 0 - ? opts.sessionKey - : activeSessionKey ?? null; + const requestedSessionKey = + typeof opts.sessionKey === "string" && opts.sessionKey.length > 0 + ? opts.sessionKey + : activeSessionKey ?? null; ... - requestedSessionKey = typeof opts.sessionKey === "string" && opts.sessionKey.length > 0 - ? opts.sessionKey - : activeSessionKey ?? null;Also applies to: 208-210
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@runtime.ts` around lines 185 - 188, The code computes requestedSessionKey twice causing duplicate/mutable state; consolidate to a single const by removing the later reassignment and using the initial const requestedSessionKey computed from opts.sessionKey and activeSessionKey (i.e., keep the expression typeof opts.sessionKey === "string" && opts.sessionKey.length > 0 ? opts.sessionKey : activeSessionKey ?? null as a single const) and update any subsequent logic to reference that const instead of reassigning requestedSessionKey.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@runtime.ts`:
- Around line 185-188: The code computes requestedSessionKey twice causing
duplicate/mutable state; consolidate to a single const by removing the later
reassignment and using the initial const requestedSessionKey computed from
opts.sessionKey and activeSessionKey (i.e., keep the expression typeof
opts.sessionKey === "string" && opts.sessionKey.length > 0 ? opts.sessionKey :
activeSessionKey ?? null as a single const) and update any subsequent logic to
reference that const instead of reassigning requestedSessionKey.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: f66c0d03-0146-4d33-be67-bff505cb43a4
📒 Files selected for processing (2)
runtime.tstest/qmd-bridge.test.ts
…undant ensureInitialized Addresses PR review feedback: - qmdSearch: replace execSync shell string with execFileSync, passing [cmd, args...] array to prevent command injection via query parameter - qmdGet: replace execSync shell string with execFileSync, passing ["get", path] array to prevent injection via path parameter - search(): remove duplicate let requestedSessionKey declaration and redundant early ensureInitialized() call; single const declaration remains after the QMD promise block - Keep unused execSync import removed
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
runtime.ts (1)
276-286:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winError message is misleading when
qmd getfails.When
qmdGet(relPath)returnsnull(CLI failure), the error thrown says "Unsupported Honcho memory path" — but theqmd://path format is supported; the QMD CLI invocation simply failed. This could cause confusion during debugging.🛠️ Proposed fix: distinguish QMD retrieval failure from unsupported path
if (typeof relPath === "string" && relPath.startsWith("qmd://")) { const qmdText = qmdGet(relPath); if (qmdText !== null) { return { path: relPath, text: sliceLines(qmdText, params.from, params.lines), source: "qmd", }; } - throw new Error(`Unsupported Honcho memory path: ${relPath}`); + throw new Error(`Failed to retrieve QMD content for path: ${relPath}`); }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@runtime.ts` around lines 276 - 286, The error thrown when qmdGet(relPath) returns null is misleading; instead of throwing "Unsupported Honcho memory path" (which implies the qmd:// scheme isn't supported), change the code in the qmd:// handling branch to throw a clear retrieval failure that includes the relPath (and any available error info) — e.g. "Failed to retrieve qmd content for <relPath>" — so callers know the QMD CLI/lookup failed; update the throw in the block that checks qmdGet(relPath) and keep the returned object using sliceLines(qmdText, params.from, params.lines) unchanged.
🧹 Nitpick comments (1)
test/qmd-bridge.test.ts (1)
137-146: 💤 Low valueUnused variable
statebeforememState.Line 138 creates
statebut it's never used —memStateis created immediately after and passed to the manager. This appears to be leftover code from an earlier iteration.♻️ Proposed fix
it("returns honcho+qmd when qmd block exists (even empty)", async () => { - const state = createState({ memory: createMemoryConfig({ command: undefined, searchMode: undefined }) }); - // Empty qmd block — isQmdConfigured checks mem?.qmd truthiness - // But qmd is { includeDefaultMemory: true } which is truthy - // So for this test, set qmd to {} const memState = createState({ memory: { backend: "qmd", qmd: {} } }); const { manager } = await getHonchoMemorySearchManager(memState, { agentId: "main" }); const s = manager.status(); expect(s.provider).toBe("honcho+qmd"); });🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/qmd-bridge.test.ts` around lines 137 - 146, In the test "returns honcho+qmd when qmd block exists (even empty)" there is an unused variable `state` created via createState(...) that is never used; remove that leftover `state` declaration and keep only `memState` (the one passed into getHonchoMemorySearchManager), ensuring the test uses `memState` exclusively and no unused local `state` remains.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@runtime.ts`:
- Line 146: The CLI argument for the qmd search is being built with
JSON.stringify(query), which causes extra quoting when using execFileSync;
update the call that builds args for execFileSync (where qmdCommand(),
qmdSearchMode(), JSON.stringify(query), "--json", "-n", String(limit) are used)
to pass the query value directly (or ensure it's a plain string) instead of
JSON.stringify; if query can be a non-string object, serialize it outside in a
way the qmd CLI expects (e.g., convert to the intended string form) before
passing it as an argument to execFileSync.
---
Outside diff comments:
In `@runtime.ts`:
- Around line 276-286: The error thrown when qmdGet(relPath) returns null is
misleading; instead of throwing "Unsupported Honcho memory path" (which implies
the qmd:// scheme isn't supported), change the code in the qmd:// handling
branch to throw a clear retrieval failure that includes the relPath (and any
available error info) — e.g. "Failed to retrieve qmd content for <relPath>" — so
callers know the QMD CLI/lookup failed; update the throw in the block that
checks qmdGet(relPath) and keep the returned object using sliceLines(qmdText,
params.from, params.lines) unchanged.
---
Nitpick comments:
In `@test/qmd-bridge.test.ts`:
- Around line 137-146: In the test "returns honcho+qmd when qmd block exists
(even empty)" there is an unused variable `state` created via createState(...)
that is never used; remove that leftover `state` declaration and keep only
`memState` (the one passed into getHonchoMemorySearchManager), ensuring the test
uses `memState` exclusively and no unused local `state` remains.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: f53c4ef2-4957-42d7-b726-0a8172f203f4
📒 Files selected for processing (2)
runtime.tstest/qmd-bridge.test.ts
execFileSync doesn't invoke a shell so JSON.stringify adds unwanted literal quote characters to the query string. Pass query directly.
execFileSync doesn't invoke a shell so JSON.stringify adds unwanted literal quote characters. Pass query directly.
We also added 13 unit tests for the QMD bridge in 038492d — covers config detection, field mapping, score clamping, qmd:// path handling, and fallback behavior. All tests passing alongside the existing 28 tests (41 total). Resolved in 992c0d2 — added JSDoc docstrings to search(), readFile(), and status() methods. |
|
Review rate limit: 2/2 reviews remaining, available now. |
|
@coderabbitai review |
✅ Actions performedReview triggered.
|
![r0c1n4nte[bot]](https://github.githubassets.com/images/gravatars/gravatar-user-420.png){kind=avatar}
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
runtime.ts (1)
316-325:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winMisleading error when QMD CLI retrieval fails for a valid qmd:// path.
When
qmdGet(relPath)returnsnull(CLI timeout, non-zero exit, file not found, etc.) for a qmd:// URI that already passed theallowedPrefixescheck, the error thrown isUnsupported Honcho memory path: .... The path is supported — the underlying retrieval failed. This conflates a config/path-format error with a runtime CLI failure and complicates debugging in production.🐛 Proposed fix: differentiate the qmd:// runtime-failure error
const qmdText = await qmdGet(relPath); if (qmdText !== null) { return { path: relPath, text: sliceLines(qmdText, params.from, params.lines), source: "qmd", }; } - throw new Error(`Unsupported Honcho memory path: ${relPath}`); + throw new Error(`Failed to read QMD memory path: ${relPath}`); }Note:
test/qmd-bridge.test.tsline ~296 asserts/Unsupported Honcho memory path/for the qmd-get-fails case; that regex will also need updating to match the new message.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@runtime.ts` around lines 316 - 325, The current code throws "Unsupported Honcho memory path: relPath" when qmdGet(relPath) returns null, which obscures that the qmd:// path passed allowedPrefixes but the CLI retrieval failed; change the thrown error inside the qmdGet null branch to a runtime-failure message (e.g., "Failed to retrieve qmd:// path: <relPath> — qmd CLI returned no content" and include any available error/exit info) so it distinguishes config/format errors from CLI/runtime failures; update the test (test/qmd-bridge.test.ts) that matches /Unsupported Honcho memory path/ to expect the new message or a more general regex for "Failed to retrieve" instead.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@runtime.ts`:
- Around line 316-325: The current code throws "Unsupported Honcho memory path:
relPath" when qmdGet(relPath) returns null, which obscures that the qmd:// path
passed allowedPrefixes but the CLI retrieval failed; change the thrown error
inside the qmdGet null branch to a runtime-failure message (e.g., "Failed to
retrieve qmd:// path: <relPath> — qmd CLI returned no content" and include any
available error/exit info) so it distinguishes config/format errors from
CLI/runtime failures; update the test (test/qmd-bridge.test.ts) that matches
/Unsupported Honcho memory path/ to expect the new message or a more general
regex for "Failed to retrieve" instead.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c093cf02-8f9e-4a71-bc0a-b62bc377d621
📒 Files selected for processing (2)
runtime.tstest/qmd-bridge.test.ts
5d4492a to
1633b16
Compare
1633b16 to
c5b83d2
Compare
|
@coderabbitai resume |
✅ Actions performedReviews resumed. |
There was a problem hiding this comment.
🧹 Nitpick comments (1)
test/qmd-bridge.test.ts (1)
155-164: ⚡ Quick winDrop the unused
statesetup in the "even empty" test.
stateon line 156 is created but never used — onlymemStateis consumed by the test. The dead line plus the inline rationale comments make the test harder to read.🧹 Proposed cleanup
it("returns honcho+qmd when qmd block exists (even empty)", async () => { - const state = createState({ memory: createMemoryConfig({ command: undefined, searchMode: undefined }) }); - // Empty qmd block — isQmdConfigured checks mem?.qmd truthiness - // But qmd is { includeDefaultMemory: true } which is truthy - // So for this test, set qmd to {} - const memState = createState({ memory: { backend: "qmd", qmd: {} } }); - const { manager } = await getHonchoMemorySearchManager(memState, { agentId: "main" }); + // isQmdConfigured() only requires mem.qmd to be truthy; an empty object qualifies. + const state = createState({ memory: { backend: "qmd", qmd: {} } }); + const { manager } = await getHonchoMemorySearchManager(state, { agentId: "main" }); const s = manager.status(); expect(s.provider).toBe("honcho+qmd"); });🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/qmd-bridge.test.ts` around lines 155 - 164, Remove the dead setup in the unit test "returns honcho+qmd when qmd block exists (even empty)": delete the unused local variable state created via createState(...) and its inline rationale comments, leaving only the memState = createState({ memory: { backend: "qmd", qmd: {} } }) setup and the calls to getHonchoMemorySearchManager and manager.status(); ensure createState, memState and getHonchoMemorySearchManager remain unchanged so the test still asserts expect(s.provider).toBe("honcho+qmd").
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@test/qmd-bridge.test.ts`:
- Around line 155-164: Remove the dead setup in the unit test "returns
honcho+qmd when qmd block exists (even empty)": delete the unused local variable
state created via createState(...) and its inline rationale comments, leaving
only the memState = createState({ memory: { backend: "qmd", qmd: {} } }) setup
and the calls to getHonchoMemorySearchManager and manager.status(); ensure
createState, memState and getHonchoMemorySearchManager remain unchanged so the
test still asserts expect(s.provider).toBe("honcho+qmd").
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 9b4966bb-7899-45ae-b198-2fb87ec1ba25
📒 Files selected for processing (2)
runtime.tstest/qmd-bridge.test.ts
- Docstrings: bring coverage to 82.5% across all source files (config.ts, helpers.ts, index.ts, peers.ts, runtime.ts, state.ts, hooks/*, tools/*) - Fix: cast config access in qmdAllowedPrefixes for TS type compat - Fix: capture custom_message events (was data loss bug) - Fix: validate timestamps with Number.isFinite() before new Date() - Fix: guard against empty sanitized peer IDs - Fix: optimize isParticipantPeerId() from O(n) to O(1) with Set cache - Fix: skip wasted Honcho API call when search limit already filled - Fix: handle broken symlinks in uniqueWorkspacePaths() - Fix: validate NaN from parseInt/parseFloat in CLI search - Fix: remove non-null assertion on agentPeer.chat() result - Fix: replace console.warn with api.logger.warn for log routing - Fix: remove unused minScore param from memory_search schema - Fix: handle empty sanitized sender IDs in resolveParticipantPeerId()
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (2)
commands/cli.ts (1)
492-492: ⚡ Quick winConsider enforcing topK upper bound for consistency with tool schema.
The validation checks
topK > 0but allows arbitrarily large values. The tool schema intools/search.tsenforcesmaximum: 100. For consistency, consider clamping topK to 100.♻️ Proposed refinement
-const searchTopK = Number.isFinite(topK) && topK > 0 ? topK : 10; +const searchTopK = Number.isFinite(topK) && topK > 0 ? Math.min(topK, 100) : 10;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@commands/cli.ts` at line 492, The current calculation of searchTopK allows arbitrarily large topK values (const searchTopK = Number.isFinite(topK) && topK > 0 ? topK : 10), which conflicts with the tool schema maximum of 100; change the logic to clamp the value to the schema bound by ensuring searchTopK = Math.min(100, Number.isFinite(topK) && topK > 0 ? topK : 10) (or equivalent) so values above 100 are reduced to 100; reference the variable searchTopK and the input topK and keep the fallback of 10 when topK is invalid.state.ts (1)
180-184: 💤 Low valueFallback format differs from peers.ts — consider aligning.
This secondary fallback uses
sender-${sha256...}(hyphen, 16 chars) whilepeers.tsusessender_${base64url...}(underscore, 32 chars). If both fallbacks ever fire for the same input (e.g., due to a future regression), the samechannelPeerIdwould map to different Honcho peer IDs. Consider aligning the format withpeers.tsfor consistency, or removing this guard if thepeers.tsfallback is considered sufficient.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@state.ts` around lines 180 - 184, The fallback peer-id generation in this file uses `sender-${createHash("sha256")...slice(0,16)}` which differs from the format used in peers.ts; update the fallback to match peers.ts (use the same base64url-encoded SHA256 slice with an underscore prefix) so identical inputs produce identical Honcho peer IDs. Concretely, in the branch after `resolveParticipantPeerId(...)` update the `resolvedPeerId` assignment to replicate peers.ts' logic (use the same base64url(sha256(channelPeerId)) encoding and slicing and the `sender_` prefix) instead of the current hyphenated hex slice; keep the guard but ensure you reference the same encoding/util used in peers.ts to maintain consistency between `resolveParticipantPeerId`, this fallback, and peers.ts.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@state.ts`:
- Around line 211-225: The participantPeerIdCache is built lazily by
rebuildParticipantPeerIdCache() but never cleared when the participantPeers map
changes, causing isParticipantPeerId() to return stale results; update all
places that mutate state.participantPeers (e.g., the code paths calling
participantPeers.set(...) and participantPeers.delete(...) or any add/remove
handlers) to invalidate the cache by setting participantPeerIdCache = null after
the mutation so rebuildParticipantPeerIdCache() will rebuild an up-to-date Set
on the next isParticipantPeerId(...) call.
---
Nitpick comments:
In `@commands/cli.ts`:
- Line 492: The current calculation of searchTopK allows arbitrarily large topK
values (const searchTopK = Number.isFinite(topK) && topK > 0 ? topK : 10), which
conflicts with the tool schema maximum of 100; change the logic to clamp the
value to the schema bound by ensuring searchTopK = Math.min(100,
Number.isFinite(topK) && topK > 0 ? topK : 10) (or equivalent) so values above
100 are reduced to 100; reference the variable searchTopK and the input topK and
keep the fallback of 10 when topK is invalid.
In `@state.ts`:
- Around line 180-184: The fallback peer-id generation in this file uses
`sender-${createHash("sha256")...slice(0,16)}` which differs from the format
used in peers.ts; update the fallback to match peers.ts (use the same
base64url-encoded SHA256 slice with an underscore prefix) so identical inputs
produce identical Honcho peer IDs. Concretely, in the branch after
`resolveParticipantPeerId(...)` update the `resolvedPeerId` assignment to
replicate peers.ts' logic (use the same base64url(sha256(channelPeerId))
encoding and slicing and the `sender_` prefix) instead of the current hyphenated
hex slice; keep the guard but ensure you reference the same encoding/util used
in peers.ts to maintain consistency between `resolveParticipantPeerId`, this
fallback, and peers.ts.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 65bb9c30-96f2-4add-a82b-6a1b0a9dd0df
📒 Files selected for processing (16)
commands/cli.tsconfig.tshelpers.tshooks/capture.tshooks/context.tshooks/gateway.tsindex.tspeers.tsruntime.tsstate.tstools/ask.tstools/context.tstools/memory-passthrough.tstools/message-search.tstools/search.tstools/session.ts
💤 Files with no reviewable changes (1)
- tools/memory-passthrough.ts
✅ Files skipped from review due to trivial changes (8)
- index.ts
- tools/message-search.ts
- hooks/gateway.ts
- tools/session.ts
- hooks/capture.ts
- hooks/context.ts
- config.ts
- tools/search.ts
…mat, clamp searchTopK - Invalidate participantPeerIdCache after every participantPeers.set() so isParticipantPeerId() does not return stale results - Use base64url SHA-256 fallback in state.ts to match peers.ts format - Clamp CLI search topK to max 100 per Honcho schema bound
The function required both backend === "qmd" AND a truthy mem.qmd
sub-object, but a minimal config like "memory": {"backend": "qmd"}
(without an explicit qmd block) broke the entire QMD bridge.
All downstream qmd helper functions use optional chaining and handle
undefined sub-objects fine.
1 participant
When memory.backend: 'qmd' is configured, the Honcho plugin's memory_search/memory_get tools only return Honcho session transcripts. This patch adds parallel QMD query execution so indexed local files (such as wiki docs) appear in results alongside session data.
Changes:
This was developed with AI assistance and tested on live infrastructure.
Summary by CodeRabbit
New Features
Bug Fixes
Tests
Tools
Documentation