Conversation
… onnx, restore KV.state The v0.9.1 dist was dead on arrival. Three independent bugs layered: 1) `getContext` was re-imported during PR #111 conflict resolution in three files (compress.ts, disk-size-manager.ts, image-quota-cleanup.ts). iii-sdk v0.11 dropped that export — the shim in src/logger.ts documents the fix. Tests mocked iii-sdk so `npm test` passed, but `node dist/cli.mjs` crashed on first import. Removed the imports, switched to the module-level logger. 2) The two PR #111 files also used the `registerFunction({ id, description }, handler)` shape. iii-sdk 0.11.2 only accepts `registerFunction(id, handler, options?)`. The dist imports succeeded but the runtime registration threw "not a function" on the test mock. Collapsed to the flat form. 3) `KV.state` was dropped from src/state/schema.ts during the same conflict resolution, which broke the disk-size-manager's persistence key. Restored `state: "mem:state"` — used only by this manager. 4) tsdown was inlining onnxruntime-node and @xenova/transformers into dist/. That rewrites the relative require() inside `onnxruntime-node/dist/binding.js` so it can't find `../bin/napi-v3/darwin/arm64/onnxruntime_binding.node` from dist/. Even without AGENTMEMORY_IMAGE_EMBEDDINGS the module graph still evaluated the bundled binding.js on startup and blew up. Added both packages (plus onnxruntime-web and the two Anthropic SDKs) to `external:` in tsdown.config.ts. Bundle shrank from 6.1 MB to 1.9 MB. The CLIP / local embedding providers lazy-load them from node_modules where relative paths work. 5) Bumped iii-sdk 0.11.0 → 0.11.2 to match the API currently shipped (Logger / durable:subscriber / durable:publisher / TriggerAction.void). 6) test/multimodal.test.ts used the old `{ id, description }` mock shape — rewrote the four registerFunction mocks to match the real `(id, cb)` signature. 812/812 pass. End-to-end smoke test with AGENTMEMORY_SLOTS=true AGENTMEMORY_REFLECT=true: - livez → 200 ok - GET /slots → all 8 defaults seeded into correct scopes (persona / user_preferences / tool_guidelines in global; rest in project) - POST /slot (invalid sizeLimit / unknown scope) → 400 with specific error - POST /slot/append overflow → 413 with currentSize + sizeLimit - POST /slot/reflect on empty session → no-op - GET /audit → every slot write + reflect emits an audit row - POST /vision-search without flag → 503 disabled
Policy: never commit lock files. Downstream developers and CI regenerate them at install time. Committing them creates churn on every transitive-dep bump and masks real dependency changes in PRs.
- import-jsonl: run synthetic compression + BM25 index per observation
(previously raw obs never got indexed; search/summaries never saw them)
- import-jsonl + observe: capture first userPrompt as session.firstPrompt
so sessions list shows what each session was about, not just an ID
- api::audit: wrap bare array in { entries, success } — viewer expected
{ entries } shape and was rendering empty despite populated data
- api::replay::sessions: call kv.list directly instead of trigger hop
(was timing out at 10s+ with 600+ sessions)
- viewer sessions tab: show firstPrompt preview under each session
- viewer memories tab: fix strength display (7 * 100 = 700%; now scaled
properly for both 0-1 and 0-10 representations, clamped to 100)
- viewer memories tab: add explainer card describing what memories are
- viewer WS: add 5s connect timeout so stuck WS upgrades fall through
to polling mode instead of showing CONNECTING... forever
import-jsonl now runs lightweight heuristic extraction on imported sessions so the Lessons and Crystals tabs populate from replay data instead of staying empty until a user manually invokes the MCP tools. - Lessons: regex-match imperative patterns in user/assistant text (always/never/don't/prefer/avoid/caveat) and save top 20 per session with low confidence (0.4) and an 'auto-import' tag so they're clearly derived rather than hand-curated. - Crystals: one crystal per imported session — narrative = first user prompt or observation preview, keyOutcomes = distinct tool names, filesAffected = extracted file paths, lessons = derived lesson IDs. - Actions stay deliberately empty (user-initiated goals, not tool logs). - Semantic/procedural still require consolidation pipeline summaries.
- Crystals tab: add header card explaining what crystals are; redesign each card with narrative, pills (project/session/counts), tool badges, files list, and resolved lesson content (fetched via /lessons and mapped by id) instead of showing raw lesson UUIDs. - Lessons tab: add header card explaining the rule/confidence/decay model so users understand the low-confidence auto-import tag. - Sessions tab: clicking a session now fetches its observations and renders a full detail panel — preview callout, 4-card grid (observations / tools / files / duration), top tool bar chart, activity type breakdown, file list, metadata block, actions row. - Pull in lessons when loading crystals so auto-generated lessons are actually readable in context.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
✅ Files skipped from review due to trivial changes (1)
📝 WalkthroughWalkthroughAdds synthetic observation compression and lesson/crystal derivation during replay import; extends Session with Changes
Sequence Diagram(s)sequenceDiagram
participant Import as JSONL Import
participant SessionKV as Session KV
participant ObsKV as Observation KV
participant Compress as Compression Module
participant Index as Search Index
participant Deriver as Lesson/Crystal Deriver
Import->>Import: Parse JSONL into raw observations
Import->>Import: Find earliest non-empty userPrompt -> firstPrompt
Import->>SessionKV: Conditionally persist firstPrompt if missing
loop for each raw observation
Import->>Compress: buildSyntheticCompression(raw)
Compress-->>Import: syntheticObs
Import->>ObsKV: Store syntheticObs under sessionId
Import->>Index: add(syntheticObs)
end
Import->>Deriver: deriveCrystalAndLessons(rawObsList, syntheticObsList, firstPrompt)
Deriver->>Deriver: Extract lesson texts, create KV.lessons (content-addressed)
Deriver->>Deriver: Aggregate lesson IDs/files -> KV.crystals
Deriver-->>Import: return created lesson/crystal IDs
Estimated code review effort🎯 4 (Complex) | ⏱️ ~75 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/functions/image-quota-cleanup.ts (1)
56-68:⚠️ Potential issue | 🟠 MajorSkip eviction when the ref-count lookup fails.
Line 61 logs the failure but leaves
refCountat0, so a transient KV/read error falls through todeleteImage()for an image that may still be referenced.🛡️ Proposed fix
try { refCount = await getImageRefCount(kv, f.filePath); } catch (err) { logger.error("Failed to read refCount", { filePath: f.filePath, error: err instanceof Error ? err.message : String(err) }); + return; } if (refCount > 0) { return; }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/functions/image-quota-cleanup.ts` around lines 56 - 68, The ref-count read failure leaves refCount at 0 and causes deleteImage to run; update the withKeyedLock callback (surrounding getImageRefCount in src/functions/image-quota-cleanup.ts) so that if getImageRefCount throws you do not proceed with eviction: on catch of the error from getImageRefCount, log the error as you do and then return early (or rethrow) from the callback to skip deleteImage; reference the withKeyedLock callback, getImageRefCount, refCount variable, and deleteImage to locate and change the logic.
🧹 Nitpick comments (2)
tsdown.config.ts (1)
25-31: onnxruntime packages are purely transitive dependencies; no direct imports found.Verification found no direct imports of
onnxruntime-nodeoronnxruntime-webanywhere in the source code—only in theexternalarray oftsdown.config.ts. This means the concern about non-lazy code paths triggeringMODULE_NOT_FOUNDerrors does not apply. These packages are only ever accessed transitively through@xenova/transformers, which is already optional and lazy-loaded (seeclip.ts).For clarity and explicit intent, consider declaring
onnxruntime-nodeandonnxruntime-webasoptionalDependenciesalongside@xenova/transformers, but this is not required.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tsdown.config.ts` around lines 25 - 31, The external array in tsdown.config.ts lists "onnxruntime-node" and "onnxruntime-web" even though there are no direct imports and they are only transitively used via "@xenova/transformers"; either remove those two packages from the external: [...] declaration to avoid misleading configuration, or document the intent with an inline comment and instead add them as optionalDependencies in package.json alongside "@xenova/transformers" to make their optional/transitive nature explicit; update the external array and package.json accordingly and ensure the change references the external array symbol in tsdown.config.ts and the optionalDependencies section in package.json.src/state/schema.ts (1)
48-48: Add a typed interface for the newKV.statescope.
KV.statenow stores state entries such assystem:currentDiskSize, butsrc/types.tsdoes not define the corresponding value shape/key map. Add a small exported interface so callers do not keep using ad-hocnumbergenerics for this scope. Based on learnings: When adding new KV scopes, add to the KV object in src/state/schema.ts and add corresponding interface in src/types.ts.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/state/schema.ts` at line 48, Add a typed interface for the KV.state scope and export it from src/types.ts so callers stop using ad-hoc number generics; define an interface (e.g., KVState or StateKV) that maps known keys like "system:currentDiskSize" to their value types (number) and export it, then update usages to use that interface as the generic for KV.state. Ensure the KV.state entry remains in src/state/schema.ts (state: "mem:state") and wire the new exported interface name into any type declarations referencing KV.state so the scope has a proper key->value shape.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/functions/observe.ts`:
- Around line 206-211: The code pushes a sanitized firstPrompt using
raw.userPrompt without ensuring it's actually a string; add a runtime guard
(e.g. typeof raw.userPrompt === "string") before calling replace/trim/slice in
the observe flow so non-string truthy values won't throw. Update the conditional
that currently reads (!session.firstPrompt && raw.userPrompt) to validate
stringness and only push the update to the updates array when raw.userPrompt is
a string; keep the same sanitization (replace(/\s+/g," ").trim().slice(0,200))
but only apply it after the type check so functions like replace/trim are never
called on non-strings.
In `@src/functions/replay.ts`:
- Around line 112-156: Replace non-deterministic IDs with content-addressed IDs
and upserts: for lessons, stop using generateId("lesson")—use
fingerprintId("lsn", content.trim().toLowerCase()) to derive lessonId, check
kv.get(KV.lessons, lessonId) and only create/kv.set if missing (or merge/update
existing) and push the canonical lessonId into lessonIds; for crystals, avoid
generateId("crystal") by either fingerprinting on sessionId (e.g.,
fingerprintId("crl", sessionId)) or first querying existing crystals for this
sessionId and reusing that id, then perform an upsert via kv.set(KV.crystals,
crystalId, crystal) so repeated imports are idempotent (update existing records
instead of always creating new ones).
In `@src/viewer/index.html`:
- Around line 2686-2691: The search currently only uses raw c.lessons (IDs) in
the items filter, so visible lesson text doesn't match; update the filter inside
the items = items.filter(...) to resolve each lesson ID via lessonMap (use
state.crystals.lessonMap) and include the lesson content/string (e.g.,
lessonMap[id].content or the appropriate field) when building the haystack,
replacing or augmenting (c.lessons || []).join(' ') so the combined string uses
the resolved lesson text for matching.
---
Outside diff comments:
In `@src/functions/image-quota-cleanup.ts`:
- Around line 56-68: The ref-count read failure leaves refCount at 0 and causes
deleteImage to run; update the withKeyedLock callback (surrounding
getImageRefCount in src/functions/image-quota-cleanup.ts) so that if
getImageRefCount throws you do not proceed with eviction: on catch of the error
from getImageRefCount, log the error as you do and then return early (or
rethrow) from the callback to skip deleteImage; reference the withKeyedLock
callback, getImageRefCount, refCount variable, and deleteImage to locate and
change the logic.
---
Nitpick comments:
In `@src/state/schema.ts`:
- Line 48: Add a typed interface for the KV.state scope and export it from
src/types.ts so callers stop using ad-hoc number generics; define an interface
(e.g., KVState or StateKV) that maps known keys like "system:currentDiskSize" to
their value types (number) and export it, then update usages to use that
interface as the generic for KV.state. Ensure the KV.state entry remains in
src/state/schema.ts (state: "mem:state") and wire the new exported interface
name into any type declarations referencing KV.state so the scope has a proper
key->value shape.
In `@tsdown.config.ts`:
- Around line 25-31: The external array in tsdown.config.ts lists
"onnxruntime-node" and "onnxruntime-web" even though there are no direct imports
and they are only transitively used via "@xenova/transformers"; either remove
those two packages from the external: [...] declaration to avoid misleading
configuration, or document the intent with an inline comment and instead add
them as optionalDependencies in package.json alongside "@xenova/transformers" to
make their optional/transitive nature explicit; update the external array and
package.json accordingly and ensure the change references the external array
symbol in tsdown.config.ts and the optionalDependencies section in package.json.
🪄 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: defaults
Review profile: CHILL
Plan: Pro
Run ID: e235bdfd-5ecf-4d42-bc9f-9e31dec407ce
⛔ Files ignored due to path filters (1)
package-lock.jsonis excluded by!**/package-lock.json
📒 Files selected for processing (13)
.gitignorepackage.jsonsrc/functions/compress.tssrc/functions/disk-size-manager.tssrc/functions/image-quota-cleanup.tssrc/functions/observe.tssrc/functions/replay.tssrc/state/schema.tssrc/triggers/api.tssrc/types.tssrc/viewer/index.htmltest/multimodal.test.tstsdown.config.ts
All six findings verified against current code on this branch; all valid.
1. observe.ts firstPrompt (L206-211) — raw.userPrompt is typed as
optional string but populated from payload.data which is unknown at
runtime; a non-string truthy value (array, object) would crash at
.replace(). Guard with typeof === 'string' before sanitizing, and
skip the update when the trimmed value is empty.
2. replay.ts deriveCrystalAndLessons — generateId('lesson') /
generateId('crystal') produced fresh IDs on every import, so
re-importing the same JSONL duplicated lessons and crystals. Switch
to fingerprintId:
- lessons keyed by fingerprintId('lesson', content.toLowerCase()) +
merge on existing (bump reinforcements, append sessionId to
sourceIds, union tags).
- crystals keyed by fingerprintId('crystal', sessionId) + preserve
original createdAt and sourceActionIds when upserting.
3. viewer/index.html crystals filter (L2686-2691) — search haystack
previously joined raw lesson IDs (lesson_xxx), so lesson text never
matched. Resolve each ID via state.crystals.lessonMap and include
resolved content in the haystack; also fold in filesAffected and
project so the search feels complete.
4. image-quota-cleanup.ts (L56-68) — when getImageRefCount threw, the
refCount variable stayed at its default 0 and the code happily
proceeded to deleteImage(). That could delete a still-referenced
image on any transient KV error. Fail-closed: log and return from
the withKeyedLock callback on error so deleteImage is never reached
without a confirmed refCount === 0.
5. types.ts / disk-size-manager.ts — add a StateScope interface that
maps KV.state scope keys to their value types
(system:currentDiskSize: number), export StateScopeKey, and have
disk-size-manager use StateScope[typeof DISK_SIZE_KEY] in kv.get /
kv.set generics instead of the ad-hoc <number>.
6. tsdown.config.ts + package.json — onnxruntime-node / onnxruntime-web
are not imported directly; they come in transitively via
@xenova/transformers. Keep them in tsdown's external array (bundling
them breaks the relative .node binding paths in dist/) and add both
to package.json optionalDependencies alongside @xenova/transformers
so the optional/lazy nature is explicit. Expand the tsdown comment
to reference the new optionalDependencies entries and name the exact
src files that lazy-import them.
812/812 tests pass.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
src/types.ts (1)
855-860: Drop the WHAT-style comment here.
StateScopeand"system:currentDiskSize"are already self-descriptive, and this comment mostly restates the shape/usage.Proposed cleanup
-/** - * KV.state scope — long-lived system counters + flags keyed by string. - * Keep keys/types in sync with the state-scope callers (e.g., - * disk-size-manager) so TypeScript enforces consistent value shapes - * instead of every caller using ad-hoc `<number>` generics. - */ export interface StateScope {As per coding guidelines, “Avoid code comments explaining WHAT — use clear naming instead”.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/types.ts` around lines 855 - 860, Remove the redundant WHAT-style comment block above StateScope that describes the KV.state scope and the "system:currentDiskSize" key; since StateScope and the key name are self-descriptive, delete the comment lines and leave the type definition and any brief doc that explains WHY only if necessary, ensuring symbols StateScope and "system:currentDiskSize" remain unchanged.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@src/types.ts`:
- Around line 855-860: Remove the redundant WHAT-style comment block above
StateScope that describes the KV.state scope and the "system:currentDiskSize"
key; since StateScope and the key name are self-descriptive, delete the comment
lines and leave the type definition and any brief doc that explains WHY only if
necessary, ensuring symbols StateScope and "system:currentDiskSize" remain
unchanged.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 46e6f2af-5a43-45ba-b0c1-c57242439454
📒 Files selected for processing (8)
package.jsonsrc/functions/disk-size-manager.tssrc/functions/image-quota-cleanup.tssrc/functions/observe.tssrc/functions/replay.tssrc/types.tssrc/viewer/index.htmltsdown.config.ts
✅ Files skipped from review due to trivial changes (1)
- src/functions/observe.ts
🚧 Files skipped from review as they are similar to previous changes (6)
- package.json
- src/functions/disk-size-manager.ts
- tsdown.config.ts
- src/functions/image-quota-cleanup.ts
- src/functions/replay.ts
- src/viewer/index.html
Follow-up to CodeRabbit review on #188. detectProvider's agent-sdk fallback at L93-97 hardcoded maxTokens: 4096, ignoring the maxTokens variable computed from env["MAX_TOKENS"] at L45. Every other provider branch (minimax, anthropic, gemini, openrouter) uses the computed value; the hardcoded one was the odd one out. Replace with the maxTokens identifier for consistency. Rejecting the two lockfile-related findings: .gitignore on this branch lists package-lock.json / pnpm-lock.yaml / yarn.lock, so the workflow comments ("Lockfiles are gitignored at the repo level") are accurate and the two-step install (npm install --package-lock-only then npm ci) is the correct pattern. Reverting to a single npm ci would fail CI because there is no committed lockfile to install from.
rohitg00
added a commit
that referenced
this pull request
Apr 22, 2026
Pulls in #184 runtime fix, #188 viewer pipeline, lockfile gitignore, CI two-step install, StateScope types, firstPrompt typeof guard, content-addressed lesson/crystal IDs, image-quota fail-closed, and onnxruntime optionalDependencies. Resolves workflow conflicts flagged by CodeRabbit — ci.yml / publish.yml on main are already the correct two-step install pattern. # Conflicts: # .github/workflows/publish.yml
This was referenced Apr 22, 2026
Merged
rohitg00
added a commit
that referenced
this pull request
Apr 22, 2026
- README provider table: surface the no-op default and call out the opt-in Claude-subscription fallback with AGENTMEMORY_ALLOW_AGENT_SDK (from #187) instead of listing 'Claude subscription' as the default. - README env block: document OPENAI_BASE_URL / OPENAI_EMBEDDING_MODEL (#186) and OPENAI_EMBEDDING_DIMENSIONS (#189), plus MINIMAX_API_KEY. - Hero stat-tests SVG: 654 -> 827 (both dark and light variants) to match current suite size after recursion guard + idempotent lesson/crystal tests + openai dimension tests landed. - website/lib/generated-meta.json regenerated by website/scripts/gen-meta.mjs (v0.9.1, 51 tools, 12 hooks, 119 endpoints, 848 tests).
Merged
rohitg00
added a commit
that referenced
this pull request
Apr 22, 2026
Rolls up #186 (OPENAI_BASE_URL / OPENAI_EMBEDDING_MODEL), #187 (Stop-hook recursion 5-layer defense + NoopProvider + AGENTMEMORY_ALLOW_AGENT_SDK opt-in), #188 (viewer empty-tabs + import-jsonl synthetic compression + auto-derived lessons/crystals + richer session detail + audit/replay/frontier shape fixes), #189 (OPENAI_EMBEDDING_DIMENSIONS + model-dimensions table), and #190 (README/website docs refresh). Bumps: package.json, plugin/.claude-plugin/plugin.json, src/version.ts, src/types.ts ExportData.version union, src/functions/export-import.ts supportedVersions, test/export-import.test.ts assertion, and packages/mcp/package.json shim (was stuck at 0.9.0).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes a pile of viewer issues where tabs rendered empty despite data existing, sessions looked like useless hex IDs with no context, and JSONL imports silently broke the search + consolidation pipelines.
Stacks on #184
Includes the runtime fix from #184 (
a12767bdrop getContext / externalize onnx / restore KV.state) and the lockfile gitignore (b1bb2b2). Merge #184 first, rebase drops to 3 commits. Kept bundled because without #184 the dist will not start and none of this UI work can be verified locally.What was broken
import-jsonlwroteRawObservationshape directly to KV, never ran synthetic compression, never indexed BM25. Result: imported sessions were invisible to search, consolidation reported "fewer than 5 summaries", memories / semantic / procedural stayed empty./agentmemory/auditreturned a bare array. Viewer expected{ entries, success }. Rendered empty despite populated data./agentmemory/replay/sessionswent viasdk.trigger → mem::replay::sessions, which timed out past ~400 sessions. Users with real history hit a 30s hang.m.strength = 7scored × 100 rendered as "700%".new WebSocket()has no browser-level timeout, so the banner stuck on "CONNECTING…" forever and the polling fallback never kicked in.lesson_xxxIDs instead of lesson content.What changes
Pipeline (src/functions/replay.ts, src/functions/observe.ts, src/types.ts, src/functions/summarize.ts ← N/A — left alone)
import-jsonl: runsbuildSyntheticCompressionon each parsed observation, writes compressed shape toKV.observations, adds to BM25 viagetSearchIndex().add(). Same path liveobserveuses, so imports behave like captured observations from Claude Code hooks.userPromptfrom parsed JSONL assession.firstPrompt. Liveobservedoes the same when it sees its firstprompt_submit.keyOutcomes= distinct tool titles,filesAffected= extracted file paths,lessons= derived lesson IDs.always/never/don't/prefer/avoid/caveat/note/warning), dedupe, save up to 20 per session withconfidence: 0.4, tagauto-import, sourceconsolidation.Sessioninterface getsfirstPrompt?+summary?fields.API shape fixes (src/triggers/api.ts)
api::auditwraps response as{ entries, success: true }so the viewer's expected shape is satisfied.api::replay::sessionsnow callskv.list<Session>(KV.sessions)+.sort()directly instead ofsdk.trigger. Sub-50ms on 600+ sessions.Viewer (src/viewer/index.html)
firstPrompt/summarypreview row under each item.lessonsfetch (stored instate.crystals.lessonMap).{ frontier: [...] }, not{ actions: [...] }.CONNECTINGafter that, forciblyws.close()→onclosefires → retry / polling-fallback chain takes over. Stops the forever "CONNECTING…" banner.Test plan
npm run build— cleannpm test— 812 / 812 passAGENTMEMORY_SLOTS=true AGENTMEMORY_REFLECT=true GRAPH_EXTRACTION_ENABLED=true CONSOLIDATION_ENABLED=true node dist/cli.mjs start+node dist/cli.mjs import-jsonl ~/.claude/projects/-Users-<user>/<uuid>.jsonlauto-importat 40% confidence.What is still empty (not a bug)
AGENTMEMORY_AUTO_COMPRESS=truewith a real API key (don't combine with the agent-sdk fallback; see fix(hooks): break Stop-hook infinite recursion via agent-sdk fallback #187 for why).Summary by CodeRabbit
New Features
Bug Fixes
Chores
Tests