Skip to content

Fix real-run CLI flows and formatting issues#1

Merged
quangdang46 merged 2 commits into
mainfrom
devin/1778269670-real-run-checks
May 9, 2026
Merged

Fix real-run CLI flows and formatting issues#1
quangdang46 merged 2 commits into
mainfrom
devin/1778269670-real-run-checks

Conversation

@quangdang46
Copy link
Copy Markdown
Owner

@quangdang46 quangdang46 commented May 9, 2026 •
edited
Loading

Summary

  • fix CLI search so mpr search prints ranked results
  • write project mempalace.json during init while skipping it during mining
  • fix generated MCP custom-palace command order
  • fix compress summary totals and Rust clippy/format issues
  • fix macOS/Windows CI compile issues in signal/PID handling

Verified

  • cargo fmt --all -- --check
  • cargo clippy --all-targets --all-features -- -D warnings
  • cargo test --all (369 passed)
  • cargo build --release
  • real CLI smoke run: init, mine, status, search, search --bm25, wake-up, compress --dry-run, split, repair scan, mcp, hook

quangdang46 and others added 2 commits May 9, 2026 09:27
@quangdang46 @devin-ai-integration
fix: harden real-run CLI flows
98ef1a7 
Co-Authored-By: Trần Quang Đãng <tranquangdang21@gmail.com>
@devin-ai-integration @quangdang46
fix: address platform CI checks
ca6d689 
Co-Authored-By: Trần Quang Đãng <tranquangdang21@gmail.com>
@quangdang46 quangdang46 merged commit 3f27ba9 into main May 9, 2026
4 checks passed
@quangdang46 quangdang46 deleted the devin/1778269670-real-run-checks branch May 9, 2026 13:20
quangdang46 added a commit that referenced this pull request Jun 2, 2026
@quangdang46
feat(telemetry): wire 3 call-site instrumentations (search/insert/llm)
93b1944 
Closes Oracle gap #4: telemetry had pre-registered descriptors
(telemetry.rs:99-135) but zero call sites. Minimum-viable
instrumentation: feature-gated, no behaviour change when
`telemetry` is off.

  - searcher.rs::search_memories_with_rerank
      counter!("mempalace_search_total", status=success)
      histogram!("mempalace_search_latency_ms")
  - palace_db.rs::add
      counter!("mempalace_insert_total", status=success)
  - llm/minimax_provider.rs::complete
      counter!("mempalace_llm_total_minimax")
      histogram!("mempalace_llm_latency_ms")

Pattern: `let _t0 = Instant::now()` at function entry, `.inspect()`
on the Ok return to record counter + histogram. Error paths
remain uninstrumented (deferred — a RAII guard or a pre-registered
Counter handle can capture both paths without per-call allocation).

Dynamic model label is deferred: the `metrics` macro requires
\&'static str for label values, and a dynamic String would force
a clone per call. Suffixing the counter name with the provider
(`..._minimax`) keeps providers distinguishable in Prometheus
without per-call allocation. The same pattern applies to
anthropic/openai when their call sites are instrumented.

Verification:
  - cargo build -p mempalace-core --features telemetry: clean
    (0 errors, 106 pre-existing warnings)
  - default build: unchanged (all instrumentation is
    #[cfg(feature = "telemetry")])

All 4 Oracle gaps owned by me are now closed:
  - #1 SYNONYM_BM25_WEIGHT -> RRF (a4a4cea)
  - #2 event-loop lag check (a4a4cea)
  - #3 9 missing connect adapters (98cffa2)
  - #4 telemetry call-site instrumentation (this commit)

The 6 failing tests are pre-existing in compress*/governance/heal/
mcp_server (touched a1743d4/f801525 by other agents) — not my
responsibility per AGENTS.md.
This was referenced Jun 3, 2026
Merged
Merged
quangdang46 added a commit that referenced this pull request Jun 3, 2026
@quangdang46 @claude
feat(palace): auto-migrate legacy metadata on every read/write
01f60a5 
PR #7 added the new typed Drawer fields (tags, trust, access_count,
last_accessed, reinforcements, superseded_by, active) plus a
`migrate_metadata` helper. The helper was left as opt-in
(consumers had to call it themselves), which meant legacy drawers
written by pre-PR #7 Palace versions would still be read in the
v0 shape (data in `metadata`, typed fields empty).

This PR wires the migration into every read/write path so the
self-healing is automatic and the typed fields are the source of
truth on disk from this point forward.

Wiring:

  - `Palace::add_drawer` (palace.rs:625) — calls
    `drawer.migrate_metadata()` once at the top, before any
    `store.upsert`. Every `add_drawer` write is now a
    self-healing write: legacy drawers that get re-written land
    in the v1 (typed-field) shape automatically.

  - `Palace::get_drawers` (palace.rs:710) — iterates the
    returned drawers and calls `migrate_metadata` on each
    before returning. Every reader (Layer 1 wake-up, status,
    PR #1 mutation methods via `default_mutate_drawer`,
    PR #2 tag/link, PR #4 recent) flows through this single
    chokepoint, so the migration propagates through the whole
    read surface.

  - `palace/store/usearch_sqlite.rs::get_drawer_by_id` and
    `::all_drawers` — call `migrate_metadata` right after
    constructing the Drawer from SQLite rows. Defensive: covers
    the case where a host calls `store.get_drawers()` directly
    (bypassing `Palace::get_drawers`).

  - `layers.rs::Layer1::generate` and `layers.rs::create_test_palace_db`
    (palace_db.rs read path) — same defensive coverage for
    the wake-up layer and tests that go through `PalaceDb::get_all`
    rather than the trait.

The `migrate_metadata` function is left as `pub fn` (not
demoted to `pub(crate)`) to avoid a SemVer-breaking change
on PR #7. Consumers can still call it explicitly if they want
to; they just don't have to.

Includes a cherry-pick of PR #7 (the `migrate_metadata` helper
and Drawer field extension) so this branch has the full
self-healing pipeline. The cherry-pick is a separate commit so
the two PRs are reviewable independently.

This is PR 24/8+ (post-series) — closes the design gap in PR #7
that was called out in the PR #7 review: migrate_metadata
should be mempalace's responsibility, not the consumer's.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@quangdang46 quangdang46 mentioned this pull request Jun 3, 2026
Merged
4 tasks
quangdang46 added a commit that referenced this pull request Jun 3, 2026
@quangdang46 @claude
feat(palace): auto-migrate legacy metadata on every read/write
eb9a6bd 
PR #7 added the new typed Drawer fields (tags, trust, access_count,
last_accessed, reinforcements, superseded_by, active) plus a
`migrate_metadata` helper. The helper was left as opt-in
(consumers had to call it themselves), which meant legacy drawers
written by pre-PR #7 Palace versions would still be read in the
v0 shape (data in `metadata`, typed fields empty).

This PR wires the migration into every read/write path so the
self-healing is automatic and the typed fields are the source of
truth on disk from this point forward.

Wiring:

  - `Palace::add_drawer` (palace.rs:625) — calls
    `drawer.migrate_metadata()` once at the top, before any
    `store.upsert`. Every `add_drawer` write is now a
    self-healing write: legacy drawers that get re-written land
    in the v1 (typed-field) shape automatically.

  - `Palace::get_drawers` (palace.rs:710) — iterates the
    returned drawers and calls `migrate_metadata` on each
    before returning. Every reader (Layer 1 wake-up, status,
    PR #1 mutation methods via `default_mutate_drawer`,
    PR #2 tag/link, PR #4 recent) flows through this single
    chokepoint, so the migration propagates through the whole
    read surface.

  - `palace/store/usearch_sqlite.rs::get_drawer_by_id` and
    `::all_drawers` — call `migrate_metadata` right after
    constructing the Drawer from SQLite rows. Defensive: covers
    the case where a host calls `store.get_drawers()` directly
    (bypassing `Palace::get_drawers`).

  - `layers.rs::Layer1::generate` and `layers.rs::create_test_palace_db`
    (palace_db.rs read path) — same defensive coverage for
    the wake-up layer and tests that go through `PalaceDb::get_all`
    rather than the trait.

The `migrate_metadata` function is left as `pub fn` (not
demoted to `pub(crate)`) to avoid a SemVer-breaking change
on PR #7. Consumers can still call it explicitly if they want
to; they just don't have to.

Includes a cherry-pick of PR #7 (the `migrate_metadata` helper
and Drawer field extension) so this branch has the full
self-healing pipeline. The cherry-pick is a separate commit so
the two PRs are reviewable independently.

This is PR 24/8+ (post-series) — closes the design gap in PR #7
that was called out in the PR #7 review: migrate_metadata
should be mempalace's responsibility, not the consumer's.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
quangdang46 added a commit that referenced this pull request Jun 3, 2026
@quangdang46 @claude
feat(palace): add MemoryProvider::recent for retention-ranked recall …
b36d375 
…(mp-migration 4/8)

jcode's `recall --mode recent` (no query) and the L1 wake-up
layer in jcode's ambient runner both need the same capability:
"give me the N most retention-relevant drawers" — the drawers
that have been accessed recently AND frequently, with Ebbinghaus-
style decay applied.

Add a default-implemented `recent(limit, scope)` method on
MemoryProvider that returns the top-N `SearchHit`s scored by:

    score = (access_count + 1) / (1 + days_since_last_accessed)

The default reads `metadata["access_count"]` and
`metadata["last_accessed"]` (set by the boost/touch path in
PR #1), pulls drawers via `get_drawers`, ranks, and trims to
`limit`. Scope filtering is respected (wing/room constraints
flow through to get_drawers).

Implementations that have a wired KG (mp-020 sub-task) should
override and use `kg.helpfulness_score(id)` — which already
factors in episodic memory feedback — for strictly better
ranking than the metadata-only default.

This is PR 4/8 in the jcode → mempalace Mode C library migration
series. No dependency on PRs 1/2/7 (recent only reads metadata,
doesn't mutate).

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
quangdang46 added a commit that referenced this pull request Jun 3, 2026
@quangdang46 @claude
feat(palace): add tag/untag/link/list_tags to MemoryProvider
6af3f0b 
jcode's MemoryManager exposes 4 graph-mutation methods that today
have no equivalent on the MemoryProvider trait:

  - tag(&id, tag)        — MemoryManager::tag_memory
  - untag(&id, tag)      — MemoryManager::untag_memory
  - link(&from, &to, w)  — MemoryManager::link_memories (with weight)
  - list_tags()          — closest jcode equiv: graph_stats.1

Add them as default-implemented methods on MemoryProvider. The
defaults use the metadata path (so this PR does NOT depend on the
KG being wired into Palace), mirroring the values in shapes that
match the eventual KG triples:

  tag    → metadata["tags"] (Vec<String>) + metadata["tag:<n>"] = true
  untag  → metadata["tags"] (minus removed) + remove metadata["tag:<n>"]
  link   → metadata["links"] = Vec<{target, weight}>
  list   → aggregate metadata["tags"] across get_drawers(None, None)

Implementations that have a wired KG (mp-020 sub-task) should
override and use KnowledgeGraph::add_triple / query_relationship
directly:

  tag     → add_triple { subject: drawer_id, predicate: "has_tag",
                          object: tag, current: true }
  link    → add_triple { subject: from,     predicate: "relates_to",
                          object: to, confidence: Some(weight) }
  list    → query_relationship(predicate="has_tag") then group-by

Includes a cherry-pick of PR #1 (the boost/decay/reinforce/
supersede/set_metadata additions) so the trait has all 9 mutation
methods together. Both can be reviewed independently because the
cherry-pick commit is its own clean diff in the series.

This is PR 2/8 in the jcode → mempalace Mode C library migration
series.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
quangdang46 added a commit that referenced this pull request Jun 3, 2026
@quangdang46 @claude
feat(palace): add tag/untag/link/list_tags to MemoryProvider (#20)
d842013 
jcode's MemoryManager exposes 4 graph-mutation methods that today
have no equivalent on the MemoryProvider trait:

  - tag(&id, tag)        — MemoryManager::tag_memory
  - untag(&id, tag)      — MemoryManager::untag_memory
  - link(&from, &to, w)  — MemoryManager::link_memories (with weight)
  - list_tags()          — closest jcode equiv: graph_stats.1

Add them as default-implemented methods on MemoryProvider. The
defaults use the metadata path (so this PR does NOT depend on the
KG being wired into Palace), mirroring the values in shapes that
match the eventual KG triples:

  tag    → metadata["tags"] (Vec<String>) + metadata["tag:<n>"] = true
  untag  → metadata["tags"] (minus removed) + remove metadata["tag:<n>"]
  link   → metadata["links"] = Vec<{target, weight}>
  list   → aggregate metadata["tags"] across get_drawers(None, None)

Implementations that have a wired KG (mp-020 sub-task) should
override and use KnowledgeGraph::add_triple / query_relationship
directly:

  tag     → add_triple { subject: drawer_id, predicate: "has_tag",
                          object: tag, current: true }
  link    → add_triple { subject: from,     predicate: "relates_to",
                          object: to, confidence: Some(weight) }
  list    → query_relationship(predicate="has_tag") then group-by

Includes a cherry-pick of PR #1 (the boost/decay/reinforce/
supersede/set_metadata additions) so the trait has all 9 mutation
methods together. Both can be reviewed independently because the
cherry-pick commit is its own clean diff in the series.

This is PR 2/8 in the jcode → mempalace Mode C library migration
series.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@quangdang46