feat: new MCP tools — get/list/update drawer, hook settings, export (resolves #635) by bensig · Pull Request #667 · MemPalace/mempalace

bensig · 2026-04-12T03:58:38Z

Summary

Merge-conflict resolution of #635 by @jphein against develop (after #647 landed).

All code is from #635 — conflicts in mcp_server.py resolved:

WAL write: kept os.open atomic approach from Security hardening: input validation, argument whitelisting, concurrency & WAL fixes #647
tool_search: kept _MAX_RESULTS + min_similarity compat from feat: new MCP tools — get/list/update drawer, hook settings, export #635
schema_props filter: removed duplicate block (already present above)

Fixes #646 (0 drawers on claude.ai export) via paginated exporter.py.

Original PR: #635
Original author: @jphein — all feature code in this PR is their work.

Co-authored-by: jphein [email protected]

Test plan

CI passes
mempalace export produces correct output on large palaces

…, search limits Infrastructure hardening for the MCP server: - Detect palace DB replacement via inode tracking (repair command support) - WAL rotation to prevent unbounded WAL growth - _fetch_all_metadata() + _get_cached_metadata() with 60s TTL for taxonomy/status - _MAX_RESULTS cap (100) with limit clamping [1, _MAX_RESULTS] - max_distance parameter for similarity threshold in search - Handle all notifications/* methods, null arguments, method=None - Remove duplicate _client_cache = None declarations - searcher.py max_distance parameter passthrough Co-Authored-By: Claude Opus 4.6 <[email protected]>

…filed), export, normalize New MCP tools: - mempalace_get_drawer: fetch single drawer by ID with full content - mempalace_list_drawers: paginated listing with wing/room filter - mempalace_update_drawer: update content/wing/room on existing drawers - mempalace_hook_settings: get/set hook behavior (silent_save, desktop_toast) - mempalace_memories_filed_away: check latest checkpoint status Also includes: - exporter.py: export palace as browsable markdown files - normalize.py: tool_use/tool_result capture for richer transcript mining - layers.py: updated for new tool integration - config.py: hook settings properties (hook_silent_save, hook_desktop_toast) Depends on PR 3 (reliability) for _MAX_RESULTS, _metadata_cache, WAL logging. Co-Authored-By: Claude Opus 4.6 <[email protected]>

Co-Authored-By: Claude Opus 4.6 <[email protected]>

- Handle explicit null in MCP params (request.get("params") or {}) - Fix search tool description: L2 → cosine distance (collection uses hnsw:space=cosine) - Guard against empty string key in tool_use_map from malformed JSONL entries Co-Authored-By: Claude Opus 4.6 <[email protected]>

Co-Authored-By: Claude Opus 4.6 <[email protected]>

- WAL write: keep os.open atomic approach from #647 - tool_search: keep _MAX_RESULTS + min_similarity compat from #635 - schema_props filter: remove duplicate (already present above conflict)

bensig · 2026-04-12T04:07:23Z

Code review

Found 3 issues:

min_similarity backwards-compat shim passes raw similarity value as max_distance, but these are opposite scales (higher similarity = stricter, lower distance = stricter). Passing min_similarity=0.8 results in max_distance=0.8 which is permissive, not strict. Should convert: dist = (1 - min_similarity) if min_similarity is not None else max_distance.

mempalace/mempalace/mcp_server.py

Lines 574 to 576 in 4f54529

    
                   "limit": limit, 
        
               } 
        
           except Exception as e:

tool_status, tool_list_wings, tool_list_rooms, and tool_get_taxonomy replace structured error reporting ({"error": str(e), "partial": True}) with bare except: pass. This directly reverts the error surfacing added in the recently merged Security hardening: input validation, argument whitelisting, concurrency & WAL fixes #647 security hardening. Callers now get empty/incomplete results with no signal something went wrong.

mempalace/mempalace/mcp_server.py

Lines 426 to 434 in 4f54529

    
           try: 
        
               wing = sanitize_name(wing, "wing") 
        
               room = sanitize_name(room, "room") 
        
               content = sanitize_content(content) 
        
           except ValueError as e: 
        
               return {"success": False, "error": str(e)} 
        
           col = _get_collection(create=True) 
        
           if not col:

_get_client inode cache: when os.stat fails (file missing), current_inode = 0. The condition current_inode and current_inode != _palace_db_inode short-circuits on falsy 0, so a missing/deleted database silently reuses the stale cached client instead of reconnecting.

mempalace/mempalace/mcp_server.py

Lines 317 to 321 in 4f54529

    
                   pass 
        
               return {"taxonomy": taxonomy} 
        
           def tool_search(

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}

1. min_similarity backwards-compat: convert similarity to distance scale (1.0 - similarity) instead of passing raw value as max_distance 2. Restore structured error reporting (error + partial fields) in tool_status, tool_list_wings, tool_list_rooms, tool_get_taxonomy — reverts silent except:pass that dropped #647 security hardening 3. inode cache: remove falsy-zero short-circuit so missing DB file triggers reconnect instead of reusing stale client 4. _fetch_all_metadata: check for empty batch before extending/advancing offset to prevent infinite loop on concurrent deletion 5. KG initialization: only override path when --palace is explicit; default runs use KnowledgeGraph's built-in default path Co-authored-by: jphein <[email protected]>

…_path KnowledgeGraph() uses its own default db_path which may differ from config.palace_path when --palace isn't specified. Only override when the user explicitly provides --palace. Per review feedback from bensig in MemPalace#667. Co-Authored-By: Claude Opus 4.6 <[email protected]>

jphein · 2026-04-12T04:34:49Z

Great catches, all 5. I've incorporated every fix into my fork:

min_similarity conversion — embarrassing but fair, 1.0 - similarity is obviously correct for cosine distance. I only ever used max_distance directly so this never bit me in production.
Structured error reporting — your result["partial"] = True pattern is much better than the bare except: pass I had. Already adopted it across my fork.
Inode zero guard — subtle one. The current_inode and ... short-circuit was masking the missing-DB case. My fix: detect mtime changes in _get_collection to prevent stale HNSW index #663 (stale HNSW) PR had an explicit os.path.isfile() guard that worked around it, but removing the falsy check is the right fix at the root.
Pagination guard — classic TOCTOU race. Moving the empty check before extend prevents the infinite loop on concurrent deletion.
KG init — reverted in my fix: standalone bug fixes (#534 #535 #536 #538 #570 #572 #637) #626 too. KnowledgeGraph() knows its own default path.

Thanks for the thorough review — learned something on each one.

…t count 701 - README: mark MemPalace#635 as merged via MemPalace#667, fix cosine (not L2) distance - CLAUDE.md: test count 692→701, drawer count 132K→134K, replace stale upstream PR list with current table, add fork change MemPalace#22 (bensig's security/pagination/error reporting fixes from MemPalace#667) Co-Authored-By: Claude Opus 4.6 <[email protected]>

Merge origin/develop into codex/pg-backend-stage2 after upstream PRs MemPalace#647 and MemPalace#667 landed. Resolve the MCP server conflict by preserving the new cached metadata helpers, WAL hardening, drawer update/list tools, and exporter changes from develop while keeping the PostgreSQL backend abstraction: MCP collection access now goes through palace.get_collection(...) instead of constructing a Chroma PersistentClient directly. Keep the large-table optimization for PostgreSQL: metadata pagination no longer performs an exact col.count() pre-pass, MCP status derives total_drawers from fetched metadata rows, and MemoryStack.status() can use backend-provided estimated_count() so PostgreSQL can use catalog/statistics estimates rather than SELECT COUNT(*) on status paths. Also make tool_update_drawer backend-neutral by using upsert() instead of the Chroma-only update() method, with tests covering the estimated count path. Verification: uv run pytest -q (671 passed, 106 deselected); uv run ruff check .; git diff --check; git diff --cached --check; precise conflict-marker scan.

Required by upstream's tool_update_drawer (PR MemPalace#667). LanceCollection implements update as get+upsert (re-embeds on content change). ChromaCollection delegates to chromadb's native update.

- Add 5 missing tools: get_drawer, list_drawers, update_drawer, hook_settings, memories_filed_away (shipped in MemPalace#667) - Add max_distance and context params to mempalace_search - Add distance field to search response examples - Update tool count from 19 to 24 Co-Authored-By: Claude Opus 4.6 <[email protected]>

- Update 19 → 24 in four places (intro, heading, file table, tree) - Add get_drawer, list_drawers, update_drawer to Palace (write) table - Add Settings section with hook_settings and memories_filed_away Co-Authored-By: Claude Opus 4.6 <[email protected]>

- Add 5 missing tools: get_drawer, list_drawers, update_drawer, hook_settings, memories_filed_away (shipped in MemPalace#667) - Add max_distance and context params to mempalace_search - Add distance field to search response examples - Update tool count from 19 to 24 Co-Authored-By: Claude Opus 4.6 <[email protected]>

- Update 19 → 24 in four places (intro, heading, file table, tree) - Add get_drawer, list_drawers, update_drawer to Palace (write) table - Add Settings section with hook_settings and memories_filed_away Co-Authored-By: Claude Opus 4.6 <[email protected]>

Remove three stale rows that are already in upstream/develop: - Epsilon mtime comparison (merged upstream as PR MemPalace#610, Arnold Wender) - .jsonl READABLE_EXTENSIONS addition (upstream-authored at SHA 560fdbd) - max_distance threshold (superset already in upstream via MemPalace#667) All three moved to "Superseded by upstream" with attribution. Add two rows that were fork-ahead but missing from the table: - cmd_export and cmd_purge CLI commands (cli.py, PR pending) - mempal_precompact_hook.sh transcript auto-mining: session ID parsing, find-by-session-id fallback, inline chunk_exchanges → upsert (PR pending) Annotate existing rows: - PID guard: branch pr/pid-file-guard (local, not pushed) - Save hook Python auto-detection: expand files list to include both .claude-plugin/ hooks (same pattern was applied there) Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

jphein and others added 6 commits April 11, 2026 19:47

fix: normalize.py handles string messages and Read offset type mismatch

a88de9e

Co-Authored-By: Claude Opus 4.6 <[email protected]>

fix: rename ambiguous var 'l' to 'line' (E741 lint)

248555d

Co-Authored-By: Claude Opus 4.6 <[email protected]>

merge: resolve conflicts with develop

4f54529

- WAL write: keep os.open atomic approach from #647 - tool_search: keep _MAX_RESULTS + min_similarity compat from #635 - schema_props filter: remove duplicate (already present above conflict)

bensig requested review from igorls and milla-jovovich as code owners April 12, 2026 03:58

bensig mentioned this pull request Apr 12, 2026

feat: new MCP tools — get/list/update drawer, hook settings, export #635

Closed

8 tasks

bensig merged commit 20c8f8e into develop Apr 12, 2026

This was referenced Apr 12, 2026

bug: _try_claude_ai_json parser silently produces 0 drawers on claude.ai export (conversations.json) #646

Closed

MCP server stdio transport: add_drawer and kg_add write to WAL but not to ChromaDB/SQLite #538

Open

This was referenced Apr 12, 2026

fix: standalone bug fixes (#534 #535 #536 #538 #570 #572 #637) #626

Closed

fix: add wing param to diary_write/diary_read, derive from transcript path #659

Merged

Ari4ka approved these changes Apr 12, 2026

View reviewed changes

jphein mentioned this pull request Apr 12, 2026

feat: deterministic hook saves — zero data loss via silent Python API #673

Merged

6 tasks

z3tz3r0 mentioned this pull request Apr 12, 2026

fix: handle large claude.ai exports and multi-conversation "messages" key #676

Open

3 tasks

This was referenced Apr 12, 2026

feat: add LanceDB backend abstraction and migration path #574

Open

feat: add multi-device sync on LanceDB backend #575

Open

This was referenced Apr 12, 2026

fix: defer KnowledgeGraph init to first use in MCP server #167

Closed

[codex] docs add safe hook wrapper guidance #505

Closed

jphein mentioned this pull request Apr 12, 2026

docs: update MCP tools reference for #667 additions #738

Closed

2 tasks

igorls mentioned this pull request Apr 13, 2026

release: v3.2.0 #762

Merged

4 tasks

igorls deleted the merge/635-resolved branch April 13, 2026 22:37

mvalentsev mentioned this pull request Apr 18, 2026

feat: mempalace_search should return drawer_id (or add delete_by_query) — no way to clean superseded drawers programmatically #993

Closed

mvalentsev mentioned this pull request Apr 24, 2026

fix(mcp): lazy per-path KnowledgeGraph cache (#1136) #1160

Merged

8 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: new MCP tools — get/list/update drawer, hook settings, export (resolves #635)#667

feat: new MCP tools — get/list/update drawer, hook settings, export (resolves #635)#667
bensig merged 7 commits intodevelopfrom
merge/635-resolved

bensig commented Apr 12, 2026 •

edited

Loading

Uh oh!

bensig commented Apr 12, 2026

Uh oh!

jphein commented Apr 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

bensig commented Apr 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

bensig commented Apr 12, 2026

Code review

Uh oh!

jphein commented Apr 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

bensig commented Apr 12, 2026 •

edited

Loading