chore(release): 3.5.0

[igorls]

Release prep for 3.5.0 (previous tag: v3.4.1).

Version bump (six sources + lockfile)

• mempalace/version.py, pyproject.toml, .claude-plugin/marketplace.json, .claude-plugin/plugin.json, .codex-plugin/plugin.json (the five version-guard checks) + the README.md badge, and uv.lock.
• Refreshed the "N MCP tools" prose from 34 → 35 everywhere (delete_by_source feat(mcp): add mempalace_delete_by_source bulk-cleanup tool (#1722) #1729 and checkpoint feat: add mempalace_checkpoint batch save tool #1851 each registered a tool). The website/reference/mcp-tools.md table already lists all 35; test_readme_claims stays green.

What's in 3.5.0

33 PRs since v3.4.1. Highlights (full notes in CHANGELOG.md):

Features — opt-in local write daemon (#1826); opt-in HTTP transport for the MCP server, loopback-default with Host/Origin DNS-rebinding guards + optional bearer token (#1801, #1806); mempalace_checkpoint batch-save tool (#1851); mempalace_delete_by_source bulk cleanup, dry-run-by-default, purges drawers and their closets (#1722, #1729); source_file search filter (#1815, #1817); Continue.dev / Gemini CLI / Pi transcript parsers (#731, #204, #169); C#/PHP/Swift/Kotlin/Java miner coverage (#1819, #1368, #1720); final mine on plugin SessionEnd (#1814, #1820).

Performance — overview/status + graph_stats answered from the SQLite aggregate to kill large-palace timeouts (#1748, #1379); embedder ORT thread cap (#1068); SQL-pushdown pagination for sqlite_exact/pgvector + single-scroll Qdrant metadata (#1842, #1840, #1832).

Bug fixes — pgvector tolerates surrogate/NUL transcript bytes (#1833, #1829); percent-encoded SQLite RO URIs; stale-HNSW divergence → SQLite fallback (#1822); diverged-index recovery points at repair --mode from-sqlite not a data-losing re-mine (#1843, #1847, #1849); MCP refuses a second writer (#1823); Windows hook miner uses CREATE_NO_WINDOW (#1848).

⚠️ Open item before tagging

test-windows is red on develop itself (not introduced by this branch): tests/test_migrate_wings.py::test_migrate_relabels_old_format_wings hits chromadb ... Failed to apply logs to the hnsw segment writer — the known ChromaDB-on-Windows HNSW compaction flake. migrate.py and that test are unchanged since v3.4.1. All other jobs (Linux 3.9/3.11/3.13, macOS, build, build-gpu, lint) are green, and the full suite passes locally (3102 passed). This needs a flaky-vs-deterministic call before we tag.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[gemini-code-assist]

Code Review

This pull request bumps the MemPalace version from 3.4.1 to 3.5.0 across all configuration files, plugins (Claude, Codex, and Cursor), and documentation, reflecting the addition of a new MCP tool (bringing the total to 35). It also updates the CHANGELOG.md with comprehensive release notes detailing new features such as an opt-in local daemon, HTTP transport for the MCP server, new batch-save and bulk-cleanup tools, performance optimizations, and various bug fixes. There are no review comments to assess, and I have no additional feedback to provide.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.