Rebuild the memory SQLite index from observations.jsonl and scratchpad findings.
npx skills add https://github.com/paulingham/.claude --skill reindex-memory
Please help me install this Agent Skill. Skill: reindex-memory Source: https://github.com/paulingham/.claude/tree/main/skills/reindex-memory Command: npx skills add https://github.com/paulingham/.claude --skill reindex-memory
Rebuilds ~/.claude/db/memory.sqlite — a derived SQLite index with FTS5 — from the canonical learning/{project_hash}/observations.jsonl files and scratchpad findings. The JSONL files remain the source of truth; this DB is a fast, queryable read index used by Stories 3-5 (/harness:recall, MCP memory server, and embedding similarity search).
Run the reindex.py script to rebuild memory from the learning observations.
High-intent search queries and answers about installing and using this skill.
Rebuilding a SQLite memory index from JSONL observations involves running a reindex script that reconstructs the FTS5 database from learning observations and scratchpad findings to restore recall accuracy.
Schema drift in a memory index happens when the database schema version changes after upgrades or learning cycles, requiring a rebuild that recreates data tables while preserving existing embeddings whenever possible.
Yes, you can reindex memory offline using only the Python standard library, making the operation suitable for fresh installs and reproducible reindexes without requiring external dependencies or network connectivity.
Memory recall search returns stale or missing results when the SQLite index has not been updated to reflect newly added learning observations, requiring an idempotent reindex to refresh the queryable read index.
The reindex process preserves existing embeddings during schema changes by recreating data tables only when the current schema version requires it, maintaining idempotence throughout the rebuild operation.
0xBigBossExtract Claude Code and Codex session JSONL transcripts into Markdown.
msd-dev-labStore and retrieve user memories across conversations with categorized folders.
peteromalletBuild, tune, and query FAISS indices for vector similarity search.
carrotwaxrGuide Prisma ORM schema design and operations for SQLite databases.
