| name | cv-db-rebuild |
| description | Rebuild cv-agent.db from CV copy files, sources.md, and ccrecall databases. Use when database is lost, corrupted, or needs full reconstruction. |
CV Agent Database Rebuild
Reconstruct data/cv-agent.db from source files using
mcp-sqlite-tools.
Quick Start
- Open/create DB:
open_database at
/home/scott/repos/cv/data/cv-agent.db
- Create schema (see references/schema.sql)
- Populate from sources (see
references/data-sources.md)
- Build FTS5 search index across all tables
- SCP to prod (see memory/MEMORY.md for push process)
Table Population Order
- profile — from
src/lib/copy/basics.md + +layout.svelte
- roles — from
src/lib/copy/work.md, early-webdev-exp.md,
non-webdev-exp.md
- achievements — from
src/lib/copy/work.md highlights +
memory/sources.md
- projects — from
src/lib/copy/projects.md +
memory/sources.md GitHub stars
- skills — from
src/lib/copy/skills.md with evidence strings
- events — from
memory/sources.md speaking section
- content_stats — from
memory/sources.md Fathom/blog/GitHub
stats
- qa_pairs — curate ~25-30 recruiter Q&A pairs across categories:
about, skills, experience, community, projects, logistics
- work_evidence — aggregate stats from ccrecall databases (see
references)
- search_index — FTS5 index across ALL content from ALL tables
Critical Rules
- Use mcp-sqlite-tools for all DB operations
- ALWAYS checkpoint WAL before SCP push: mcp-sqlite-tools uses WAL
mode. Edits go to
.db-wal, not the main file. Run
PRAGMA wal_checkpoint(TRUNCATE) before any file-based transfer.
- Push process: stop container → SCP → delete WAL/SHM on prod → start
- XtendOps ccrecall (
~/Downloads/ccrecall.db) contains customer PII
— extract AGGREGATE stats ONLY
- Personal ccrecall (
~/.claude/ccrecall.db) — per-project session
counts, tools, date ranges
- Use en-GB spellings throughout
- After rebuild, verify with:
SELECT COUNT(*) FROM search_index
(expect ~140 rows)
References
