vector-db-ingest

Vector DB Ingest (vector-db-ingest)

Ingests and indexes repository files into the ChromaDB vector store for semantic retrieval.

Contents

• Critical Constraints
• Quick start
• Workflow
• Verification

Critical Constraints

1. Profile Sovereignty: Always specify --profile to ensure the correct manifest and batch configuration are loaded.
2. Lock Concurrency: Ensure no concurrent process holds a lock on the database folder during ingestion.
3. Transparency: Report target profile, indexed file count, and any processing errors.

Quick start

Run incremental ingestion for the default wiki profile (last 24 hours):

python3 scripts/ingest.py --profile wiki --since 24

Workflow

1. Prerequisite Check: Ensure profiles exist in .agent/learning/vector_profiles.json (or initialize via vector-db-init).
2. Execute Ingest: Run scripts/ingest.py with --profile and appropriate target scope:
   • Specific file: python3 scripts/ingest.py --profile wiki --file path/to/file.md
   • Specific directory: python3 scripts/ingest.py --profile wiki --folder path/to/folder
   • Incremental: python3 scripts/ingest.py --profile wiki --since 24
   • Full rebuild: python3 scripts/ingest.py --profile wiki --full
3. Report Output: State target profile, number of chunks created, and total runtime.

Verification

Confirm indexed content is immediately retrievable via semantic query:

python3 scripts/query.py "test query" --profile wiki --limit 3