vector-db-init

Vector DB Initialization (vector-db-init)

Prepares the local environment and manifests for ChromaDB vector embeddings and semantic search.

Contents

• Critical Constraints
• Quick start
• Operational Parameters
• Workflow
• Verification
• References

Critical Constraints

1. In-Process Default: Operates directly on disk via chroma_data_path; no daemon required unless multi-process concurrency is needed.
2. Dependency Install: Requires chromadb and sentence-transformers available in the python environment.
3. Manifest Sovereignty: Configuration persists in .agent/learning/vector_profiles.json.

Quick start

Execute automatic profile scaffolding for local vector storage:

python3 scripts/init.py

Operational Parameters

Settings configured in .agent/learning/vector_profiles.json:

• chroma_host: Empty string for in-process direct disk mode; IP:port for server mode.
• batch_size: Embedding batch size (default: 1,000 files).
• embedding_model: nomic-ai/nomic-embed-text-v1.5.
• parent_chunk_size: 2,000 chars; child_chunk_size: 400 chars.

Workflow

1. Dependency Verification: Ensure dependencies are installed in virtualenv.
2. Target Discovery: Identify project directories to index (e.g. docs/, plugins/).
3. Manifest Generation: Write .agent/learning/vector_knowledge_manifest.json.
4. Scaffold Profile: Run python3 scripts/init.py to create the profile configuration.
5. Next Steps: Guide user to run vector-db-ingest to index content.

Verification

Confirm configuration file generation and validity:

test -f .agent/learning/vector_profiles.json && echo "Profile configured"

References

• acceptance-criteria.md — Acceptance criteria for vector database bootstrapping.
• vector-db-bootstrap-guide.md — Step-by-step initialization manual.