Semantic search skill for retrieving code and documentation from the ChromaDB vector store. Use when you need concept-based search across the repository (Phase 2 of the 3-phase search protocol). V2 includes L4/L5 retrieval constraints.
From vector-dbnpx claudepluginhub richfrem/agent-plugins-skills --plugin vector-dbThis skill is limited to using the following tools:
acceptance-criteria.mdassets/resources/architecture_sequence.mmdassets/resources/deployment_model.mmdassets/resources/rag_design_choices.mdassets/resources/stabilizers/README.mdassets/resources/stabilizers/vector_consistency_check.mdevals/evals.jsonevals/results.tsvfallback-tree.mdreferences/acceptance-criteria.mdreferences/fallback-tree.mdreferences/logic.mmdreferences/unpacking.mmdreferences/workflow.mmdrequirements.inrequirements.txtscripts/ingest.pyscripts/init.pyscripts/operations.pyscripts/query.pyGuides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Designs KPI dashboards with metrics selection (MRR, churn, LTV/CAC), visualization best practices, real-time monitoring, and hierarchy for executives, operations, and product teams.
Transforms raw data into narratives with story structures, visuals, and frameworks for executive presentations, analytics reports, and stakeholder communications.
This skill requires Python 3.8+ and standard library only. No external packages needed.
To install this skill's dependencies:
pip-compile ./requirements.in
pip install -r ./requirements.txt
See ./requirements.txt for the dependency lockfile (currently empty — standard library only).
Semantic (meaning-based) search against the ChromaDB vector store. Use for Phase 2 of the 3-phase search protocol -- after the RLM Summary Ledger (Phase 1) returns insufficient results.
| Script | Role |
|---|---|
scripts/query.py | Semantic search -- CLI entry point |
scripts/operations.py | Core Parent-Child retrieval library |
scripts/vector_config.py | Profile config helper (vector_profiles.json) |
scripts/vector_consistency_check.py | Integrity validation |
Write operations (ingest, cleanup) are handled by dedicated agents: vdb-ingest, vdb-cleanup.
curl -sf http://127.0.0.1:8110/api/v1/heartbeat
If connection refused: run vector-db-launch skill (.agents/skills/vector-db-launch/SKILL.md).
For first-time setup: run vector-db-init skill (scripts/init.py).
Profiles are project-defined in vector_profiles.json (see vector-db-init skill). Any number can exist. Discover what's available:
cat .agent/learning/vector_profiles.json
Common default is knowledge -- your project may define more (e.g. separate profiles for code vs docs). When topic is ambiguous, search all profiles.
python3 ./scripts/query.py \
"your natural language question" --profile knowledge --limit 5
Results include ranked parent chunks with RLM Super-RAG context pre-injected.
Do not cat, strings, or sqlite3 the .vector_data/ directory.
Binary blobs will corrupt your context window and the retrieval pipeline.
All access goes through query.py. No exceptions.
When search returns empty results, explicitly state:
> Not Found in Vector Store
> Searched profile: [profile_name] for "[query]"
> Profile covers: [scope]
> Not searched: [out-of-scope areas]