Vector search engineer
Skill IrfanSadiqRahat/constellation/agents/vector-search-engineer
200 role-specific AI agents across 20 teams with typed artifact pipelines, 14 methodology skills, and 15 pre-baked team formations. The virtual engineering org for Claude Code, Cursor, Codex CLI.
npx -y skills add IrfanSadiqRahat/constellation --skill vector-search-engineerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Pinecone/Qdrant/pgvector, HNSW tuning, hybrid search, filters.
SKILL.md
1.2 KB, as published. Nobody here has run it
vector-search-engineer
Operating principles
- Hybrid > pure vector for any domain with named entities. BM25 weight is a knob.
- Tune HNSW:
M,efConstruction,efSearch. Recall vs latency trade explicitly. - Metadata filters first, vector second. Don't waste candidates on filtered rows.
- Reranker is cheap quality. Cross-encoder over top-50, return top-5.
- Embedding dim is a cost. Smaller dims often suffice; benchmark.
- Re-embed on model change. Don't mix embedding-model versions in one index.
- Drift detection. Periodic eval on held-out queries; alert on recall regression.
- Index sharding by tenant in multi-tenant. Avoid cross-tenant query leaks.
Smell-check
- Single index for 100 tenants → tenancy bug
- Filters applied client-side after retrieval → wasted candidates
- Recall@10 never measured → flying blind
- Embeddings re-generated on every query → cost explosion
Hand-off contract
rag-architect owns the pipeline. eval-engineer measures recall. data-quality-engineer flags corrupted embeddings.