databricks-vector-search — Claude Code skill: what it does
Skill for creating Databricks vector indexes and querying embeddings in Delta tables. Useful for RAG applications and semantic search.
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a claude-codeWhat databricks-vector-search does
The skill teaches the agent to work with Databricks Vector Search: create endpoints (STANDARD or STORAGE_OPTIMIZED), build three index types (Delta Sync with automatic embedding generation, Delta Sync with precomputed vectors, Direct Access for manual control), and query results by text or vector. The agent can configure hybrid search (ANN + BM25) — helpful when queries contain exact terms like SKUs or error codes. Filters for Standard endpoints use dictionary format; Storage-Optimized endpoints require SQL-like syntax via the databricks-vectorsearch package.
Delta Sync indexes auto-sync with the source Delta table. TRIGGERED mode runs sync on demand, CONTINUOUS tracks updates. Direct Access indexes let you add, update, and delete vectors via upsert_data_vector_index and delete_data_vector_index. Storage-Optimized endpoints handle over a billion vectors but latency is higher — 300–500ms versus 20–50ms for Standard.
What it helps with
- create a vector index from a Delta table with automatic embedding generation
- run semantic search by text or vector
- configure hybrid search with BM25 for exact terms
- sync an index with a Delta table in real time
- choose between Standard and Storage-Optimized for billions of vectors
Example prompts for the agent
- Create a vector index
catalog.schema.docs_indexon endpointmy-vs-endpointfrom tablecatalog.schema.documents, embed columncontentviadatabricks-gte-large-en - Query index
catalog.schema.my_indexfor 10 nearest results by text "What is machine learning?", returnid,content, andmetadata - Upsert two records with
id,text,embedding, andmetadatainto Direct Access indexcatalog.schema.direct_indexusingupsert_data_vector_index
How to install databricks-vector-search
1Skills CLI (Vercel)
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a claude-code2Manually: put the files in place
~/.claude/skills/databricks-vector-search/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/databricks-vector-search/
1Skills CLI (Vercel)
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a codex2Manually: put the files in place
.agents/skills/databricks-vector-search/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a cursor2Manually: put the files in place
.cursor/skills/databricks-vector-search/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install databricks/databricks-agent-skills databricks-vector-search2Manually: put the files in place
.github/skills/databricks-vector-search/SKILL.md1Skills CLI (Vercel)
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a gemini-cli2Platform CLI
gemini skills install https://github.com/databricks/databricks-agent-skills --path skills/databricks-vector-searchManually: put the files in place
.devin/skills/databricks-vector-search/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about databricks-vector-search
How to install databricks-vector-search in Claude Code?
Run npx skills add databricks/databricks-agent-skills --skill databricks-vector-search -a claude-code, the skill will appear in ~/.claude/skills/databricks-vector-search/SKILL.md.
What does databricks-vector-search do?
The skill lets the agent create Databricks vector indexes for semantic search and RAG, configure endpoints, sync indexes with Delta tables, and query results by text or vector.
When to choose Storage-Optimized over Standard endpoint?
Storage-Optimized handles over a billion vectors and costs seven times less, but latency is 300–500ms versus 20–50ms for Standard. Best for large archives where response time is not critical.
