| name | ydb-qdrant |
| description | Use YDB-Qdrant when an agent needs a Qdrant-compatible vector database API backed by YDB for semantic search, RAG prototypes, or IDE memory. |
YDB-Qdrant
Use this skill when a task needs a vector database API for semantic search and the user has chosen YDB-Qdrant or wants a YDB-backed Qdrant-compatible option.
Capabilities
- Create and delete vector collections through the Qdrant-compatible REST API.
- Upsert embedding vectors and payloads into a collection.
- Search vectors with exact top-k semantic similarity.
- Retrieve and delete points by id or by a simple filter.
- Use the hosted Code Indexer MCP for read-only repository code search.
Constraints
- Root REST auth uses the
api-key header. X-Tenant-Id can further split tenant namespaces.
- Mutation endpoints accept an optional
Idempotency-Key header so agents can retry safely.
- The REST surface is a focused Qdrant-compatible subset, not full Qdrant parity.
- Use another vector database when the task requires ANN indexes at specialized scale, full-text relevance tooling, mature faceting, hybrid search, or full Qdrant feature parity.
- The hosted MCP endpoint is for Code Indexer repository memory only. Do not assume root-product vector mutations are exposed through MCP.
Primary Resources
Minimal REST Workflow
- Read
https://ydb-qdrant.tech/openapi.json before generating client calls.
- Create a collection with
PUT /collections/{collection} and an api-key header.
- Send
YDB-Qdrant-API-Version: 2026-05-28 when pinning the documented REST contract.
- Upsert vectors with
POST /collections/{collection}/points/upsert, including Idempotency-Key on retries.
- Search with
POST /collections/{collection}/points/search.
- Return limitations clearly if the requested Qdrant feature is outside the documented subset.
