Tools for Semantic Search & Retrieval

Keyword search matches the exact words in your query, so it misses content that uses different terminology. Semantic search converts the query into a vector embedding and finds content with the closest meaning — so an agent surfaces the right page, ticket, or code even when it shares no keywords with how the question was asked.

Agents can be assigned focused connectors for Confluence, GitHub, and Jira; a Federated Vector Search tool that queries all sources at once; Hybrid Search that combines vector and keyword recall; a Knowledge Graph Query; and the supporting pipeline of embedding generation, reranking, and chunk citation. Each one returns ranked, scored results an agent can ground its answers in.

Yes. These tools return the source page, file, or record for every hit, and the chunk citation tool attaches passage-level references — so an agent can produce a grounded answer that a human can open and verify rather than a black-box guess.

Hybrid search runs a vector (semantic) query and a keyword query together and merges the results, giving you meaning-aware recall without losing exact matches for names, codes, or identifiers. It is the safest default when queries mix natural language with specific terms.

No. The vector index and the search itself run on-premise or in your sovereign cloud, scoped per tenant and audit-logged. Nothing is sent to a third-party retrieval or embedding service.

From the agent’s toolkit on VDF AI, attach the retrieval tool you need under role-based policy. The agent can then call it whenever a workflow requires grounded recall, and several such agents can be orchestrated together as a VDF AI Network.