The Federated Vector Search Tool
Run one semantic query and get aggregated, ranked matches from Jira, GitHub, and Confluence together — the retrieval backbone for any agent that needs to answer across systems, on infrastructure you control.
The answer is in one of three systems — but which?
Enterprise knowledge is scattered: the requirement is in Jira, the implementation is in GitHub, and the decision is in Confluence. Searching each one separately is slow, and keyword search misses anything phrased differently.
Knowledge is siloed
The full picture lives across three tools at once, and no single search box spans them.
Keywords miss meaning
Exact-match search fails when the answer uses different words than the question.
Results aren’t comparable
Three separate searches return three lists with no shared ranking to tell you what’s most relevant.
Hosted search leaks IP
Pushing tickets, code, and docs to an external search service is exactly what regulated teams can’t do.
One semantic query, fanned out and merged
Fan-out
Query every source in parallel
Jira, GitHub, and Confluence, together.
The tool embeds the query once and runs it against each connected vector store, then merges the hits into a single ranked list so the most relevant result wins regardless of where it lives.
- Parallel search across 3 sources
- Embedding-based, meaning-aware matching
- Aggregated, de-duplicated results
- Tunable result depth per source
In a single call
Relevance
Ranked by similarity, not keywords
Find it even when it’s phrased differently.
Because matching is semantic, the tool surfaces the right ticket, file, or page even when it shares no exact terms with the query — and attaches a similarity score so agents can threshold low-confidence hits.
Threshold low-confidence hits
Governance
Tenant-isolated and on-premise
Vectors stay inside your perimeter.
Every query is scoped by user_id for strict multi-tenant isolation, and the whole index runs on-premise or in your sovereign cloud with role-based access and full audit logging.
Per-tenant by user_id
Parameters
The all_vectors_search tool accepts these inputs when an agent calls it. Required inputs are flagged.
default: 10 Optional Maximum results per source (1–25).
Where federated search pays back
Cross-system Q&A
Answer "how did we implement X?" by pulling the ticket, the code, and the design doc in one query.
Onboarding assistant
Let new hires ask questions and get answers grounded in real tickets, repos, and wiki pages.
Incident response
During an incident, surface the related change, the runbook, and the prior ticket together, fast.
Agent grounding
Give any agent a single retrieval call that grounds its answers across all three systems.
Duplicate detection
Find the existing ticket or doc before a team creates a duplicate of it.
Knowledge audits
Discover where a topic is documented — or where coverage is missing — across every source.
Assigned to agents, orchestrated as networks
On VDF AI, an industry’s use cases map to agents, and you assign tools like this one to those agents. Compose multiple agents into a governed, on-premise network.
What changes after you assign it
Questions about the Federated Vector Search tool
What is the federated vector search tool?
It is a retrieval tool that runs one semantic query across your Jira, GitHub, and Confluence vector stores at the same time and returns a single ranked, aggregated result set. On VDF AI you assign it to an agent so the agent can ground its answers across all three systems in a single call.
How is this different from searching each tool separately?
Separate searches give you three disconnected lists with no shared ranking, and each one is keyword-based. This tool embeds the query once, fans it out to every source, and merges the hits into one list ranked by similarity — so the most relevant result wins no matter where it lives.
How does multi-tenant isolation work?
Every query takes a user_id and is strictly scoped to that tenant’s vectors, so results never cross tenant boundaries. Combined with on-premise deployment and role-based access, it is safe for shared, regulated environments.
Can an agent use this together with other tools?
Yes. It is commonly assigned alongside the file summarizer and document generator so an agent can retrieve, condense, and produce a deliverable — and several agents using it can be composed into a VDF AI Network.
Does our data leave our infrastructure?
No. The vector index and the search run on-premise or in your sovereign cloud. Queries and results are logged for audit, and nothing is sent to a third-party service.
Assign Federated Vector Search to these agents
These VDF AI agents can be assigned this tool. Open an agent to see the full toolkit it can run.
Tools that work well alongside this one
Where this tool delivers value
Give your agents one query across every system
See federated vector search assigned to an agent and orchestrated in a network — governed and on-premise.