Enterprise Generative Search — Azure and Open Source  ·  View 22 of 41  ·  Runtime

Ask to Answer — The Critical Path

One request, twenty-one messages, and the re-retrieval that is the normal case rather than the exception.

Editable source SVG draw.io All views
Knowledge Worker Search Gateway Query Understanding Planner Retriever Facade Azure AI Search Model Gateway Verifier 1. POST /v1/answer 2. validate OBO token, start trace 3. interpret and authorise 4. class, entities, group set, fingerprint 5. QueryPlan 6. retrieve, depth 80 7. hybrid query with ACL filter 8. 94 candidates 9. fuse, rerank, diversify 10. 8 evidence items 11. synthesise over fenced evidence 12. first token, 1.2 s 13. draft with claim markers 14. verify claims against evidence 15. one claim unsupported 16. targeted re-retrieval, hop 2 17. 2 further items 18. re-verify 19. all claims supported 20. answer, citations, provenance 21. stream completed, 4.1 s Ask to Answer — The Critical Path, One Request The unsupported claim on message 15 is the normal case, not the exception. Budget allows one extra hop; a second failure abstains rather than guessing. v 1.0 · owner Data and AI Global Practice

What this view proves

  • The gateway never talks to the index. Retrieval is reached through the facade so that filters, evidence ids and traces cannot be bypassed by a service taking a shortcut.
  • The first token is streamed before verification completes. Verification can revise or withdraw a claim, and the UI is built to accept that — a stream that cannot be corrected forces the choice between latency and grounding.
  • Message 15 — one claim unsupported — is the expected path, not an error. The budget allows one extra hop; a second failure abstains.

Numbers

  • Total 4.1 s for this explanatory query: 0.3 s understanding and authorisation, 0.45 s retrieval, 0.11 s reranking, 2.4 s synthesis, 0.18 s verification, 0.7 s the second hop.
  • 94 candidates from four backends reduced to 8 passages and about 6,000 evidence tokens.

Deliberately omitted

  • The cache probe, the safety screen and the trace writes are not drawn; they would triple the message count and are covered in views 25, 38 and 34.
  • Tool invocation is a separate sequence in view 27 because it has a different authorisation path.
◀  Freshness Lifecycle All views The Bounded Agentic Loop  ▶
Solution Architecture v1.0 · Data & AI Global Practice · 2026-09 · 41 views · Microsoft Azure and open source22-ask-to-answer-sequence