Persist provider reasoning and assistant text separately on agent_reasoning_audit
(DeepSeekAssistantMessage path), extract diagnosis_run.conclusion, enrich RAG
tool audit (step_id/query/qualityScore), gate empty mysql tools, drop devtools,
and align MVP docs after live E2E verification.
Document the lookup_knowledge flow from L0 hints through L1 retrieval,
post-processing, packing, and Agent projection so the boundaries and
current limitations are easier to follow.