Foundry IQ Search¶
Foundry IQ and Azure AI Search are the enterprise context plane. They should index governed, extracted, request-scoped evidence rather than raw opaque links.
Search is not the authority layer. It helps agents find governed evidence, but the backend resolver and route contracts decide whether evidence can support a savings/audit, dashboard, or deliverable claim.
First Index¶
The first ITR evidence index target is:
leafenterprise-it-request-evidence
Search documents are emitted by LeafEnterprise/salesforce_it_requests/evidence.py and can be uploaded by LeafEnterprise/salesforce_it_requests/materialize.py when Azure Search env vars are configured.
Document Properties¶
Search-ready documents include:
- stable hashed chunk id,
- request number,
- Salesforce record id,
- source system,
- source type and id,
- source version id,
- chunk ordinal,
- title,
- extracted text content,
- content hash,
- storage URI when available,
- sync run id,
- metadata.
Publication Rule¶
Keep module and sensitivity boundaries explicit. Cross-index routing should stay in LeafEnterprise-owned tools instead of assuming one broad search index is safe or sufficient.
Indexing Priorities¶
| Corpus | Preferred index posture |
|---|---|
| Salesforce ITR evidence | Request-scoped chunks with request number, source record id, source version id, content hash, and sync run id. |
| Savings/audit packets | Scope-aware packet summaries with formula/evidence links, not raw workbook dumps. |
| Contract packets | Exact packet summaries and extracted terms; raw Graph URLs remain backend-only. |
| Domo takeover evidence | Golden comparison and flaw-gate metadata, clearly marked as migration/reconciliation evidence. |
| Client dashboard payload docs | Backend API contracts, scope requirements, and blocked-state semantics. |