Backend Capability Map

This page summarizes the backend systems LeafEnterprise uses to replace hidden BI ownership with deterministic, inspectable operations.

Provisioned And Target Runtime Capabilities

Capability	Current / target component	Backend responsibility
Claims runtime	AWS SQL Server	Source scoped claims/audit rows for current live reports.
Migration evidence	Azure SQL dedicated_domo_takeover and ADLS Gen2	Preserve Domo takeover data, bronze/silver normalized evidence, and curated comparison packets.
Artifact storage	ADLS Gen2 and local artifact ledgers	Store materialized files, extracted text, output artifacts, replay inputs, and worker outputs.
Contract packet retrieval	Microsoft Graph / OneDrive	Retrieve and promote governed contract/pricing/rebate packet evidence.
Request intake	Salesforce ITR mirror	Preserve request rows, history, files, notes, feed items, comments, and run-plan lineage.
Enterprise retrieval	Azure AI Search / Foundry IQ	Index governed chunks and packets for agent/tool context.
Long-running execution	Azure Service Bus and AKS lane workers	Dispatch, run, trace, and artifact long-running agent/backend work.
Model-backed assistance	Azure OpenAI, OpenAI Agents SDK, Codex SDK	Draft, summarize, plan, and execute through backend guardrails and deterministic fallback.
API governance	APIM target plus Cloudflare Access perimeter	Protect internal tools, future MCP traffic, and hosted docs/app access.

Integration Ownership

Integration	LeafEnterprise-owned module or contract	Trust boundary
AWS SQL	Report context and workbook generation paths.	Backend credentials only; browser sees scoped payloads.
Azure SQL / ADLS	Domo takeover jobs and artifact/materialization pipelines.	Mutation requires explicit operator flags or worker policy.
OneDrive Graph	contract_guarantees and graph_contract_packets resolver lane.	Read-only; no token, raw URL, or local path exposure.
Salesforce	salesforce_it_requests mirror, materialization, evidence bundle, search-document contract.	Backend auth only; browser receives normalized evidence.
Agent workers	LeafEnterprise/agents, Service Bus dispatch, AKS worker entrypoint.	Policy-gated; every task has status, trace, and artifact posture.
OpenAI / Codex SDKs	Backend provider layer and task runner.	AI is assistant/execution support, not calculation truth.

Deterministic Ownership Pattern

Every backend process should follow the same pattern:

flowchart LR
  Source[Source system] --> Mirror[Mirror or materialize]
  Mirror --> Normalize[Normalize / packetize]
  Normalize --> Validate[Validate scope and evidence]
  Validate --> Compute[Deterministic compute]
  Compute --> Publish[Publish scoped output]
  Publish --> Present[Domo / frontend / workbook / PDF]
  Validate --> Agent[Agent worker if long-running]
  Agent --> Publish

The presentation layer may change. The backend ownership pattern should not.