How ecvanlare/cloud-standards-assistant works
This repository is an Azure-hosted, agentic assistant for cloud and DevOps standards. Its main interaction runs from the browser chat through a FastAPI backend to a Foundry agent, which retrieves indexed guidance and can call live ASB and Terraform-version tools before returning an answer with citations and trace metadata. Separate product branches cover corpus ingestion, safety, observability, evaluation, and multi-environment infrastructure. The graph follows documented capabilities and sampled call sites; unsampled subsystem wiring is shown only where README documentation supports it.
Main entry points: standards-assistant.json Foundry agent · main.py Chat BFF · index.json AI Search index · envs Azure resources
Components 18
Chat and agent
- Chat UI
serving/app/static/app.js - Chat BFF
serving/app/main.py - Foundry agent
agents/standards-assistant.json - Answer and citations
serving/app/main.py
Knowledge retrieval
- Standards corpus
corpus/README.md - Search ingestion
search - Azure Blob Storage
- AI Search index
search/index.json
Live tools and safety
- ASB version function
tools/functions/asb_version/function_app.py - Registry proxy function
tools/functions/terraform_registry/function_app.py - Terraform Registry
- RAI content filters
safety
Telemetry and evaluation
- Application Insights
- Golden evaluation set
eval/golden_set.jsonl - Regression evaluation
eval/scripts/run_eval.py - Foundry evaluation
eval/scripts/run-foundry-eval.py
Azure platform
- Azure resources
terraform/envs
Other components
- Assistant user
Connections 24
- Assistant user → Chat UI: asks
- Chat UI → Chat BFF: sends question
- Chat BFF → Foundry agent: submits turn
- Foundry agent → Chat BFF: returns response
- Chat BFF → Answer and citations: formats response
- Answer and citations → Chat UI: renders
- Standards corpus → Azure Blob Storage: uploaded to
- Search ingestion → Azure Blob Storage: reads corpus
- Search ingestion → AI Search index: chunks and indexes
- Azure Blob Storage → AI Search index: indexer ingests
- Foundry agent → AI Search index: retrieves guidance
- Foundry agent → ASB version function: calls version tool
- Foundry agent → Registry proxy function: calls registry tool
- Registry proxy function → Terraform Registry: fetches versions
- Foundry agent → RAI content filters: applies guardrails
- Chat BFF → Application Insights: sends OTel
- Foundry agent → Application Insights: emits traces
- Golden evaluation set → Regression evaluation: supplies cases
- Regression evaluation → Foundry agent: runs questions
- Golden evaluation set → Foundry evaluation: supplies dataset
- Foundry evaluation → Foundry agent: evaluates agent
- Azure resources → Foundry agent: provisions platform
- Azure resources → AI Search index: provisions search
- Azure resources → Application Insights: provisions monitoring