Proposed service dependencies and data flow
Parse uploaded documents (Excel/PDF), extract structured KVPs, validate data quality, and store intermediate results.
Orchestrate LLM prompt chain, retrieve historical context via RAG, generate variance analysis and compliant narratives.
Handle Workiva OAuth2 authentication, map data to container IDs, execute PUT requests to update ESG sections.
Private deployment with reasoning capabilities. Temperature: 0.1-0.3 for consistency. Max tokens: 4096.
Historical Factbooks, style guides, previous ESG reports, regulatory templates, and approved narratives.
B3 Energy Section - Narrative and Data cells for automated population.
| From \ To | Event Grid | func_ingest | func_reason | func_bridge | OpenAI | AI Search | Key Vault | Workiva |
|---|---|---|---|---|---|---|---|---|
| SharePoint | Event | - | - | - | - | - | - | - |
| Event Grid | - | Trigger | Trigger | - | - | - | - | - |
| func_ingest | - | - | KVPs | - | - | - | Secrets | - |
| func_reason | - | - | - | Draft | Prompt | Query | Secrets | - |
| func_bridge | - | - | - | - | - | - | Creds | PUT |
Project Northern Lights Component Interactions v1.0 | Danta Labs for KPMG Lighthouse
Service Mesh Architecture | Azure Functions