Skip to main content

Knowledge & Document Pipeline Data Flow

Status

Owner: Tech lead · Status: draft · Canonical source: training/diagrams/FleziGalaxy-Architectures/04-workflow-document-knowledge-pipeline.html

Details document ingestion, encrypted object storage, parsing, chunking, embedding generation, and hybrid retrieval.

Ingestion Architecture​

Existing Controls​

  • File extension allowlisting and MIME type verification.
  • Server-Side Encryption (SSE-256) on Azure Blob Storage.
  • Workspace-scoped access boundaries on document search queries.

Evidence (Files & Configuration)​

  • flezi.app.api.backend/app/features/files/storage.py
  • flezi.app.ai.agent/app/document_pipeline/

Limitations & Roadmap​

  • Customer-Managed Keys (CMK) support on Azure Key Vault for document blob storage.