How intel/enterprise-rag works
Intel AI for Enterprise RAG is a platform layer for governed document ingestion and RAG workloads, including ChatQnA, DocSum, AudioQnA, and translation. Users ingest files or links through the admin UI, then ask questions in a chat UI; the composed pipeline embeds the query, retrieves and reranks documents, applies guardrails, builds a prompt, and calls the LLM. The README documents this flow and the platform integrations. Source excerpts confirm selected service and EDP behavior; links for unsampled subsystems follow README statements rather than inspected implementation.
Main entry points: main.go Pipeline router · tasks.py Enhanced data prep · ChatConversationRoute.tsx ChatQnA UI · DataIngestionTab.tsx Ingestion UI
Components 30
User applications
- ChatQnA UI
src/ui/apps/chatqna/src/app/routes/chat/ChatConversationRoute.tsx - Ingestion UI
src/ui/packages/data-ingestion/src/components/DataIngestionTab/DataIngestionTab.tsx - DocSum UI
src/ui/apps/docsum/src/app/routes/docsum/DocSumRoute.tsx - AudioQnA UI
src/ui/apps/audioqna/src/app/routes/chat/ChatConversationRoute.tsx
Pipeline control
- Pipeline composer
deployment/scripts/compose_pipeline.py - GMConnector resource
src/gmc/api/v1alpha3/gmconnector_types.go - GMC operator
src/gmc/internal/controller/gmconnector_controller.go - Pipeline router
src/gmc/cmd/router/main.go
Document ingestion
- Enhanced data prep
src/edp/app/tasks.py - Text extraction
src/comps/text_extractor/erag_text_extractor_microservice.py - Text splitting
src/comps/text_splitter/erag_text_splitter_microservice.py - Vector ingestion
src/comps/ingestion/erag_ingestion_microservice.py
RAG services
- Embedding service
src/comps/embeddings/erag_embedding_microservice.py - Retriever
src/comps/retrievers/erag_retriever_microservice.py - Reranking service
src/comps/reranks/erag_reranking_microservice.py - Guardrail services
src/comps/guardrails - Prompt template
src/comps/prompt_template/erag_prompt_template_microservice.py - LLM service
src/comps/llms/erag_llm_microservice.py - Chat history
src/comps/chat_history/erag_chat_history_microservice.py - Document summarization
src/comps/docsum/erag_docsum_microservice.py - Speech services
src/comps/asr/erag_asr_microservice.py - Translation workload
src/comps
Shared platform
- Microservice framework
src/comps/cores/mega/micro_service.py - Document protocols
src/comps/cores/proto/docarray.py
Other components
- Enterprise user
- Platform gateway
- Keycloak identity
- Object storage
- Vector database
- Platform inference
Connections 33
- Enterprise user → ChatQnA UI: asks questions
README.md - Enterprise user → Ingestion UI: uploads documents
README.md - Enterprise user → DocSum UI: submits documents
README.md - Enterprise user → AudioQnA UI: asks by voice
README.md - ChatQnA UI → Platform gateway: sends requests
README.md - Ingestion UI → Enhanced data prep: starts ingestion
README.md - DocSum UI → Platform gateway: submits summaries
- AudioQnA UI → Platform gateway: sends audio
- Platform gateway → Keycloak identity: authenticates users
README.md - Platform gateway → Pipeline router: routes requests
README.md - Pipeline composer → GMConnector resource: renders flow
README.md - GMConnector resource → GMC operator: reconciles workload
README.md - GMC operator → Pipeline router: manages services
README.md - Enhanced data prep → Object storage: reads documents
README.md - Enhanced data prep → Text extraction: extracts content
README.md - Enhanced data prep → Text splitting: splits documents
README.md - Enhanced data prep → Embedding service: requests embeddings
README.md - Enhanced data prep → Vector ingestion: submits chunks
src/edp/app/tasks.py - Vector ingestion → Vector database: writes vectors
README.md - Pipeline router → Embedding service: embeds query
README.md - Pipeline router → Retriever: retrieves candidates
README.md - Retriever → Vector database: searches vectors
README.md - Pipeline router → Reranking service: reranks candidates
README.md - Pipeline router → Guardrail services: filters input
README.md - Pipeline router → Prompt template: builds prompt
README.md - Pipeline router → LLM service: requests completion
README.md - LLM service → Platform inference: requests inference
README.md - Pipeline router → ChatQnA UI: streams answer
README.md - ChatQnA UI → Chat history: manages conversations
- Pipeline router → Document summarization: runs summary flow
README.md - Pipeline router → Speech services: runs voice flow
README.md - Pipeline router → Translation workload: runs translation flow
README.md - Microservice framework → Document protocols: serves typed payloads
src/comps/cores/mega/micro_service.py