myAi/RAG_SPLIT_README.md

MyAi RAG split cleanup

Public api

The existing api project is now only the public gateway for the existing frontend.

It keeps:

• contact API
• file download API
• Google/config APIs
• health API
• api/rag/* proxy endpoints

It no longer contains local RAG processing code. The removed responsibilities are:

• PDF extraction
• chunking
• embeddings
• vector storage
• OpenAI/Ollama calls
• job text extraction
• CV matching business logic

api/Controllers/RagController.cs is intentionally kept. It proxies the current frontend calls:

• POST /api/rag/cv -> cv-matcher-api /api/cv/upload
• POST /api/rag/match-job -> cv-matcher-api /api/cv/match-job

Required public API config:

"CvMatcherApi": {
  "BaseUrl": "http://cv-matcher-api:8080",
  "InternalApiKey": "change-this-internal-key"
}

cv-matcher-api

Business API for CV/job workflows.

Main endpoints:

• POST /api/cv/upload
• POST /api/cv/match-job
• POST /api/cv/find-jobs
• GET /health
• Swagger: /swagger

Responsibilities:

• CV matcher business logic
• job URL/text extraction
• final LLM scoring
• result persistence
• email sending
• calls rag-api for generic semantic indexing/search

rag-api

Generic semantic search API.

Main endpoints:

• POST /api/rag/documents
• POST /api/rag/documents/json
• POST /api/rag/search
• GET /api/rag/documents/{id}
• GET /health
• Swagger: /swagger

Responsibilities:

• generic document indexing
• automatic document type classification when type is missing
• PDF/text extraction
• chunking
• embedding creation
• embedding and chat completion cache
• semantic search over generic documents

Logging and Swagger

All three APIs now have:

• Serilog startup logging
• Serilog request logging
• structured JSON console logs
• health endpoint
• Swagger/OpenAPI support

Swagger is enabled by default and can be disabled per service with:

"Swagger": {
  "Enabled": false
}

Internal API security

Both internal APIs support API-key protection:

"InternalApi": {
  "RequireApiKey": true,
  "ApiKey": "change-this-internal-key"
}

Requests must include:

X-Internal-Api-Key: change-this-internal-key