Feature/graphdb

[qchapp]

Now loading the catalog from the graphdb endpoint instead of a static jsonl catalog.

[Copilot]

Pull Request Overview

This PR refactors the retriever module architecture by splitting the monolithic embedders.py into focused, single-responsibility modules and adds GraphDB catalog synchronization functionality.

• Splits retriever/embedders.py into separate modules: software_doc.py, text_embedder.py, reranker.py, and vector_index.py for better maintainability
• Introduces GraphDB integration with automatic catalog syncing via new catalog/sync.py module
• Updates RAGImagingPipeline initialization to load from persisted index instead of requiring docs parameter

Reviewed Changes

Copilot reviewed 21 out of 22 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
src/ai_agent/retriever/software_doc.py	Extracted SoftwareDoc schema with field validation and format token extraction
src/ai_agent/retriever/text_embedder.py	Extracted text embedding interface and BGE-M3 implementation
src/ai_agent/retriever/reranker.py	Extracted CrossEncoder reranking logic
src/ai_agent/retriever/vector_index.py	Extracted FAISS index management with fingerprinting and persistence
src/ai_agent/catalog/sync.py	New module for GraphDB SPARQL queries and catalog synchronization
src/ai_agent/cli.py	Extended CLI with sync command and background catalog refresh
src/ai_agent/api/pipeline.py	Updated to load index from disk; removed docs parameter from constructor
src/ai_agent/ui/app.py	Updated to load docs from FAISS metadata after index initialization
tests/full_test.py	Updated imports and monkeypatch targets for new module structure
src/ai_agent/utils/full_processing.py	Refactored as reusable function; added logging

Comments suppressed due to low confidence (1)

src/ai_agent/utils/full_processing.py:103

• Test is always true, because of this condition.