Clean sys.path and add PYTHONPATH

Fix hardcoded absolute path for Linux deployment

Add YAML metadata for Docker space

Remove ngrok.exe from tracking

add .env.example

Merge remote changes and resolve rename/delete conflict for test files

phase 4: Complete agent UI and final evaluation run

Delete src/agents/test_retrievers.py

Delete src/agents/test_decomposer.py

Delete Makefile

Phase 3: full 30-question evaluation complete - ViLexAgent groundedness 4.35 vs 2.20 baseline

Phase 2: full 5-node pipeline working end-to-end - cross-reference and synthesis verified

fix(retrievers): apply Singleton pattern for Jina v5 to resolve config_class error and optimize VRAM

bug detected at embedding model loading, fix in next commit

Phase 2: state schema and query decomposer node verified

Phase 1: baseline retriever verified - all 4 query types returning correct results

Phase 1: international corpus acquired, parsed, and indexed (112 chunks)

Phase 1: embedding and indexing complete - 45544 chunks in Qdrant

Phase 1: HTML parsing and article-level chunking (45544 chunks)

Phase 1: domain filtering and metadata-content join (8596 docs)

Phase 1: data acquisition - downloaded all 3 dataset configs

Phase 0: project scaffold, Docker stack, environment setup