Spaces:
Sleeping
Sleeping
Commit History
-
docs: thêm màu vào Mermaid pipeline diagrams
cd36943 -
docs: chuyển pipeline diagrams sang Mermaid
2415a88 -
fix: eval dùng 8b-instant + checkpoint resume + cắt context để tiết kiệm token
4455cfa -
fix: use llama-3.1-8b-instant for eval to avoid shared TPD exhaustion
c76b5a4
-
fix: query rewriter falls back to original query on rate limit error
f79d236 -
data: expand golden set từ 12 lên 30 câu hỏi
ee705f2 -
ui: show ColPali unavailable message on HF Spaces
c746d00 -
feat: show rewritten query in chat UI
When query rewriter changes the original question, display a badge "Query được viết lại: ..." below the user bubble to make the rewriting step visible in the pipeline. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>e535e98 -
feat: expand source to show retrieved text chunk
Click "Xem đoạn văn" on each source to preview the exact text the LLM used. Adds text field to Source model and passes it through the SSE stream. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>7053098 -
fix: graceful fallback when colpali_engine not installed
colpali_retriever imports are now conditional so /compare endpoint works on HF Spaces without colpali_engine; also fix local Qdrant path. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>9e7ded3 -
fix: rebuild BM25 index from Qdrant if pkl not found
538d976 -
fix: download bm25_index.pkl from HF repo at build time
f48746c -
fix: remove bm25_index.pkl COPY from Dockerfile (uploaded via HF API)
8fbcfcb -
feat: add compare pipeline tab + has_context badge in UI
862a5f6 -
docs: add key results to intro, add GitHub link
4e3e867 -
chore: track pkl files with git lfs
93855f1 -
ui: update sample questions to financial report queries
270dbbb -
feat: contextual retrieval eval + update README results table
b3dbec1 -
feat: contextual retrieval with llama-3.1-8b-instant, Recall 0.40→0.72
79b4e9d -
docs: chuyển README sang tiếng Việt
1544382 -
feat: add ColPali eval script + update comparison results in README
- evaluate_colpali.py: head-to-head ColPali vs Text RAG source relevance eval - eval_colpali_results.json: Text RAG 0.790 vs ColPali 0.520 (10 questions) - README: update comparison table with actual numbers and per-query-type breakdown Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>0165cfe
-
docs: update README with ColPali comparison, eval results, architecture
- Add Pipeline A (Text RAG) vs Pipeline B (ColPali) architecture diagrams - Add retrieval comparison table with findings - Add RAGAS baseline results (Faith 0.633 / Rel 0.767 / Recall 0.400) - Update stack table with ColPali and Qdrant Cloud/local split - Add architecture decisions for ColPali local storage and contextual retrieval - Update local setup commands for new scripts Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>e0403fd -
feat: add ColPali visual retrieval + contextual retrieval + PDF eval
- src/contextualizer.py: Contextual Retrieval with Gemini Flash, retry on rate limit - ingest.py: --contextual flag for contextual ingest - src/colpali_retriever.py: ColPali v1.2 visual retrieval with Qdrant MaxSim - ingest_colpali.py: PDF pages -> images -> ColPali embeddings pipeline - compare_retrieval.py: side-by-side comparison ColPali vs text RAG - evaluate.py: accept golden set path as CLI arg - data/golden_set_pdf.json: PDF financial report evaluation set - eval_results.json: baseline results (Faithfulness 0.633 / Relevancy 0.767 / Recall 0.400) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>af92b4d
-
feat: dark mode, mobile responsive, copy button, skeleton loader, sample questions
25622cc -
docs: add live demo link and architecture decisions
099fb8b -
fix: add nginx buffer flush padding for SSE on HF Spaces
cba7c68 -
fix: async SSE generator with blinking cursor for streaming
73964c0 -
feat: streaming response via SSE with token-by-token display
b18c987 -
fix: add python-multipart for file upload
6a99458 -
feat: PDF upload with drag & drop, ephemeral in-memory search
f7f0189 -
perf: preload reranker model at startup
b1ddd58 -
fix: remove *.pkl from dockerignore
3d49947 -
fix: explicit bm25 copy path to bust cache
edbe951 -
fix: allow bm25_index.pkl in Docker build
3160026 -
fix: copy bm25_index.pkl into Docker image
694cc30 -
fix: commit bm25_index.pkl for HF Spaces deployment
a088361 -
debug: print QDRANT_URL
995eaeb -
remove debug log
d5f100c -
debug: print GROQ key prefix in logs
8ccdbf2 -
fix: strip whitespace from API keys
cc2a682 -
fix: remove local hf_models copy, let model download on startup
67adee2 -
fix: add HuggingFace Spaces configuration header
d5aebee -
docs: replace roadmap with Future Work section
ce5886b -
docs: final README with full feature list and quick start
7f57072 -
Docker containerization with baked HuggingFace model cache
6f9d8f6 -
improve chat UI, fix score bars, lower rerank threshold
f8ff11b -
Vietnamese tokenization, query rewriting, RAGAS evaluation
aa2c264 -
Vietnamese tokenization, query rewriting, RAGAS evaluation
06f35c0 -
Vietnamese tokenization (underthesea) + query rewriting
d287ce2 -
Week 4: reranking + citation - bge-reranker-v2-m3, inline citations
c696a51