tangbotony/GraTAG
GraTAG — Production AI Search via Graph-Based Query Decomposition and Triplet-Aligned Generation with Rich Multimodal Representations
Implements graph-based query decomposition (DAG-structured sub-queries with GRPO alignment) and triplet-aligned generation (relation extraction + REINFORCE alignment) to improve coherence and reduce hallucination in retrieval-augmented search. Integrates multimodal visualization (timeline + Hungarian algorithm image-text matching), MongoDB/Elasticsearch/Milvus for persistence and retrieval, and supports both synchronous and streaming LLM inference via vLLM/HF TGI-compatible endpoints.
204 stars.
Stars
204
Forks
29
Language
Python
License
—
Category
Last pushed
Feb 26, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/rag/tangbotony/GraTAG"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Compare
Higher-rated alternatives
beir-cellar/beir
A Heterogeneous Benchmark for Information Retrieval. Easy to use, evaluate your models across...
superlinear-ai/raglite
🥤 RAGLite is a Python toolkit for Retrieval-Augmented Generation (RAG) with DuckDB or PostgreSQL
HKUDS/LightRAG
[EMNLP2025] "LightRAG: Simple and Fast Retrieval-Augmented Generation"
illuin-tech/vidore-benchmark
Vision Document Retrieval (ViDoRe): Benchmark. Evaluation code for the ColPali paper.
HKUDS/RAG-Anything
"RAG-Anything: All-in-One RAG Framework"