FareedKhan-dev/best-llm-finder-pipeline

Implements role-specific LLM evaluation by running models through three production-grade pipelines—agentic RAG with recursive document navigation, multi-agent systems with parallel ideation and critique loops, and vision-based form processing with OCR refinement—measuring performance via task-specific metrics (faithfulness, latency, cost) rather than isolated benchmarks. Uses OpenAI-compatible APIs to support multiple model providers (Nebius, Ollama, Together AI), enabling cost-optimized routing by deploying smaller models for fast classification and larger models for synthesis and reasoning tasks.

132 stars. No commits in the last 6 months.