CambioML/pykoi
Open-source Python library for improving language models through RLHF with feedback collection, fine-tuning, and LLM comparison capabilities.

Velocity · 7d
+0.4
★ / day
Trend
→steady
star history
Pykoi provides a unified interface for RLHF and RLAIF workflows, enabling users to collect real-time user feedback on LLM outputs and continuously improve models through reinforcement learning. It supports integration with OpenAI, Amazon Bedrock, and Huggingface models, and includes tools for reward modeling, model comparison, and visualization of chat histories on a dashboard.