lm-sys/FastChat
FastChat is an open platform for training, serving, and evaluating large language model based chatbots, featuring multi-model serving, extensive model support, and large-scale evaluation datasets.
Awesome AI Agents › Chatbot Platforms
choosewhatulike/trainable-agents
Character-LLM is a project that introduces trainable agents specifically designed for role-playing as historical or fictional characters. Unlike traditional prompted agents, these Character-LLMs are trained to embody the knowledge, personality traits, and emotional characteristics of specific figures such as Beethoven, Cleopatra, Julius Caesar, and others. This allows the agents to respond and interact in a manner consistent with the character they represent without requiring additional prompts or reference documents. The core innovation behind this project is the Experience Reconstruction method, a data generation process that creates detailed and diverse experience data for each character to train the models effectively. The repository provides code, datasets, and pre-trained model weights for nine characters, enabling users to deploy these agents as chatbots that simulate conversations with these personalities. The training datasets are available on Hugging Face and include extensive experience data for each character, with thousands of scenes and hundreds of thousands of words. The models are based on the Llama 1 architecture, and due to licensing, the repository releases weight differences that users can apply to base models to obtain the full Character-LLM models. The project is linked to an EMNLP 2023 paper and offers detailed instructions for downloading datasets, applying model weight deltas, and running the models with meta prompts to simulate character-specific dialogues. This work advances the field of AI-driven role-playing by providing trainable, character-specific language models that can be used for educational, entertainment, or research purposes.
https://github.com/choosewhatulike/trainable-agents
Also in Chatbot Platforms
CopilotKit/CopilotKit
CopilotKit is a React-based framework for building deeply integrated AI assistants, chatbots, and in-app AI agents that enhance user interactions through natural language and contextual understanding.
microsoft/TypeChat
TypeChat is a library that simplifies building natural language interfaces by using types to define intents and schemas, replacing traditional prompt engineering with schema engineering for improved validation and reliability.
run-llama/rags
RAGs is a Streamlit app that enables users to build and interact with personalized Retrieval-Augmented Generation pipelines over their own data using natural language instructions.
SylphAI-Inc/AdalFlow
AdalFlow is a PyTorch-like library that enables building and auto-optimizing large language model applications, including chatbots, RAG, and agents, with a unified framework for prompt optimization and model-agnostic flexibility.
nexu-io/nexu
Open-source desktop client that runs an OpenClaw agent locally and bridges it into WeChat, Feishu, Slack and Discord through a graphical setup, with bring-your-own-key and OAuth model access.
LLPhant/LLPhant
LLPhant is a comprehensive PHP generative AI framework inspired by Langchain, supporting multiple LLMs and designed to simplify building powerful AI applications.
langchain-ai/streamlit-agent
Reference implementations of various LangChain agents as interactive Streamlit applications demonstrating conversational AI capabilities and integrations.