ChatPilot

??Chinese | English

---

ChatPilot: Chat Agent Web UI

ChatPilot : Chat Agent WebUI, implements AgentChat dialogue, supports Google search, file URL dialogue (RAG), code interpreter function, reproduces Kimi Chat (file, drag in; URL, send out), supports OpenAI/Azure API.

Features

• This project implements Agent Assistant calling based on Agentica and supports the following functions:
  • Tool calling: Support Agent calling external tools
    • Internet search tool: Google Search API (Serper/DuckDuckGo)
    • URL automatic parsing tool: reproduces the function of sending Kimi Chat URL
    • Python code interpreter: supports E2B virtual environment and local python compiler environment to run code
  • Multiple LLM access: supports multiple LLM models to be accessed in multiple ways, including using Ollama API to access various local open source models; using litellm API to access various cloud service deployment models; using OpenAI API to access GPT series models
  • RAG: Support Agent calling RAG file Q&A
• Supports separation of front-end and back-end services. The front-end uses Svelte and the back-end uses FastAPI.
• Support voice input and output, support image generation
• Support user management, permission control, support import and export of chat records

Demo

Official Demo: https://chat.mulanai.com

Getting Started

Run ChatPilot in Docker

 export OPENAI_API_KEY=sk-xxx
export OPENAI_BASE_URL=https://xxx/v1

docker run -it 
 -e OPENAI_API_KEY= $WORKSPACE_BASE 
 -e OPENAI_BASE_URL= $OPENAI_BASE_URL 
 -e RAG_EMBEDDING_MODEL= " text-embedding-ada-002 " 
 -p 8080:8080 --name chatpilot- $( date +%Y%m%d%H%M%S ) shibing624/chatpilot:0.0.1

You'll find ChatPilot running at http://0.0.0.0:8080 Enjoy! ?

Start service locally

git clone https://github.com/shibing624/ChatPilot.git
cd ChatPilot
pip install -r requirements.txt

# Copying required .env file, and fill in the LLM api key
cp .env.example .env

bash start.sh

Okay, now your application is running: http://0.0.0.0:8080 Enjoy! ?

Build front-end web

Two ways to build the front end:

1. Download the packaged and compiled front-end buid.zip and extract it to the project web directory.
2. If you modify the web front-end code, you need to use npm to rebuild the front-end yourself:

git clone https://github.com/shibing624/ChatPilot.git
cd ChatPilot/

# Building Frontend Using Node.js >= 20.10
cd web
npm install
npm run build

Output: The project web directory outputs the build folder, which contains the front-end compilation output file.

Multiple LLM access

Use OpenAI API to access GPT series models

• Use the OpenAI API to configure environment variables:

 export OPENAI_API_KEY=xxx
export OPENAI_BASE_URL=https://api.openai.com/v1
export MODEL_TYPE= " openai "

• If you use Azure OpenAI API, you need to configure the following environment variables:

 export AZURE_OPENAI_API_KEY=
export AZURE_OPENAI_API_VERSION=
export AZURE_OPENAI_ENDPOINT=
export MODEL_TYPE= " azure " 

Use Ollama API to connect to various local open source models

Start the ollama service with ollama serve , and then configure OLLAMA_API_URL : export OLLAMA_API_URL=http://localhost:11413

Use litellm API to access each cloud service deployment model

1. Install litellm package:

pip install litellm -U

2. Modify configuration file

The default litellm config file of chatpilot is in ~/.cache/chatpilot/data/litellm/config.yaml

Modify its content as follows:

 model_list :
#  - model_name: moonshot-v1-auto # show model name in the UI
#    litellm_params: # all params accepted by litellm.completion() - https://docs.litellm.ai/docs/completion/input
#      model: openai/moonshot-v1-auto # MODEL NAME sent to `litellm.completion()` #
#      api_base: https://api.moonshot.cn/v1
#      api_key: sk-xx
#      rpm: 500      # [OPTIONAL] Rate limit for this deployment: in requests per minute (rpm)

  - model_name : deepseek-ai/DeepSeek-Coder # show model name in the UI
    litellm_params : # all params accepted by litellm.completion() - https://docs.litellm.ai/docs/completion/input
      model : openai/deepseek-coder # MODEL NAME sent to `litellm.completion()` #
      api_base : https://api.deepseek.com/v1
      api_key : sk-xx
      rpm : 500
  - model_name : openai/o1-mini # show model name in the UI
    litellm_params : # all params accepted by litellm.completion() - https://docs.litellm.ai/docs/completion/input
      model : o1-mini # MODEL NAME sent to `litellm.completion()` #
      api_base : https://api.61798.cn/v1
      api_key : sk-xxx
      rpm : 500
litellm_settings : # module level litellm settings - https://github.com/BerriAI/litellm/blob/main/litellm/__init__.py
  drop_params : True
  set_verbose : False 

Contact

• Issue(suggestion):
• Email me: xuming: [email protected]
• WeChat Me: Add my WeChat ID: xuming624, Remarks: Name-Company-NLP to join the NLP communication group.

Citation

If you use ChatPilot in your research, please cite it in the following format:

APA:

Xu, M. ChatPilot: LLM agent toolkit (Version 0.0.2) [Computer software]. https://github.com/shibing624/ChatPilot

BibTeX:

@misc{ChatPilot,
  author = {Ming Xu},
  title = {ChatPilot: llm agent},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = { url {https://github.com/shibing624/ChatPilot}},
}

License

The licensing agreement is The Apache License 2.0, which is free for commercial use. Please attach the link to ChatPilot and the license agreement in the product description.

Contribute

The project code is still very rough. If you have any improvements to the code, you are welcome to submit it back to this project. Before submitting, please pay attention to the following two points:

• Add corresponding unit tests in tests
• Use python -m pytest -v to run all unit tests to ensure that all unit tests pass

You can then submit a PR.

Reference

• Open WebUI
• langchain-ai/langchain
• shibing624/agentica