Open Assistant API is an open-source, self-hosted AI intelligent assistant API that’s compatible with the official OpenAI interface. It can be seamlessly used with the official OpenAI Client to create LLM applications.
It also supports One API for integration with a broader range of commercial and private models.
Additionally, it includes integration with the R2R RAG engine.
Below is an example of using the official OpenAI Python openai library:
import openai
client = openai.OpenAI(
base_url="http://127.0.0.1:8086/api/v1",
api_key="xxx"
)
assistant = client.beta.assistants.create(
name="demo",
instructions="You are a helpful assistant.",
model="gpt-4-1106-preview"
)
Why Choose Open Assistant API
Feature Open Assistant API OpenAI Assistant API
Ecosystem Strategy Open Source Closed Source
RAG Engine Supports R2R Supported
Internet Search Supported Not Supported
Custom Functions Supported Supported
Built-in Tools Extendable Not Extendable
Code Interpreter In Development Supported
Multimodal Supported Supported
LLM Support More LLMs Only GPT
Message Streaming Output Supported Supported
Local Deployment Supported Not Supported
LLM Support: Compatible with a wider range of models via One API integration.
Toolset: Currently supports online search and can be expanded with additional tools.
RAG Engine: Supports file types like txt, html, markdown, pdf, docx, pptx, xlsx, png, mp3, mp4, and more.
Message Streaming: Provides smoother message streaming for enhanced user interaction.
Open Source: Completely open-source for local deployment and customization.
Quick Start
The easiest way to launch the Open Assistant API is by using the Docker Compose file. Ensure Docker and Docker Compose are installed on your system.
Configuration
Navigate to the project root, open docker-compose.yml, and fill in your OpenAI API key and Bing search key (optional).
sh
Copy code
# openai api_key (supports OneAPI api_key)
OPENAI_API_KEY=<openai_api_key>
# bing search key (optional)
BING_SUBSCRIPTION_KEY=<bing_subscription_key>
Configuring the R2R RAG engine is recommended for a better RAG experience. You can set this up through the R2R repository.
sh
# RAG config
# FILE_SERVICE_MODULE=app.services.file.impl.oss_file.OSSFileService
FILE_SERVICE_MODULE=app.services.file.impl.r2r_file.R2RFileService
R2R_BASE_URL=http://<r2r_api_address>
R2R_USERNAME=<r2r_username>
R2R_PASSWORD=<r2r_password>
Run
Run with Docker Compose:
sh
Copy code
docker compose up -d
Access API
API Base URL: http://127.0.0.1:8086/api/v1
Documentation: http://127.0.0.1:8086/docs
Complete Usage Example
Below is a basic example to create and run an AI assistant using the OpenAI client. For other advanced features like streaming output, online tools, or custom functions, check the examples directory. Before starting, install the OpenAI Python library:
sh
Copy code
# !pip install openai
export PYTHONPATH=$(pwd)
python examples/run_assistant.py
Permissions
User isolation is implemented based on tokens for SaaS deployment support. To enable, configure APP_AUTH_ENABLE.
Use Bearer token for authentication. Include Authorization: Bearer *** in the header.
Token management details can be found in the API documentation under the token section. Admin token, set as APP_AUTH_ADMIN_TOKEN, defaults to "admin".
When creating a token, specify the base URL and API key. The assistant will use these settings to access the model.
Tools
Following OpenAPI/Swagger standards, the assistant can integrate various tools to enhance external connectivity.
Enables integration with other systems or services for tasks like code execution or data access.
Tools need to be created before they can be used with the assistant. See test cases for more details: Assistant With Action.
For tools requiring authentication, add auth details at runtime. Detailed parameter formats are in the API documentation. More details are in the test cases: Run With Auth Action.
Community and Support
Join our Slack channel for discussions, updates, and support.
Connect with community members on Discord.
Join our WeChat group:
Special Thanks
This project builds on the following incredible open-source projects:
OpenOpenAI: Node-based Assistant API
One API: Multi-model management
R2R: RAG engine
OpenAI-Python: OpenAI Python client
OpenAI API: OpenAI API interface definition
LangChain: LLM application development library
OpenGPTs: LangChain GPTs
TaskingAI: TaskingAI client SDK
Contributing
Please refer to our Contribution Guide to learn how to get involved and contribute.