Practical overview
When it fits
Use it when you want to operate the assistant orchestration service yourself and need an OpenAI-compatible integration path with model, retrieval or tool extensions.
Who it is for
- Developers building LLM applications that need an assistant API they can self-host.
- Teams integrating assistant features with OpenAI-compatible clients.
- Operators combining model providers with retrieval or custom tools.
Documented uses
- Create assistants through an OpenAI-compatible client interface.
- Connect assistants to commercial or private model services through One API.
- Add retrieval using the R2R RAG engine.
- Extend assistants with function calls and tools, including documented online search support.
Requirements & limits
The project is described as open source under the MIT license. The README documents self-hosted deployment using Docker Compose and configuration for a model API key.
- Docker and Docker Compose are prerequisites for the documented startup route.
- A model API key is required in the documented configuration; the Bing search key is optional.
- R2R is an integration choice for replacing the default RAG implementation, not a prerequisite for basic startup.
- The README describes the RAG implementation as preliminary and lists Code Interpreter as under development.
- Token-based user isolation can be enabled through configuration; token management requires an administrator token.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Start the API with Docker Compose
Configure and start the self-hosted service, then locate its API and interface documentation.
Before you start
- Docker and Docker Compose installed.
- A model API key to enter in docker-compose.yml; the README says a One API key is supported.
- A Bing search key is optional.
Steps
In the project root, edit docker-compose.yml and set OPENAI_API_KEY; optionally set BING_SUBSCRIPTION_KEY.
Run docker compose up -d.
Use http://127.0.0.1:8086/api/v1 as the API base URL and http://127.0.0.1:8086/docs for interface documentation.
Expected outcome
The README's local API address and documentation address are available as the next integration points after startup.
Guide limitations
- This guide covers startup and locating the API, not a verified deployment or successful request.
- The README recommends configuring R2R to replace the default RAG implementation for better RAG capabilities.
Create an assistant with the OpenAI Python client
Use the documented client example to create an assistant against an Open Assistant API instance.
Before you start
- An Open Assistant API instance and its API base URL.
- The Python openai library, installed with pip install openai.
- An API key value for the client configuration and a model identifier.
Steps
Create an OpenAI client with base_url set to the instance API URL and api_key set to the configured key.
Call client.beta.assistants.create with a name, instructions and model, following the README example.
Expected outcome
The example demonstrates a client request to create an assistant using the OpenAI-compatible interface.
Guide limitations
- The example documents assistant creation; it does not provide a complete account of running a conversation.
- The model identifier shown is an example and should not be read as a model availability guarantee.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- MLT-OSS/open-assistant-api · Official homepage title · retained in its source language