- [March 2024] Add gemma-7b and qwen-7b models(based on Ollama)
- [February 2024] Add mistral-7b model (based on Ollama)
- [February 2024] Add gemini-pro model (based on Open API)
- [January 2024] refactor the config-template.yaml to control the backend and the frontend settings at the same time, click to find more introduction about the
config-template.yaml
- [January 2024] Add internlm2-chat-7b model (based on LMDeploy)
- [January 2024] Released version v0.0.1, officially open source!
AOE, an acronym from DOTA2 for Area Of Effect, denotes an ability that can affect a group of targets within a certain area. Here, AOE in AI implies that user can obtain parallel outputs from multiple LLMs with one single prompt at the same time.
Currently, there are many open-source frameworks based on the ChatGPT for chat, but the LGC(LLM Group Chat) framework is still not coming yet.
The emergence of OpenAOE fills this gap: OpenAOE can help LLM researchers, evaluators, engineering developers, and even non-professionals to quickly access the market's well-known commercial and open-source LLMs, providing both single model serial response mode and multi-models parallel response mode.
OpenAOE can:
- return one or more LLMs' answers at the same time by a single prompt.
- provide access to commercial LLM APIs, with default support for gpt3.5, gpt4, Google Palm, Minimax, Claude, Spark, etc., and also support user-defined access to other large model APIs. (API keys need to be prepared in advanced)
- provide access to open-source LLM APIs. ( We recommend to use LMDeploy to deploy with one click)
- provide backend APIs and a WEB-UI to meet the needs of different requirements.
Tip
Require python >= 3.9
We provide three different ways to run OpenAOE: run by pip
, run by docker
and run by source code
as well.
pip install -U openaoe
openaoe -f /path/to/your/config-template.yaml
There are two ways to get the OpenAOE docker image by:
- pull the OpenAOE docker image
docker pull opensealion/openaoe:latest
- or build a docker image
git clone https://github.com/internlm/OpenAOE
cd OpenAOE
docker build . -f docker/Dockerfile -t opensealion/openaoe:latest
docker run -p 10099:10099 -v /path/to/your/config-template.yaml:/app/config.yaml --name OpenAOE opensealion/openaoe:latest
- clone this project
git clone https://github.com/internlm/OpenAOE
- [optional] build the frontend project when the frontend codes are changed
cd OpenAOE/openaoe/frontend
npm install
npm run build
cd OpenAOE # this OpenAOE is the clone directory
pip install -r openaoe/backend/requirements.txt
python -m openaoe.main -f /path/to/your/config-template.yaml
Tip
/path/to/your/config-template.yaml
is a configuration file loaded by OpenAOE at startup,
which contains the relevant configuration information for the LLMs,
including: API URLs, AKSKs, Tokens, etc.
A template configuration yaml file can be found in openaoe/backend/config/config-template.yaml
.
Note that, this config-template.yaml
DOES NOT contain any API access data, you should add them by yourself.
You are always welcome to fork this project to contribute your work and find the TODOs in furture.
If you want to add more LLMs' APIs or features based on OpenAOE, the following info might be helpful.
The technology stack we use includes:
- Backend framework based on python + fastapi;
- Frontend framework based on typescript + Sealion-Client (encapsulated based on React) + Sealion-UI.
- Build tools:
- conda: quickly create a virtual python env to install necessary packages
- npm: build the frontend project
Tip
The build tools can be installed quickly by pip install -U sealion-cli
- Frontend codes are in
openaoe/frontend
- Backend codes are in
openaoe/backend
- Project entry-point is
openaoe/main.py
- Add new model info like
name
,avatar
,provider
, etc inopenaoe/frontend/src/config/model-config.ts
- Add a new model basic API request payload configuration in
openaoe/frontend/src/config/api-config.ts
- Modify your new model's payload specifically in
openaoe/frontend/src/services/fetch.ts
, you may need to change the payload structure and handle corner cases according to your model's API definition.