Skip to content

Latest commit

 

History

History
 
 

microservice

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
.DS_Store
.DS_Store
 
 
.gitignore
.gitignore
 
 
README.md
README.md
 
 
action.py
action.py
 
 
app.py
app.py
 
 
observation.py
observation.py
 
 
requirements.txt
requirements.txt
 
 
session.py
session.py
 
 

README.md

Anterion Agent - Microservice

About

This repo accepts requests from the SWE-agent and sends them to the frontend, for the appropriate user.

Usage

.env

Create a .env file in this repo and fill it with the following details:

OPENAI_API_KEY=<OPENAI_API_KEY_GOES_HERE>
ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY_GOES_HERE>
SWE_AGENT_PATH=<PATH_TO_THE_SWE_AGENT_DIRECTORY>
PYTHON_PATH=<PATH_TO_SWE_AGENT_CONDA_ENV_PYTHON_BIN>

DOCKER_HOST_VOLUME_PATH=<PATH_TO_DOCKER_VOLUME_DIRECTORY>
DOCKER_CONTAINER_VOLUME_PATH=/usr/app

SWE_AGENT_PER_INSTANCE_COST_LIMIT=<MAXIMUM_AMOUNT_OF_USD_TO_SPEND_PER_TASK>
SWE_AGENT_TIMEOUT=25
SWE_AGENT_MODEL_NAME=<FILL_THIS_IN_WITH_DESIRED_MODEL_NAME>

Full List of Models

OpenAI Models:

  • gpt3
  • gpt3-legacy
  • gpt4 (NOTE: This is the default model)
  • gpt4-legacy
  • gpt4-0125
  • gpt3-0125

Anthropic Models:

  • claude (NOTE: This is Claude 2.0)
  • claude-opus
  • claude-sonnet
  • claude-haiku

Installation

Firstly, you will need the SWE-agent conda env setup before using this submodule. Once you have done that, activate the environment by doing the following:

conda activate swe-agent

Then install the requirements for this submodule using:

pip install -r requirements.txt

Then run the program using the following command:

python -m uvicorn app:app --port 3000 --reload