Tools for running a SWE agent using Composio platform
Project description
SWE Development Kit
Table of Contents
Overview
swekit
is a framework for building SWE agents on by utilising composio tooling ecosystem. SWE Kit allows you to
- Scaffold agents which works out-of-the-box with choice of your agentic framework,
crewai
,llamaindex
, etc... - Tools to add or optimise your agent's abilities
- Benchmark your agents against
SWE-bench
Dependencies
Before getting started, ensure you have the following set up:
-
Installation:
pip install swekit composio-core
-
Install agentic framework of your choice and the Composio plugin for the same: Here we're using
crewai
for the example:pip install crewai composio-crewai
-
GitHub Access Token:
The agent requires a github access token to work with your repositories, You can create one at https://github.com/settings/tokens with necessary permissions and export it as an environment variable using
export GITHUB_ACCESS_TOKEN=<your_token>
-
LLM Configuration: You also need to setup API key for the LLM provider you're planning to use. By default the agents scaffolded by
swekit
usesopenai
client, so exportOPENAI_API_KEY
before running your agent
Getting Started
Creating a new agent
-
Scaffold your agent using:
swekit scaffold crewai -o <path>
This creates a new agent in
<path>/agent
with four key files:main.py
: Entry point to run the agent on your issueagent.py
: Agent definition (edit this to customise behaviour)prompts.py
: Agent promptsbenchmark.py
: SWE-Bench benchmark runner
-
Run the agent:
cd agent python main.py
You'll be prompted for the repository name and issue.
Docker Environment
The SWE-agent runs in Docker by default for security and isolation. This sandboxes the agent's operations, protecting against unintended consequences of arbitrary code execution.
To run locally instead, modify workspace_env
in agent/agent.py
. Use caution, as this bypasses Docker's protective layer.
Running the Benchmark
SWE-Bench is a comprehensive benchmark designed to evaluate the performance of software engineering agents. It comprises a diverse collection of real-world issues from popular Python open-source projects, providing a robust testing environment.
To run the benchmark:
- Ensure Docker is installed and running on your system.
- Execute the following command:
cd agent python benchmark.py --test-split=<test_split>
- By default,
python benchmark.py
runs only 1 test instance. - Specify a test split ratio to run more tests, e.g.,
--test-split=1:300
runs 300 tests.
- By default,
Note: We utilize SWE-Bench-Docker to ensure each test instance runs in an isolated container with its specific environment and Python version.
To extend the functionality of the SWE agent by adding new tools or extending existing ones, refer to the Development Guide.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for swekit-0.1.0rc1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1196ea2f0a6ed1d4d250e8343831e027f94a8fcfd1121f421e9e3b8e306fbf98 |
|
MD5 | 67142ec073e1ad856b31621b4d3dd1ac |
|
BLAKE2b-256 | aed5360e65648566dab3c5907c04b9220f64160f2cc06b9bb107136169b99b40 |