inspect-swe

library0.2.63pypypi✓ verified 79d ago

Inspect Swe is a library for software engineering agents, built on top of Inspect AI. It provides tools and workflows for creating AI agents that can interact with code, run tests, and perform software engineering tasks. Version 0.2.50 is current, with active development.

llm-agents ai-ml testing

pip install inspect-swe

github ↗package ↗docs ↗

INSTALL

IMPORT

SIG · INSPECT-SWE

inspect-swe

llm-agentspythonv0.2.63

Install

19.4s avg

Import

—

Disk

294MB

Pass rate

8/ 10

Env Coverage8 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v0.2.63 · pip install

no network on importno background threads

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

glibc

py 3.10

✓ —

✓ 23.8s

py 3.11

✓ —

✓ 20.5s

py 3.12

✓ —

✓ 16.8s

py 3.13

✓ —

✓ 16.7s

py 3.9

✕ build_error

294MB installed

● package 294MB

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

ACPAgent

✓ from inspect_swe import ACPAgent

✗ from inspect_swe import SweAgent

ACPAgentParams

✓ from inspect_swe import ACPAgentParams

SandboxPlatform

✓ from inspect_swe import SandboxPlatform

Basic usage: create a SweAgent and run it on a simple coding task.

from inspect_swe import SweAgent from inspect_ai import Task task = Task("Write a function that adds two numbers")

Debug

Known issues

breakingBreaking change in v0.2.0: SweAgent no longer accepts 'model' as a constructor argument. Use 'config.model' instead.

fix

Change SweAgent(model='gpt-4') to SweAgent(config=SweConfig(model='gpt-4'))

affects: >=0.2.0

deprecatedThe 'run' method is deprecated since v0.2.40. Use 'solve' or 'execute' instead.

fix

Replace agent.run(task) with agent.solve(task) or agent.execute(task)

affects: >=0.2.40

gotchaSweEnvironment requires a running Docker daemon when using local Docker backend. Failure to start Docker will raise RuntimeError.

fix

Ensure Docker Desktop or daemon is running before creating SweEnvironment.

affects: all

gotchaWhen using SweEnvironment with 'sandbox' backend, the environment is reset after each task. State is not persisted.

fix

Use 'persistent' backend if state across tasks is needed.

affects: all

Upgrade

Version history

0.2.63latest on PyPI · released Jun 10, 2026

Audit

Dependencies

inspect-airequiredCore dependency - inspect-swe builds on Inspect AI framework

Agent activity

28 hits · last 30 days

node

Amazon

OpenAI (training)

Resources

githubhttps://github.com/meridianlabs-ai/inspect_swe ↗

packagehttps://pypi.org/project/inspect-swe/ ↗

docshttps://meridianlabs-ai.github.io/inspect_swe/ ↗