←新闻详情

Cognition 模型 / Devin 博客更新于 10月12日 19:14

Cognition 宣布 Devin 开源计划,向开源维护者赠送 500 个免费 ACU

Announcing The Devin Open Source Initiative | Cognition

AI 导读

Cognition 随 Devin 正式可用(GA)推出 Devin Open Source Initiative,向入选的开源维护者提供 Devin Teams 计划下 500 个免费 ACU。Devin 已在 Anthropic MCP、Dagger、nanoGPT、llama_index、zod 等仓库提交 PR,可后台处理小任务,但官方提醒不要期待 100% 完美的 PR,最终仍需人工把关。

正文

当前显示原文,中文全文译文尚未就绪。

Update: instead of emailing us, please fill out this form.

Today we are announcing the Devin Open Source Initiative.

As part of Devin’s general availability launch, we’re giving selected open source maintainers 500 free ACUs on a Devin Teams plan. Fill out the application form here: https://form.typeform.com/to/AITkMdpD

Open source projects often have a long backlog of issues. There are so many little things to fix that maintainers can quickly become overwhelmed. Devin is the ideal partner on these tasks — by working on minor tasks in the background, Devin helps the maintainers focus on the critical tasks.

Below are some OSS contributions Devin has already helped to make:

Anthropic MCP

Here Devin works on the inspector project, a debugging UI for MCP servers, to add a new feature for displaying the capabilities negotiation.

Devin first uses the browser to research how capability negotiation works:

After Devin writes the code, a few key Devin features enable proper testing:

1. Repo setup. Before starting Devin sessions, we set up Devin’s VM by installing uv for Python MCP servers. 2. Knowledge. We can either tell Devin manually how to test the code, or add this to its knowledge base so that Devin automatically remembers in future sessions.

Given this setup, Devin can verify that the code changes work in the browser:

The code, now tested, is ready for the maintainer to review.

Dagger

Devin’s PR addresses a low-priority task in the Dagger project.

While it’s impressive that Devin is able to eventually solve the issue, several rounds of back and forth are sometimes needed.

Devin’s Github integration works with PR comments and CI checks to make this process smoother. Any PR comments are automatically sent to Devin by a webhook.

Finally, to get the most out of Devin, don’t expect 100% perfect PRs. While Devin can get you 80% of the way there, a human should still ensure quality of the final result. This means e.g. removing stray diffs before merging. In this PR, Devin added debugging logs which will later need to be removed.

Many more

We had Devin contribute to many more open source repos, including:

karpathy/nanoGPT/pull/578 (View Devin run); modelcontextprotocol/inspector/pull/105 (View Devin run); run-llama/llama_index/pull/17201 (View Devin run); hyperliquid-dex/hyperliquid-rust-sdk/pull/68 (View Devin run); dagger/dagger/pull/9130 (View Devin run); modelcontextprotocol/inspector/pull/104 (View Devin run); modelcontextprotocol/servers/pull/256 (View Devin run); ekzhang/bore/pull/146 (View Devin run); colinhacks/zod/pull/3893 (View Devin run); google/go-github/pull/3369 (View Devin run)

We’re excited to see how Devin can help build your open source project! Fill out the application form here: https://form.typeform.com/to/AITkMdpD

来源

Cognition 模型 / Devin 博客