docs: Disclosure of AI agents with respect to development process

オープン
  • 何か問題がありますか？
#584 コメント 1 件 リアクション 2 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
30/100
issue の種類
ドキュメント
明瞭さ
説明が足りない
活発さ
静か
技術スタック
github
領域
documentation

調査の方向性

Issue にはドキュメントファイルもテストも記載されていないため、まず要求された開示内容を PR #532 およびリポジトリに既にあるコントリビューター向けのガイダンスと併せて確認します。プロジェクトが、AI agents または自動化が issue や pull request への対応に参加するタイミングを説明する、明確で目に見える声明を備えれば完了です。

索引モデルが issue の本文から書いたものです。

説明

documentation

TL;DR: Recent activity on this project by maintainers has left some ambiguity to their development process, and what users or contributors should expect when engaging with the comfy-cli repo. This should be documented with clear disclosure of how AI has been adopted into the process.

If the observations are correct, that maintainers are no longer directly engaging with issues and PRs via the Github browser interface, but instead abstracted via an API and delegating agents to act on their behalf through developer GH accounts, including acts such as posting comments and PR feedback/approval... transparency of this delegation should be clearly disclosed, as presently the ability to discern any form human interaction in the review process is looking questionable?

I appreciate the labels like agent-coded to provide context, but when AI tooling has been used to manage issues (closing them in bulk with commentary tailored to each message), and also in PR descriptions, discussion and feedback/approval... It is less obvious the prose in these cases is not authored by a human unless certain mannerisms/content is present to distinguish this.

Could the project please disclose this is part of the process used now? There's been an uptick in PR activity for example, but examples like this PR give a strong impression that even the reviewer involved is agentic or through some AI tool generating that comment content?

For users of the project, this concern may be a niche concern. For OSS the added transparency however would be appreciated as the accounts are assumed to belong to real humans while these interactions are either automated or managed via some higher-level interface/service, rather than from Github itself which is what is available to the public eye.

Traditionally engagement with issues/PRs from those with commit access on this project has stalled, and with the recent closing of issues (all actioned in the same minute), it's unclear if there's a real human that can be engaged with as notifications may not reach the account owner if a third-party service is managing issues and PRs via an API alone while AI agents submit and review PRs with whatever degree of oversight by said account owners.

An example of where this is problematic is that any user engaging with the comfy-cli repo is highly likely to do so via the GH web interface (especially for issues), and they're most likely to be doing so directly as a human rather than delegated via an automated task. When a maintainer responds (such as closing an issue for any given reason), attempting to engage with that maintainer may be unsuccessful, and in turn any actions suggested by the comment may result in similar automated actions taking place unintentionally. This is not ideal for the maintainers (whom may be unaware of the attempt to reach out), nor for the public image of the repo (and thus the organization itself).

I don't have an issue with the use of AI, but as a user I'd very much appreciate upfront disclosure when engagement will not only be ignored (for reasons I understand), but attempts to contribute not only becoming stale from lack of upstream engagement but by automation quietly dismissing efforts via accounts that will not respond when a follow-up query to their action is commented (an issue I've experienced personally several times over the past year by the Comfy-Org organization).

Without providing such transparency, these additional practices with AI tooling do accumulate across the various users engaging with the project that it can erode trust and leave a negative impression, which while ComfyUI has reached levels of success a good driver of that was via community and OSS... I would hope this request is fair and won't be dismissed.

主要言語
Python
スター
1k
フォーク
176
平均マージ
1日 11時間
マージ済み PR（30日）
87

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

Comfy-Org/comfy-cli のほかの issue

Comfy-Org/comfy-cli の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。