{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-11-02T02:00:36.453Z","headline":"OpenRouter 教程：用 Responses API 构建可在 Linux 沙箱验证 bug 的 AI 代码审查脚本","description":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py，通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，把变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug 并返回 V4A 补丁。","url":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","mainEntityOfPage":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","datePublished":"2026-11-01T00:00:00.000Z","dateModified":"2026-11-01T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://openrouter.ai/blog/tutorials/ai-code-review","https://aihot.news/items/yww15c701iat30p8ttjf029ij"],"canonicalUrl":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","directAnswer":{"@type":"Answer","text":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py 脚本。该脚本通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，将变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug，并返回可用标志应用的 V4A 补丁。","url":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","dateCreated":"2026-11-01T00:00:00.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"OpenRouter：Announcements source article","url":"https://openrouter.ai/blog/tutorials/ai-code-review","datePublished":"2026-11-01T00:00:00.000Z","provider":{"@type":"Organization","name":"OpenRouter：Announcements","url":"https://openrouter.ai/blog/tutorials/ai-code-review"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/yww15c701iat30p8ttjf029ij","datePublished":"2026-11-01T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/yww15c701iat30p8ttjf029ij"}}],"aggregationSource":"OpenRouter：Announcements","originalPublisher":{"name":"OpenRouter：Announcements","url":"https://openrouter.ai/blog/tutorials/ai-code-review"},"geoDeepAnswer":null,"article":{"id":"yww15c701iat30p8ttjf029ij","slug":"yww15c701iat30p8ttjf029ij","url":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","title":"OpenRouter 教程：用 Responses API 构建可在 Linux 沙箱验证 bug 的 AI 代码审查脚本","title_en":"","summary":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py，通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，把变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug 并返回 V4A 补丁。","source":"OpenRouter：Announcements","sourceUrl":"https://openrouter.ai/blog/tutorials/ai-code-review","aiHotUrl":"https://aihot.news/items/yww15c701iat30p8ttjf029ij","publishedAt":"2026-11-01T00:00:00.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["An AI code reviewer checks a Git change for bugs and security issues before you push or merge. Reviewers that only read the diff have to guess what the rest of the file looks like and whether a suspected bug actually fails at runtime, and their suggested fixes come back as text you copy in by hand.","The reviewer in this guide uses the OpenRouter Responses API：https://openrouter.ai/docs/api_reference/responses/overview to run your changed files in a Linux sandbox, confirm each bug by running a check, and return fixes as patches you can apply with a flag. Across the four models we tested on a five-bug change, a review takes one to four minutes and costs between $0.015 and $0.045.","The script combines three OpenRouter server tools：https://openrouter.ai/docs/guides/features/server-tools and one function tool on the Responses API:","You need Python 3.11 or newer, Git, and an OpenRouter API key from the quickstart：https://openrouter.ai/docs/quickstart exported in your shell:","The Files API, Shell tool, and Apply Patch tool run on the global endpoint ( https://openrouter.ai/api/v1 ); in-region endpoints return 403 . Your changed files and context files are uploaded to your OpenRouter workspace and sent to the model provider, so only run the script on code you are allowed to share and keep secrets out of those files.","Save the following script as review.py in the root of your repository:","Several choices in review.py come from issues we hit while building it:","Create a small repository where the existing tests pass even after a change introduces five bugs.","Save this five-bug change as buggy.diff :","Commit the initial files, apply buggy.diff , and run the existing tests:","The two existing tests still pass because they do not cover what the diff changed. (Our sandbox had Python 3.11.14 and SQLite 3.50.4; unittest works out of the box, while pytest is not installed.)","Now run the reviewer with the test file as context:","Here is the output from one of our anthropic/claude-haiku-5.5 runs:","The script exits 1 because three findings are HIGH . Four findings are confirmed by running code in the sandbox, and their commands and outputs are saved in .git/ai-review.json under evidence . The removed conn.close() is unconfirmed because CPython closes the local connection when the function returns. Pass --apply to copy the patched app/orders.py into your working tree.","Save this script as .git/hooks/pre-push and run chmod +x .git/hooks/pre-push :","When we committed review.py on main , committed the five-bug change on top, and ran git push origin main , the hook ran the review ( 5 findings, 1 patch, $0.0117, 96 s ), exited 1 , and Git printed error: failed to push some refs . The remote stayed at the previous commit.","Save this workflow as .github/workflows/ai-review.yml . We did not run this workflow in GitHub Actions; we ran the same git show and review.py commands locally.","Running git show \"$BASE:review.py\" uses the base branch’s copy of the script so a pull request cannot modify the reviewer that checks it, and python -I keeps the pull request’s files off Python’s import path. Fork pull requests skip the job, and timeout-minutes: 15 covers the one-to-four-minute review time.","Add OPENROUTER_API_KEY under Settings, Secrets and variables, Actions, and make review a required status check in your branch protection rule：https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/managing-protected-branches/managing-a-branch-protection-rule. For a broader CI gating setup, see how to gate pull requests on LLM evals in CI：https://openrouter.ai/blog/tutorials/how-to-gate-pull-requests-on-llm-evals-in-ci/.","We ran three reviews per model on the sample above with default settings and --context tests/test_orders.py . “Bugs found” counts the five planted bugs across completed reviews. “Confirmed by running code” means the model ran a check in the sandbox that showed the bug. “Patch fixed all five” means we applied the accepted patches to a clean copy and our own tests passed for all five bugs.","Start with anthropic/claude-haiku-5.5 when you want the lowest cost per review ($0.0153), or anthropic/claude-sonnet-5.5 ( --model anthropic/claude-sonnet-5.5 ) when you want faster reviews (60 s) and patches that fixed all five bugs in every run ($0.0447). You can compare more models in the model catalog：https://openrouter.ai/models and test them on your own repository with Ori Eval：https://openrouter.ai/docs/guides/ori/eval.","Each review bill includes model tokens plus container：https://openrouter.ai/docs/guides/features/containers sandbox time at $0.0001 per second. The Files API has no separate charge (up to 100 MiB per file and 10 GiB per workspace, and workspace files do not expire).","Sandbox billing runs from the first shell command until the response completes, so model thinking time during the review also counts. As a result, a slower model pays more for the sandbox even when its tokens are cheap: sandbox time was $0.0074 to $0.0159 of each Haiku review, compared with $0.0047 to $0.0068 for the faster Sonnet runs.","A diff-only Chat Completions reviewer costs about $0.001 per review on Haiku 5.5. Running the code in a sandbox costs 10 to 40 times more, and in exchange the findings come with reproducer evidence and tested patches.","The script uploads your changed files, any context files you pass, and the Git diff to your OpenRouter workspace with the Files API, and the model provider receives their contents during the review. Uploaded files cannot be downloaded back and can only be used inside OpenRouter shell sandboxes. The script deletes every uploaded file in a finally block before it exits.","In our three-run tests on a five-bug change, a review averaged $0.0153 with Claude Haiku 5.5, $0.0167 with GPT-6 Luna, $0.0307 with DeepSeek V4.1 Flash, and $0.0447 with Claude Sonnet 5.5. That total includes model tokens and the Linux sandbox at $0.0001 per second from the first shell command until the response completes.","By default the sandbox has no outbound internet access, so the model cannot run pip install. It can still read code that imports third-party packages and propose patches, though any check that imports those packages will fail. If your checks need packages from PyPI, you can set network_policy on the shell tool to allowlist hosts such as pypi.org and files.pythonhosted.org.","No. When the model proposes a fix with the apply_patch tool, the script applies it to a temporary copy of your files and tells the model whether it applied cleanly. Your working tree is only updated if you pass the —apply flag.","Model usage data, product updates, and research reports. One email each week."],"articleImages":[{"sourceUrl":"https://openrouter.ai/blog/images/ai-code-review.png","alt":"Build an AI Code Reviewer with OpenRouter","afterParagraph":0,"url":"/media/articles/yww15c701iat30p8ttjf029ij/573bb9bb4f538839.png"},{"sourceUrl":"https://openrouter.ai/blog/images/ai-code-review-dark.png","alt":"Build an AI Code Reviewer with OpenRouter","afterParagraph":0,"url":"/media/articles/yww15c701iat30p8ttjf029ij/ed1a614684564bd5.png"}],"mediaStatus":"ok","articleBodyZh":["AI 代码审查器会在你推送或合并之前检查 Git 变更中的漏洞和安全问题。仅查看差异的审查员必须猜测文件的其余部分是什么样子，以及怀疑的漏洞是否会在运行时失败，而且他们建议的修复通常以文本形式返回，你需要手动复制进来。","本指南中的审查器使用 OpenRouter Responses API：https://openrouter.ai/docs/api_reference/responses/overview 在 Linux 沙箱中运行你更改的文件，通过运行检查确认每个漏洞，并返回可以应用的补丁修复。在我们测试的四个模型上，对一个包含五个漏洞的变更进行审查，大约需要一到四分钟，费用在 $0.015 到 $0.045 之间。","该脚本结合使用了三个 OpenRouter 服务器工具：https://openrouter.ai/docs/guides/features/server-tools 和 Responses API 的一个函数工具：","你需要 Python 3.11 或更新版本、Git，以及从快速入门获取的 OpenRouter API 密钥：https://openrouter.ai/docs/quickstart 并在你的 shell 中导出：","Files API、Shell 工具和 Apply Patch 工具运行在全球端点 ( https://openrouter.ai/api/v1 )；区域端点会返回 403。你更改的文件和上下文文件会被上传到你的 OpenRouter 工作区并发送给模型提供者，因此只在你有权共享的代码上运行脚本，并确保文件中不包含机密。","将以下脚本保存为 review.py，并放在你的仓库根目录：","review.py 中的若干选择来自我们在构建过程中遇到的问题：","创建一个小型仓库，即使在更改引入五个漏洞后，现有测试仍然通过。","将这五个漏洞的变更保存为 buggy.diff：","提交初始文件，应用 buggy.diff，并运行现有的测试：","这两个现有测试仍然通过，因为它们没有覆盖 diff 所更改的内容。（我们的沙箱使用 Python 3.11.14 和 SQLite 3.50.4；unittest 开箱即用，而 pytest 未安装。）","现在使用测试文件作为上下文运行审查器：","以下是我们的一次 anthropic/claude-haiku-5.5 运行的输出：","该脚本退出码为 1，因为有三个发现被评为 HIGH。通过在沙箱中运行代码确认了四个发现，其命令和输出保存在 .git/ai-review.json 的 evidence 下。移除的 conn.close() 未被确认，因为 CPython 会在函数返回时关闭本地连接。使用 --apply 将修补后的 app/orders.py 复制到你的工作区。","将此脚本保存为 .git/hooks/pre-push 并运行 chmod +x .git/hooks/pre-push：","当我们在 main 分支提交 review.py，在其上提交了五个 bug 的更改，并运行 git push origin main 时，hook 运行了 review（5 个发现，1 个补丁，$0.0117，96 秒），退出码为 1，Git 打印错误：failed to push some refs。远程仓库保持在上一次提交。","将此工作流保存为 .github/workflows/ai-review.yml。我们没有在 GitHub Actions 中运行此工作流，而是在本地运行相同的 git show 和 review.py 命令。","运行 git show \"$BASE:review.py\" 使用的是基础分支的脚本副本，因此 pull 请求无法修改检查它的审查器，而 python -I 可以把 pull 请求的文件从 Python 的导入路径中排除。分叉的 pull 请求会跳过此作业，timeout-minutes: 15 覆盖了一到四分钟的审查时间。","在 Settings、Secrets and variables、Actions 下添加 OPENROUTER_API_KEY，并在分支保护规则中将 review 设置为必需的状态检查：https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/managing-protected-branches/managing-a-branch-protection-rule。要进行更广泛的 CI 门控设置，请参见如何在 CI 中基于 LLM 评估对 pull 请求进行门控：https://openrouter.ai/blog/tutorials/how-to-gate-pull-requests-on-llm-evals-in-ci/。","我们针对上述示例使用默认设置和 --context tests/test_orders.py 对每个模型运行了三次审查。“发现的 bug”统计了在完成的审查中发现的五个预植入 bug。“通过运行代码确认”意味着模型在沙箱中运行检查并显示了该 bug。“补丁修复了全部五个”意味着我们将接受的补丁应用到干净副本，并且我们自己的测试通过了所有五个 bug。","当你想以最低每次审查成本（$0.0153）进行审查时，可以从 anthopric/claude-haiku-5.5 开始；如果你想要更快的审查（60 秒）以及每次运行都修复所有五个漏洞的补丁，可以使用 anthopric/claude-sonnet-5.5（--model anthopric/claude-sonnet-5.5）。你可以在模型目录中比较更多模型：https://openrouter.ai/models，并使用 Ori Eval 在你自己的仓库中进行测试：https://openrouter.ai/docs/guides/ori/eval。","每次审查账单包括模型令牌加上容器：https://openrouter.ai/docs/guides/features/containers 沙盒时间，每秒 $0.0001。Files API 没有额外费用（每个文件最多 100 MiB，每个工作区最多 10 GiB，工作区文件不会过期）。","沙盒计费从第一个 shell 命令开始直到响应完成，因此审查期间模型的思考时间也会被计算在内。因此，即使令牌便宜，较慢的模型也会为沙盒支付更多费用：每次 Haiku 审查的沙盒时间费用为 $0.0074 到 $0.0159，而快速的 Sonnet 审查为 $0.0047 到 $0.0068。","仅差异的 Chat Completions 审查在 Haiku 5.5 上每次审查约 $0.001。运行代码在沙盒中成本要高 10 到 40 倍，但作为回报，这些发现附带了可复现的证据和已测试的补丁。","该脚本将你修改的文件、你提供的任何上下文文件以及 Git 差异上传到你的 OpenRouter 工作区，通过 Files API，在审查期间模型提供者会收到其内容。上传的文件不能下载，只能在 OpenRouter shell 沙盒中使用。脚本在退出前的 finally 块中删除所有上传的文件。","在我们对一个五漏洞更改进行的三次测试中，使用 Claude Haiku 5.5 的平均审查费用为 $0.0153，GPT-6 Luna 为 $0.0167，DeepSeek V4.1 Flash 为 $0.0307，Claude Sonnet 5.5 为 $0.0447。该总费用包括模型令牌和从第一个 shell 命令开始直到响应完成的 Linux 沙盒费用，每秒 $0.0001。","默认情况下，沙箱没有出站互联网访问，因此模型无法运行 pip install。不过，它仍然可以读取导入第三方包的代码并提出补丁，尽管任何导入这些包的检查都会失败。如果您的检查需要来自 PyPI 的包，您可以在 shell 工具上设置 network_policy，以允许访问 pypi.org 和 files.pythonhosted.org 等主机。","不能。当模型使用 apply_patch 工具提出修复时，脚本会将其应用到文件的临时副本上，并告诉模型是否应用成功。只有在您使用 —apply 标志时，您的工作区才会更新。","模型使用数据、产品更新和研究报告。每周发送一次电子邮件。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py 脚本。该脚本通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，将变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug，并返回可用标志应用的 V4A 补丁。","background":"来源说明：仅读 diff 的 AI 审查者只能猜测文件其余内容、无法确认疑似 bug 是否真的在运行时失败，其修复建议也需手工复制。OpenRouter 在一个含五个 bug 的变更上测试了四个模型，单次审查耗时一到四分钟、成本 0.015 到 0.045 美元。相关工具运行在全局端点，区域内端点返回 403。","viewpoint":"Aioga 判断：让模型在沙箱中实际运行检查来验证 bug、并以可直接应用的补丁形式返回修复，可能比纯文本 diff 审查更能降低误报和手工粘贴成本，值得关注。Aioga 观察：脚本上传文件到工作区并发送给模型提供方的提示，表明使用场景存在代码分享边界。","implications":"可能影响：开发者或可参考该教程构建验证式代码审查流程，但需要满足 Python 3.11+、Git 与 API key 等前置条件。建议仅在允许共享的代码上运行并避免包含敏感信息。来源的四个模型、五个 bug 样本不足以代表所有场景表现。","nextStep":"后续观察：可关注 OpenRouter 后续是否发布更多 Responses API 与服务端工具教程、区域内端点何时支持相关工具，以及社区在真实仓库中对该审查脚本准确性与成本的实际反馈。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-11-01T03:03:29.200Z","sourceHash":"db2ce63969d1f60c","review":{"approved":true,"groundedness":96,"clarity":91,"duplicationRisk":18,"blockingIssues":[],"notes":["“可用标志应用的 V4A 补丁”表述略不自然，可改为“可通过标志应用的 V4A 补丁”。","“相关工具运行在全局端点”可明确为 Files API、Shell 工具和 Apply Patch 工具，以避免范围歧义。","“四个模型、五个 bug 样本不足以代表所有场景表现”属于合理的审慎推论；如需更严谨，可改为“不能据此推断所有场景的表现”。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","OpenRouter：Announcements"],"translations":{"zh-CN":{"title":"OpenRouter 教程：用 Responses API 构建可在 Linux 沙箱验证 bug 的 AI 代码审查脚本","summary":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py，通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，把变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug 并返回 V4A 补丁。","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter 教程：用 Responses API 构建可在 Linux 沙箱验证 bug 的 AI 代码审查脚本 - Aioga AI资讯","description":"OpenRouter 发布教程，介绍一个 233 行、仅用标准库的 review.py，通过 Responses API 组合 Files API、Shell 沙箱工具和 Apply Patch 工具，把变更文件上传到托管 Linux 沙箱，让模型实际运行检查确认 bug 并返回 V4A 补丁。","url":"https://www.aioga.com/news/yww15c701iat30p8ttjf029ij/","articleBody":["AI 代码审查器会在你推送或合并之前检查 Git 变更中的漏洞和安全问题。仅查看差异的审查员必须猜测文件的其余部分是什么样子，以及怀疑的漏洞是否会在运行时失败，而且他们建议的修复通常以文本形式返回，你需要手动复制进来。","本指南中的审查器使用 OpenRouter Responses API：https://openrouter.ai/docs/api_reference/responses/overview 在 Linux 沙箱中运行你更改的文件，通过运行检查确认每个漏洞，并返回可以应用的补丁修复。在我们测试的四个模型上，对一个包含五个漏洞的变更进行审查，大约需要一到四分钟，费用在 $0.015 到 $0.045 之间。","该脚本结合使用了三个 OpenRouter 服务器工具：https://openrouter.ai/docs/guides/features/server-tools 和 Responses API 的一个函数工具：","你需要 Python 3.11 或更新版本、Git，以及从快速入门获取的 OpenRouter API 密钥：https://openrouter.ai/docs/quickstart 并在你的 shell 中导出：","Files API、Shell 工具和 Apply Patch 工具运行在全球端点 ( https://openrouter.ai/api/v1 )；区域端点会返回 403。你更改的文件和上下文文件会被上传到你的 OpenRouter 工作区并发送给模型提供者，因此只在你有权共享的代码上运行脚本，并确保文件中不包含机密。","将以下脚本保存为 review.py，并放在你的仓库根目录：","review.py 中的若干选择来自我们在构建过程中遇到的问题：","创建一个小型仓库，即使在更改引入五个漏洞后，现有测试仍然通过。","将这五个漏洞的变更保存为 buggy.diff：","提交初始文件，应用 buggy.diff，并运行现有的测试：","这两个现有测试仍然通过，因为它们没有覆盖 diff 所更改的内容。（我们的沙箱使用 Python 3.11.14 和 SQLite 3.50.4；unittest 开箱即用，而 pytest 未安装。）","现在使用测试文件作为上下文运行审查器：","以下是我们的一次 anthropic/claude-haiku-5.5 运行的输出：","该脚本退出码为 1，因为有三个发现被评为 HIGH。通过在沙箱中运行代码确认了四个发现，其命令和输出保存在 .git/ai-review.json 的 evidence 下。移除的 conn.close() 未被确认，因为 CPython 会在函数返回时关闭本地连接。使用 --apply 将修补后的 app/orders.py 复制到你的工作区。","将此脚本保存为 .git/hooks/pre-push 并运行 chmod +x .git/hooks/pre-push：","当我们在 main 分支提交 review.py，在其上提交了五个 bug 的更改，并运行 git push origin main 时，hook 运行了 review（5 个发现，1 个补丁，$0.0117，96 秒），退出码为 1，Git 打印错误：failed to push some refs。远程仓库保持在上一次提交。","将此工作流保存为 .github/workflows/ai-review.yml。我们没有在 GitHub Actions 中运行此工作流，而是在本地运行相同的 git show 和 review.py 命令。","运行 git show \"$BASE:review.py\" 使用的是基础分支的脚本副本，因此 pull 请求无法修改检查它的审查器，而 python -I 可以把 pull 请求的文件从 Python 的导入路径中排除。分叉的 pull 请求会跳过此作业，timeout-minutes: 15 覆盖了一到四分钟的审查时间。","在 Settings、Secrets and variables、Actions 下添加 OPENROUTER_API_KEY，并在分支保护规则中将 review 设置为必需的状态检查：https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/managing-protected-branches/managing-a-branch-protection-rule。要进行更广泛的 CI 门控设置，请参见如何在 CI 中基于 LLM 评估对 pull 请求进行门控：https://openrouter.ai/blog/tutorials/how-to-gate-pull-requests-on-llm-evals-in-ci/。","我们针对上述示例使用默认设置和 --context tests/test_orders.py 对每个模型运行了三次审查。“发现的 bug”统计了在完成的审查中发现的五个预植入 bug。“通过运行代码确认”意味着模型在沙箱中运行检查并显示了该 bug。“补丁修复了全部五个”意味着我们将接受的补丁应用到干净副本，并且我们自己的测试通过了所有五个 bug。","当你想以最低每次审查成本（$0.0153）进行审查时，可以从 anthopric/claude-haiku-5.5 开始；如果你想要更快的审查（60 秒）以及每次运行都修复所有五个漏洞的补丁，可以使用 anthopric/claude-sonnet-5.5（--model anthopric/claude-sonnet-5.5）。你可以在模型目录中比较更多模型：https://openrouter.ai/models，并使用 Ori Eval 在你自己的仓库中进行测试：https://openrouter.ai/docs/guides/ori/eval。","每次审查账单包括模型令牌加上容器：https://openrouter.ai/docs/guides/features/containers 沙盒时间，每秒 $0.0001。Files API 没有额外费用（每个文件最多 100 MiB，每个工作区最多 10 GiB，工作区文件不会过期）。","沙盒计费从第一个 shell 命令开始直到响应完成，因此审查期间模型的思考时间也会被计算在内。因此，即使令牌便宜，较慢的模型也会为沙盒支付更多费用：每次 Haiku 审查的沙盒时间费用为 $0.0074 到 $0.0159，而快速的 Sonnet 审查为 $0.0047 到 $0.0068。","仅差异的 Chat Completions 审查在 Haiku 5.5 上每次审查约 $0.001。运行代码在沙盒中成本要高 10 到 40 倍，但作为回报，这些发现附带了可复现的证据和已测试的补丁。","该脚本将你修改的文件、你提供的任何上下文文件以及 Git 差异上传到你的 OpenRouter 工作区，通过 Files API，在审查期间模型提供者会收到其内容。上传的文件不能下载，只能在 OpenRouter shell 沙盒中使用。脚本在退出前的 finally 块中删除所有上传的文件。","在我们对一个五漏洞更改进行的三次测试中，使用 Claude Haiku 5.5 的平均审查费用为 $0.0153，GPT-6 Luna 为 $0.0167，DeepSeek V4.1 Flash 为 $0.0307，Claude Sonnet 5.5 为 $0.0447。该总费用包括模型令牌和从第一个 shell 命令开始直到响应完成的 Linux 沙盒费用，每秒 $0.0001。","默认情况下，沙箱没有出站互联网访问，因此模型无法运行 pip install。不过，它仍然可以读取导入第三方包的代码并提出补丁，尽管任何导入这些包的检查都会失败。如果您的检查需要来自 PyPI 的包，您可以在 shell 工具上设置 network_policy，以允许访问 pypi.org 和 files.pythonhosted.org 等主机。","不能。当模型使用 apply_patch 工具提出修复时，脚本会将其应用到文件的临时副本上，并告诉模型是否应用成功。只有在您使用 —apply 标志时，您的工作区才会更新。","模型使用数据、产品更新和研究报告。每周发送一次电子邮件。"]},"en":{"title":"OpenRouter Tutorial: Build an AI Code Review Script Using Responses API to Verify Bugs in a Linux Sandbox","summary":"OpenRouter released a tutorial, introducing a 233-line review.py script using only the standard library. By combining Files API, Shell sandbox tools, and Apply Patch tools through the Responses API, change files are uploaded to a hosted Linux sandbox, allowing the model to actually run, check, and confirm bugs, then return a V4A patch.","category":"Industry","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter Tutorial: Build an AI Code Review Script Using Responses API to Verify Bugs in a Linux Sandbox - Aioga AI News","description":"OpenRouter released a tutorial, introducing a 233-line review.py script using only the standard library. By combining Files API, Shell sandbox tools, and Apply Patch tools through...","url":"https://www.aioga.com/en/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:01:46.980Z"},"ja":{"title":"OpenRouterチュートリアル:Responses APIを使ってLinuxサンドボックスのバグを検証するためのAIコードレビュースクリプトを作成する","summary":"OpenRouterは標準ライブラリのみを使用した233行 review.py を紹介するチュートリアルを公開しました。Responds APIとFiles API、Shell Box サンドボックスツール、Apply Patchツールを組み合わせることで、変更ファイルを管理されたLinuxサンドボックスにアップロードします。モデルはバグを確認し確認し、V4Aパッチを返すために実行します。","category":"業界動向","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouterチュートリアル:Responses APIを使ってLinuxサンドボックスのバグを検証するためのAIコードレビュースクリプトを作成する - Aioga AIニュース","description":"OpenRouterは標準ライブラリのみを使用した233行 review.py を紹介するチュートリアルを公開しました。Responds APIとFiles API、Shell Box サンドボックスツール、Apply Patchツールを組み合わせることで、変更ファイルを管理されたLinuxサンドボックスにアップロードします。モデルはバグを確認し確認し、V4...","url":"https://www.aioga.com/ja/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:01:51.203Z"},"ko":{"title":"OpenRouter 튜토리얼: 리눅스 샌드박스에서 버그를 검증하기 위해 Responses API를 사용해 AI 코드 리뷰 스크립트를 구축하세요","summary":"OpenRouter는 표준 라이브러리만 사용하는 233라인 review.py 을 소개하는 튜토리얼을 공개했습니다. Responds API와 Files API, 셸 박스 샌드박스 도구, Apply Patch 도구를 결합하면, 변경 파일을 관리되는 리눅스 샌드박스에 업로드합니다. 모델은 이를 실행하여 버그를 확인하고 V4A 패치를 반환합니다.","category":"업계 동향","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter 튜토리얼: 리눅스 샌드박스에서 버그를 검증하기 위해 Responses API를 사용해 AI 코드 리뷰 스크립트를 구축하세요 - Aioga AI 뉴스","description":"OpenRouter는 표준 라이브러리만 사용하는 233라인 review.py 을 소개하는 튜토리얼을 공개했습니다. Responds API와 Files API, 셸 박스 샌드박스 도구, Apply Patch 도구를 결합하면, 변경 파일을 관리되는 리눅스 샌드박스에 업로드합니다. 모델은 이를 실행하여 버그를 확인하고 V4A...","url":"https://www.aioga.com/ko/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:01:51.206Z"},"es":{"title":"Tutorial de OpenRouter: construir un script de revisión de código AI capaz de verificar bugs en un sandbox de Linux usando la API de Responses","summary":"OpenRouter ha publicado un tutorial que presenta un review.py de 233 líneas, usando solo la biblioteca estándar, que combina la API de Responses con la API de Files, herramientas de sandbox Shell y la herramienta Apply Patch, subiendo los archivos modificados a un sandbox de Linux alojado para que el modelo ejecute la verificación de bugs y devuelva el parche V4A.","category":"Industria","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Tutorial de OpenRouter: construir un script de revisión de código AI capaz de verificar bugs en un sandbox de Linux usando la API de Responses - Aioga Noticias de IA","description":"OpenRouter ha publicado un tutorial que presenta un review.py de 233 líneas, usando solo la biblioteca estándar, que combina la API de Responses con la API de Files, herramientas d...","url":"https://www.aioga.com/es/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:14.343Z"},"fr":{"title":"Tutoriel OpenRouter : construire un script de révision de code IA capable de vérifier les bugs dans un bac à sable Linux via l'API Responses","summary":"OpenRouter a publié un tutoriel présentant un fichier review.py de 233 lignes, utilisant uniquement la bibliothèque standard, qui combine l'API Responses, l'API Files, l'outil SandBox Shell et l'outil Apply Patch pour envoyer les fichiers modifiés vers un bac à sable Linux hébergé, permettant au modèle d'exécuter réellement les vérifications de bugs et de retourner un correctif V4A.","category":"Industrie","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Tutoriel OpenRouter : construire un script de révision de code IA capable de vérifier les bugs dans un bac à sable Linux via l'API Responses - Aioga Actualités IA","description":"OpenRouter a publié un tutoriel présentant un fichier review.py de 233 lignes, utilisant uniquement la bibliothèque standard, qui combine l'API Responses, l'API Files, l'outil Sand...","url":"https://www.aioga.com/fr/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:14.375Z"},"de":{"title":"OpenRouter Tutorial: Erstellen eines KI-Code-Review-Skripts, das Bugs in einer Linux-Sandbox überprüfen kann","summary":"OpenRouter veröffentlichte ein Tutorial zu review.py, 233 Zeilen lang, nur mit Standardbibliothek, das über die Responses API zusammen mit Files API, Shell-Sandbox-Tools und Apply-Patch-Tool Änderungsdateien in eine gehostete Linux-Sandbox hochlädt, damit das Modell sie tatsächlich ausführen, Bugs prüfen und einen V4A-Patch zurückgeben kann.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter Tutorial: Erstellen eines KI-Code-Review-Skripts, das Bugs in einer Linux-Sandbox überprüfen kann - Aioga KI-News","description":"OpenRouter veröffentlichte ein Tutorial zu review.py, 233 Zeilen lang, nur mit Standardbibliothek, das über die Responses API zusammen mit Files API, Shell-Sandbox-Tools und Apply-...","url":"https://www.aioga.com/de/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:13.066Z"},"pt-BR":{"title":"Tutorial do OpenRouter: crie um script de revisão de código com a Responses API para validar bugs em um sandbox Linux","summary":"O OpenRouter publicou um tutorial que apresenta o review.py, um script de 233 linhas que usa apenas a biblioteca padrão. Por meio da Responses API, ele combina a Files API, a ferramenta de sandbox Shell e a ferramenta Apply Patch para enviar arquivos alterados a um sandbox Linux hospedado, onde o modelo pode executar verificações para confirmar bugs e retornar patches V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Tutorial do OpenRouter: crie um script de revisão de código com a Responses API para validar bugs em um sandbox Linux - Aioga Notícias de IA","description":"O OpenRouter publicou um tutorial que apresenta o review.py, um script de 233 linhas que usa apenas a biblioteca padrão. Por meio da Responses API, ele combina a Files API, a ferra...","url":"https://www.aioga.com/pt-BR/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:21.469Z"},"ru":{"title":"Руководство OpenRouter: создаём скрипт ИИ-проверки кода, который проверяет баги в песочнице Linux с помощью Responses API","summary":"OpenRouter опубликовала руководство по review.py — скрипту на 233 строки, использующему только стандартную библиотеку. С помощью Responses API он объединяет Files API, инструмент Shell в песочнице и инструмент Apply Patch: изменённые файлы загружаются в управляемую песочницу Linux, где модель может запустить проверки, подтвердить наличие багов и вернуть патч V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Руководство OpenRouter: создаём скрипт ИИ-проверки кода, который проверяет баги в песочнице Linux с помощью Responses API - Aioga Новости ИИ","description":"OpenRouter опубликовала руководство по review.py — скрипту на 233 строки, использующему только стандартную библиотеку. С помощью Responses API он объединяет Files API, инструмент S...","url":"https://www.aioga.com/ru/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:24.554Z"},"ar":{"title":"دليل OpenRouter: بناء برنامج نصي لمراجعة الشيفرات بالذكاء الاصطناعي يستطيع التحقق من الأخطاء البرمجية داخل بيئة Linux معزولة باستخدام Responses API","summary":"نشرت OpenRouter دليلاً يشرح برنامج review.py مؤلفاً من 233 سطراً، ولا يستخدم سوى المكتبات القياسية. ويجمع البرنامج بين Files API وأداة Shell sandbox وأداة Apply Patch عبر Responses API، لرفع الملفات التي طرأت عليها تغييرات إلى بيئة Linux معزولة مستضافة، حيث يشغّل النموذج الفحوصات فعلياً للتحقق من الأخطاء البرمجية وإرجاع تصحيح V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"دليل OpenRouter: بناء برنامج نصي لمراجعة الشيفرات بالذكاء الاصطناعي يستطيع التحقق من الأخطاء البرمجية داخل بيئة Linux معزولة باستخدام Responses API - Aioga أخبار الذكاء الاصطناعي","description":"نشرت OpenRouter دليلاً يشرح برنامج review.py مؤلفاً من 233 سطراً، ولا يستخدم سوى المكتبات القياسية. ويجمع البرنامج بين Files API وأداة Shell sandbox وأداة Apply Patch عبر Responses...","url":"https://www.aioga.com/ar/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:24.004Z"},"hi":{"title":"OpenRouter ट्यूटोरियल: Responses API का उपयोग करके Linux सैंडबॉक्स में बग सत्यापित करने वाले AI कोड रिव्यू स्क्रिप्ट का निर्माण","summary":"OpenRouter ने एक ट्यूटोरियल जारी किया, जिसमें एक 233 लाइन का, केवल स्टैण्डर्ड लाइब्रेरी का review.py दिखाया गया है, जो Responses API के माध्यम से Files API, Shell सैंडबॉक्स टूल और Apply Patch टूल को संयोजित करके परिवर्तित फ़ाइलों को होस्ट किए गए Linux सैंडबॉक्स में अपलोड करता है, ताकि मॉडल वास्तविक रूप से चलाकर बग की जाँच करे और V4A पैच लौटाए।","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter ट्यूटोरियल: Responses API का उपयोग करके Linux सैंडबॉक्स में बग सत्यापित करने वाले AI कोड रिव्यू स्क्रिप्ट का निर्माण - Aioga AI समाचार","description":"OpenRouter ने एक ट्यूटोरियल जारी किया, जिसमें एक 233 लाइन का, केवल स्टैण्डर्ड लाइब्रेरी का review.py दिखाया गया है, जो Responses API के माध्यम से Files API, Shell सैंडबॉक्स टूल और...","url":"https://www.aioga.com/hi/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:47.611Z"},"it":{"title":"Tutorial di OpenRouter: creare con la Responses API uno script di revisione del codice AI in grado di verificare i bug in una sandbox Linux","summary":"OpenRouter ha pubblicato un tutorial che presenta review.py, uno script di 233 righe che usa soltanto la libreria standard. Tramite la Responses API combina la Files API, lo strumento Shell sandbox e lo strumento Apply Patch, caricando i file modificati in una sandbox Linux gestita, dove il modello può eseguire controlli concreti per verificare i bug e restituire una patch V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Tutorial di OpenRouter: creare con la Responses API uno script di revisione del codice AI in grado di verificare i bug in una sandbox Linux - Aioga Notizie IA","description":"OpenRouter ha pubblicato un tutorial che presenta review.py, uno script di 233 righe che usa soltanto la libreria standard. Tramite la Responses API combina la Files API, lo strume...","url":"https://www.aioga.com/it/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:32.684Z"},"nl":{"title":"OpenRouter tutorial: Bouw AI-code reviewscripts met behulp van de Responses API om bugs in de Linux-sandbox te verifiëren","summary":"OpenRouter bracht een tutorial uit waarin een 233-regels review.py wordt geïntroduceerd die alleen de standaardbibliotheek gebruikt. Door de Responds API te combineren met de Files API, de Shell box sandbox tool en de Apply Patch tool, upload je het wijzigingsbestand naar een beheerde Linux-sandbox. Het model voert het uit om de bug te controleren en te bevestigen en de V4A-patch terug te sturen.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter tutorial: Bouw AI-code reviewscripts met behulp van de Responses API om bugs in de Linux-sandbox te verifiëren - Aioga AI-nieuws","description":"OpenRouter bracht een tutorial uit waarin een 233-regels review.py wordt geïntroduceerd die alleen de standaardbibliotheek gebruikt. Door de Responds API te combineren met de Files...","url":"https://www.aioga.com/nl/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:02:48.832Z"},"tr":{"title":"OpenRouter Eğitimi: Linux sandbox'ta hata doğrulayabilen AI kod inceleme betiği oluşturmak için Responses API kullanımı","summary":"OpenRouter, 233 satırlık ve yalnızca standart kütüphane kullanan review.py dosyasını tanıtan bir eğitim yayınladı; bu dosya Responses API'yi Files API, Shell sandbox araçları ve Apply Patch aracı ile birleştirerek değişiklik dosyalarını barındırılan Linux sandbox'a yükler, böylece model hatayı gerçek olarak çalıştırıp kontrol eder ve V4A yama döndürür.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"OpenRouter Eğitimi: Linux sandbox'ta hata doğrulayabilen AI kod inceleme betiği oluşturmak için Responses API kullanımı - Aioga AI Haberleri","description":"OpenRouter, 233 satırlık ve yalnızca standart kütüphane kullanan review.py dosyasını tanıtan bir eğitim yayınladı; bu dosya Responses API'yi Files API, Shell sandbox araçları ve Ap...","url":"https://www.aioga.com/tr/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:03:11.763Z"},"vi":{"title":"Hướng dẫn OpenRouter: Sử dụng Responses API xây dựng script AI kiểm tra mã có thể xác nhận bug trong sandbox Linux","summary":"OpenRouter phát hành hướng dẫn, giới thiệu một file review.py dài 233 dòng, chỉ dùng thư viện tiêu chuẩn, thông qua Responses API kết hợp Files API, công cụ sandbox Shell và Apply Patch, tải các file thay đổi lên sandbox Linux được quản lý, để mô hình thực sự chạy kiểm tra xác nhận bug và trả về bản vá V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Hướng dẫn OpenRouter: Sử dụng Responses API xây dựng script AI kiểm tra mã có thể xác nhận bug trong sandbox Linux - Tin tức AI Aioga","description":"OpenRouter phát hành hướng dẫn, giới thiệu một file review.py dài 233 dòng, chỉ dùng thư viện tiêu chuẩn, thông qua Responses API kết hợp Files API, công cụ sandbox Shell và Apply...","url":"https://www.aioga.com/vi/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:03:10.669Z"},"id":{"title":"Tutorial OpenRouter: Membuat skrip review kode AI yang dapat memverifikasi bug di sandbox Linux menggunakan Responses API","summary":"OpenRouter merilis tutorial yang memperkenalkan review.py sepanjang 233 baris, hanya menggunakan pustaka standar, melalui Responses API yang menggabungkan Files API, alat sandbox Shell, dan alat Apply Patch, mengunggah file yang diubah ke sandbox Linux yang dihosting, sehingga model dapat menjalankan pemeriksaan untuk memastikan bug dan mengembalikan patch V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Tutorial OpenRouter: Membuat skrip review kode AI yang dapat memverifikasi bug di sandbox Linux menggunakan Responses API - Berita AI Aioga","description":"OpenRouter merilis tutorial yang memperkenalkan review.py sepanjang 233 baris, hanya menggunakan pustaka standar, melalui Responses API yang menggabungkan Files API, alat sandbox S...","url":"https://www.aioga.com/id/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:03:09.862Z"},"th":{"title":"บทเรียน OpenRouter: สร้างสคริปต์รีวิวโค้ด AI โดยใช้ Responses API เพื่อตรวจสอบบั๊กในแซนด์บ็อกซ์ของ Linux","summary":"OpenRouter ได้ปล่อยบทเรียนแนะนํา review.py 233 บรรทัดที่ใช้เฉพาะไลบรารีมาตรฐาน โดยการรวม Responds API กับ Files API, เครื่องมือ Shellbox sandbox และเครื่องมือ Apply Patch คุณจะอัปโหลดไฟล์การเปลี่ยนแปลงไปยัง sandbox Linux ที่มีการจัดการ โมเดลจะรันเพื่อตรวจสอบและยืนยันบั๊ก และส่งคืนแพตช์ V4A","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"บทเรียน OpenRouter: สร้างสคริปต์รีวิวโค้ด AI โดยใช้ Responses API เพื่อตรวจสอบบั๊กในแซนด์บ็อกซ์ของ Linux - ข่าว AI Aioga","description":"OpenRouter ได้ปล่อยบทเรียนแนะนํา review.py 233 บรรทัดที่ใช้เฉพาะไลบรารีมาตรฐาน โดยการรวม Responds API กับ Files API, เครื่องมือ Shellbox sandbox และเครื่องมือ Apply Patch คุณจะอัปโ...","url":"https://www.aioga.com/th/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:03:35.957Z"},"pl":{"title":"Poradnik OpenRouter: Buduj skrypty do przeglądu kodu AI za pomocą API Responses do weryfikacji błędów w piaskownicy Linuksa","summary":"OpenRouter opublikował samouczek wprowadzający 233-liniowy review.py korzystający wyłącznie ze standardowej biblioteki. Łącząc API Responds z API Files, narzędziem sandbox Shell i narzędziem Apply Patch, przesyłasz plik zmiany do zarządzanego sandboxu Linuksa. Model uruchamia go, aby sprawdzić i potwierdzić błąd oraz zwrócić poprawkę V4A.","category":"行业动态","source":"OpenRouter：Announcements","aggregationSource":"OpenRouter：Announcements","pageTitle":"Poradnik OpenRouter: Buduj skrypty do przeglądu kodu AI za pomocą API Responses do weryfikacji błędów w piaskownicy Linuksa - Aioga Wiadomości AI","description":"OpenRouter opublikował samouczek wprowadzający 233-liniowy review.py korzystający wyłącznie ze standardowej biblioteki. Łącząc API Responds z API Files, narzędziem sandbox Shell i...","url":"https://www.aioga.com/pl/news/yww15c701iat30p8ttjf029ij/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"65657f6f39307a5f","translatedAt":"2026-11-01T01:03:35.873Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}