-
-
Notifications
You must be signed in to change notification settings - Fork 747
Fix: 移除 _remove_image_from_context中的flag逻辑 #2064
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix: 移除 _remove_image_from_context中的flag逻辑 #2064
Conversation
## 审查者指南
此 PR 移除了 `_remove_image_from_context` 方法中基于标志的逻辑,以防止在图像上下文之后意外删除助手消息,从而避免格式错误的 OpenAI API 请求。
### 文件级别更改
| 变更 | 详情 | 文件 |
| ------ | ------- | ----- |
| 从图像上下文过滤中删除基于标志的删除逻辑 | <ul><li>删除标志变量初始化</li><li>删除在检测到图像后跳过消息的条件块</li><li>消除图像内容分支中的标志分配</li></ul> | `astrbot/core/provider/sources/openai_source.py` |
### 可能相关的 issue
- **#2048**: PR 通过更正 `_remove_image_from_context` 中的消息格式,修复了 issue 日志中显示的 API 错误。
- **#2048**: 此 PR 移除了图像处理中的标志逻辑,该逻辑导致图像出现错误,从而解决了 issue 中的 bug。
- **#2048**: 此 PR 移除了 `_remove_image_from_context` 中错误的标志逻辑,该逻辑通过删除带有 `tool_calls` 的助手消息,导致图像转录期间出现 BadRequestError,从而修复了该 issue。
---
<details>
<summary>提示和命令</summary>
#### 与 Sourcery 互动
- **触发新的审查:** 在 pull request 上评论 `@sourcery-ai review`。
- **继续讨论:** 直接回复 Sourcery 的审查评论。
- **从审查评论生成 GitHub issue:** 通过回复审查评论,要求 Sourcery 从审查评论创建一个 issue。您也可以回复审查评论并使用 `@sourcery-ai issue` 从中创建一个 issue。
- **生成 pull request 标题:** 在 pull request 标题中的任何位置写入 `@sourcery-ai` 以随时生成标题。您也可以在 pull request 上评论 `@sourcery-ai title` 以随时(重新)生成标题。
- **生成 pull request 摘要:** 在 pull request 正文中的任何位置写入 `@sourcery-ai summary` 以随时在您想要的位置生成 PR 摘要。您也可以在 pull request 上评论 `@sourcery-ai summary` 以随时(重新)生成摘要。
- **生成审查者指南:** 在 pull request 上评论 `@sourcery-ai guide` 以随时(重新)生成审查者指南。
- **解决所有 Sourcery 评论:** 在 pull request 上评论 `@sourcery-ai resolve` 以解决所有 Sourcery 评论。如果您已经解决了所有评论并且不想再看到它们,这将非常有用。
- **驳回所有 Sourcery 审查:** 在 pull request 上评论 `@sourcery-ai dismiss` 以驳回所有现有的 Sourcery 审查。如果您想重新开始新的审查,这将特别有用 - 不要忘记评论 `@sourcery-ai review` 以触发新的审查!
#### 自定义您的体验
访问您的 [仪表板](https://app.sourcery.ai) 以:
- 启用或禁用审查功能,例如 Sourcery 生成的 pull request 摘要、审查者指南等。
- 更改审查语言。
- 添加、删除或编辑自定义审查说明。
- 调整其他审查设置。
#### 获取帮助
- [联系我们的支持团队](mailto:support@sourcery.ai) 提出问题或反馈。
- 访问我们的 [文档](https://docs.sourcery.ai) 获取详细指南和信息。
- 通过在 [X/Twitter](https://x.com/SourceryAI)、[LinkedIn](https://www.linkedin.com/company/sourcery-ai/) 或 [GitHub](https://github.com/sourcery-ai) 上关注我们,与 Sourcery 团队保持联系。
</details> Original review guide in EnglishReviewer's GuideThis PR removes the flag-based logic in the _remove_image_from_context method to prevent unintended deletion of assistant messages following image contexts, thereby avoiding malformed OpenAI API requests. File-Level Changes
Possibly linked issues
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这么看确实,并不会删除一整个 role="user" 的 part,所以移除 flag 是没问题的。感谢! |
解决了 #2048
Motivation
修复一种导致'error': {'message': "Messages with role 'tool' must be a response to a preceding message with 'tool_calls'", 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}的情况,此情况通常出现在使用图片转述插件时
bug产生原因:
flag逻辑会移除带图片消息的下一条消息,如果被移除的是assistant消息,message可能会违反OpenAI API 要求的消息格式(tool 消息前必须有带 tool_calls 的 assistant 消息)导致api返回400 BadRequestError
以使用图像转述插件为例
第一次调用大模型(如 deepseek-chat),因模型不支持图片而触发 _remove_image_from_context 。
模型返回 tool_calls 请求。
插件执行成功后,ToolLoopAgent 将工具结果和历史记录再次发给大模型。
第二次调用大模型(如 deepseek-chat),因为_remove_image_from_context 检测的是self.req,所以仍能检测到图片,_remove_image_from_context 被再次触发。此时,它的 flag 逻辑会错误地将包含 tool_calls 的 assistant 消息删除。而这个assistant消息后面有转述插件的tool消息,这破坏了 OpenAI API 要求的消息格式(tool 消息前必须有带 tool_calls 的 assistant 消息),最终导致 400 BadRequestError,使整个流程失败。
Modifications
修改位于 astrbot/core/provider/sources/openai_source.py 文件中的 _remove_image_from_context 方法。
移除了方法内部的 flag 判断及其相关逻辑。
直接移除flag不会引起问题的原因:
例如
用户发送了一张“猫弹钢琴”的图片。(视觉模型)AI 回复:“这是一只猫在弹钢琴。”
此时切换到纯文本模型继续对话,历史记录被降级处理。flag 逻辑会删除图片,并且删除 “这是一只猫在弹钢琴”这条回复。
原flag逻辑处理后:[
{"role": "user", "content": "[图片]"}
]
去除flag逻辑后:[
{"role": "user", "content": "[图片]"},
{"role": "assistant", "content": "这是一只猫在弹钢琴。"}
]
可见去除flag逻辑并不影响模型理解上下文,反而flag 逻辑会删除图片,并且删除 “这是一只猫在弹钢琴”这条回复,使模型困惑。
Check
requirements.txt
和pyproject.toml
文件相应位置。好的,这是翻译成中文的 pull request 总结:
Sourcery 总结
移除
_remove_image_from_context
中基于标志的消息移除机制,以避免在工具调用之前剥离助手回复,并修复处理基于图像的插件时出现的invalid_request_error
。Bug 修复:
invalid_request_error
。增强功能:
_remove_image_from_context
。Original summary in English
Summary by Sourcery
Remove the flag-based removal of messages in _remove_image_from_context to avoid stripping assistant replies before tool calls and fix invalid_request_error when handling image-based plugins.
Bug Fixes:
Enhancements: