fix(conversation): set context length to 1 for gemini-2.5-flash-image model

[kangfenmao]

Summary

• Fixed continuous conversation failures for the gemini-2.5-flash-image (Nano Banana) model by setting its context length to 1
• Added isGeminiGenerateImageModel utility function to identify Nano Banana models
• This ensures the model only receives the most recent message, preventing context-related errors

Changes

• Added new utility function isGeminiGenerateImageModel in image.ts
• Modified ConversationService.prepareMessages to override contextCount to 1 for gemini-2.5-flash-image models in ConversationService.ts
• Removed the previous +2 offset from the final takeRight call to align with the new contextCount handling

Test Plan

• Test continuous conversations with gemini-2.5-flash-image model
• Verify that conversations no longer fail after the first message
• Confirm that other models are not affected by this change
• Run yarn build:check to ensure no lint/type errors

[DeJeune]

Note

This issue/comment/review was translated by Claude.

Now that it's set to 1, is there still a way to edit the image generated in the previous assistant message?

---

Original Content

设置成1了，还有办法编辑上一条助手消息生成的图片吗？

[kangfenmao]

Note

This issue/comment/review was translated by Claude.

You can't edit it anymore, each time there's only one current submitted message, no more context.

---

Original Content

不能编辑了，每次只有当前提交的一条消息，没有上下文了

[DeJeune]

Note

This issue/comment/review was translated by Claude.

You can modify it in src/renderer/src/aiCore/prepareParams/messageConverter.ts

---

Original Content

src/renderer/src/aiCore/prepareParams/messageConverter.ts 可以在这里修改