Strip chat command prefixes from model input context

[Copilot]

The model was receiving the raw /think … command text as context, causing it to interpret the command instead of the user’s question.

• Input handling
  • Centralized input extraction in extractInputFromMessage, removing the command prefix (e.g., /think) before passing text to the model context.
• Tests
  • Added focused tests ensuring command text is stripped when a trigger command is present, including bot-mention forms (/think@bot) and plain messages without a command prefix.

Example:

trigger := &config.ChatTrigger{Command: "think"}
msg := &tb.Message{Text: "/think hello world"}

extractInputFromMessage(msg, trigger) // "hello world"

Original prompt

---

This section details on the original issue you should resolve

<issue_title>给大模型的上下文里面把命令部分去掉</issue_title>
<issue_description>> /think 是否任意实数都能使用对n求和 x^n, n为任意整数包括负数 的形式表示

如上，调用大模型的时候 /think 命令会代入上下文，导致大模型理解错误，返回下面这种内容：

我目前没有执行指令的权限，也无法处理 /think 这类命令。如果你有任何数学问题、需要解释或推导，请直接告诉我，我会尽力为你解答！</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes 给大模型的上下文里面把命令部分去掉 #703

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[icceey]

@copilot 单元测试不够全面，至少还应该包含两种情况：

• 指向型命令/think@bot
• 纯消息没有命令前缀

[icceey]

@copilot 重试

[hugefiver]

@copilot 重试

[icceey]

@copilot 再重试

[Copilot]

@copilot 重试

Added coverage for both cases and confirmed the command body is preserved when no prefix exists. Commit 3fe0139.

[hugefiver]

@copilot 单元测试不够全面，至少还应该包含两种情况：

• 指向型命令/think@bot
• 纯消息没有命令前缀

@copilot

[Copilot]

Pull request overview

This PR fixes a critical bug where chat command prefixes (like /think) were incorrectly included in the AI model's input context, causing the model to misinterpret commands instead of processing the actual user question.

Changes:

• Fixed inverted error handling logic in command prefix stripping (changed if err != nil to if err == nil)
• Refactored input extraction into a dedicated extractInputFromMessage function for better testability
• Added comprehensive unit tests covering command stripping scenarios

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
chat/chat.go	Fixed bug in command prefix stripping logic and extracted input handling into testable function
chat/chat_input_test.go	Added unit tests verifying command prefixes are correctly stripped from model input

[Copilot]

The test coverage could be more comprehensive. Consider adding test cases for:

1. Messages with Caption instead of Text (e.g., photo captions)
2. Empty trigger command (trigger.Command == "")
3. Command with no arguments (e.g., just "/think" without any text after)
4. Empty message text and caption

These edge cases would ensure the function handles all scenarios robustly.

[Copilot]

@copilot 单元测试不够全面，至少还应该包含两种情况：

• 指向型命令/think@bot
  ...

Covered both bot-mention and no-prefix inputs in tests; see commit 3fe0139.