fix: 避免 assistant.content=nil 注入 "null" 导致工具历史混杂

[CJackHwang]

Motivation

• 修复当上一轮 assistant 仅返回 tool_calls 且 content = null 时，归一化过程把 nil 序列化为字面量 null 并注入到提示词中，导致工具历史块与自然语言混杂，干扰后续工具识别和执行。

Description

• 在 normalizeOpenAIContentForPrompt 中对 nil 值做了提前返回空字符串处理以避免注入 "null"（修改文件：internal/adapter/openai/message_normalize.go）。
• 新增回归单元测试 TestNormalizeOpenAIMessagesForPrompt_AssistantNilContentDoesNotInjectNullLiteral 覆盖 assistant.content=nil + tool_calls 场景，确保不会注入 null 字面量且工具历史块仍被保留（修改文件：internal/adapter/openai/message_normalize_test.go）。

Testing

• 运行了 go test ./internal/adapter/openai ./internal/util，所有测试均通过。

---

Codex Task

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ds2api	Ready	Preview, Comment	Mar 6, 2026 1:24pm

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0e261ff0a0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve content fallback when text is empty

Delegating normalizeOpenAIContentForPrompt to prompt.NormalizeContent changes behavior for array content items that include both text and content: the previous implementation only consumed text when it was non-empty, then fell back to content, but NormalizeContent treats an empty text string as present and continues, so non-empty content is dropped. For OpenAI-compatible payloads that send placeholder "text": "" with real text in "content", this now strips message text from the prompt and can alter downstream tool-selection and model behavior.

Useful? React with 👍 / 👎.