feat: add extra_body parameter to LLM for provider-specific options

[xujialiu]

Summary

• Adds extra_body parameter to LLM and LLMExecutor, passed through to client.chat.completions.create()
• This enables users to send provider-specific parameters (e.g. {"enable_thinking": False} for Zhipu GLM-4.7)

Problem

Zhipu's GLM-4.7 is a thinking/reasoning model that enables reasoning mode by default. This causes two issues:

1. The model spends ~10x more tokens on internal reasoning (reasoning_content) before producing actual content (content), wasting API credits
2. If the token budget is exhausted during reasoning, delta.content can be empty, causing translation failures

Solution

Pass-through extra_body dict to the OpenAI client so users can disable thinking mode:

llm = LLM(
    key="your-key",
    url="https://open.bigmodel.cn/api/paas/v4",
    model="glm-4.7",
    token_encoding="o200k_base",
    extra_body={"enable_thinking": False},
)

This is a generic solution that works for any provider-specific parameter, not just Zhipu.

Changes

• epub_translator/llm/core.py: Add extra_body parameter to LLM.__init__
• epub_translator/llm/executor.py: Add extra_body to LLMExecutor.__init__ and pass to chat.completions.create()

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 67b87d6 and 17533fa.

📒 Files selected for processing (2)

• epub_translator/llm/core.py
• epub_translator/llm/executor.py

---

Summary by CodeRabbit

Release Notes

• New Features
  • Added extensible configuration support for language models. Users can now pass custom parameters during initialization to fine-tune model behavior, unlock advanced features, and achieve greater control and flexibility in language model configuration. This enhancement enables more sophisticated use cases and better adaptation to specific application requirements.

Walkthrough

This pull request introduces an optional extra_body parameter to the LLM request pipeline. The parameter is added to the LLM class constructor, forwarded to the LLMExecutor constructor, stored as an instance attribute, and passed through to the underlying OpenAI chat completions API call. The change involves no modifications to error handling, control flow, or core logic.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~4 minutes

🚥 Pre-merge checks | ✅ 2

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title follows the required format 'feat(): ' and accurately describes the main change—adding an extra_body parameter to LLM for provider-specific options.
Description check	✅ Passed	The description is well-detailed and directly related to the changeset, explaining the problem, solution, and example usage for the extra_body parameter addition.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Moskize91]

LGTM