Simplify draft llm

[enyst]

End-user friendly description of the problem this fixes or functionality that this introduces

• Include this change in the Release Notes. If checked, you must provide an end-user friendly description for your change below
  Users can now define a different LLM as a custom LLM config named with a predefined name: draft_editor. To use it for edit drafts, you can, as before, set the agent option to use llm editor (codeact_enable_llm_editor).

---

Give a summary of what the PR does, explaining any non-trivial design decisions

This PR proposes to simplify the code handling the draft editor LLM:

• make it a custom configuration with a specific name: [llm.draft_editor]
  • the user can fill it in with a different model than the main LLM
  • fallback to the main LLM
• clean up the special handling code, it's no longer special
• the actual pre-condition remains the same: the LLMBasedFileEditTool is only enabled if it's set in the agent, codeact_enable_llm_editor.

Also bug fix:

• don't load the draft editor when disabled in config

Note: this was based on top of #4415, the diff will be very simple once that is merged.

---

To run this PR locally, use the following command:

docker run -it --rm   -p 3000:3000   -v /var/run/docker.sock:/var/run/docker.sock   --add-host host.docker.internal:host-gateway   -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:6a7a772-nikolaik   --name openhands-app-6a7a772   docker.all-hands.dev/all-hands-ai/openhands:6a7a772

[enyst]

@openhands-agent-exp Read the diff of this PR carefully. We need to document the draft_editor like a predefined custom LLM configuration. Check where the documentation for custom configs is, and add it there in an appropriate way. It's the first predefined custom config, but we will have more!

[openhands-agent]

OpenHands started fixing the pr! You can monitor the progress here.

[openhands-agent]

Overview of Changes:

✅ Successfully Resolved:

• Added draft_editor documentation in appropriate location
• Created structured "Predefined Custom Configurations" section
• Provided comprehensive config details including TOML, parameters, and use cases
• Maintained documentation consistency
• Established framework for future predefined configs

🔍 Current Status:
All identified issues appear to have been successfully resolved. The changes are thorough and well-structured, addressing both immediate documentation needs and future scalability.

No remaining issues are apparent from the provided information.

[csmith49]

Is this "predefined" in the sense that the user need not specify anything to use this config? I read this as just an example draft editor configuration.

[enyst]

It's actually just "predefined name". 🤔 That is, openhands knows what to do with a configuration with this name, if the user fills it in.
Maybe it's not the best word.

[csmith49]

I see. Might help to clarify that the given configuration is an example that would automatically be used as the draft editor LLM if provided.

[enyst]

Just like the user can define any other custom configuration like [llm.eval-gpt4o] etc, and they can use them or not, use them in an agent, or in general, there are some ... maybe we should call them reserved names? [llm.draft_editor] will be used for the draft llm editor feature.

[csmith49]

"Reserved names" makes sense, I think that fits the concept. Implies there's some hard-coded logic around the name.

[enyst]

Done!

[csmith49]

Much simpler, I like this approach quite a bit.