docs(anthropic): add preset config and expand setup instructions; fix max_tokens truncation

[citizen204]

What

The Anthropic provider (feat: add Anthropic/Claude as LLM provider, #196) was added with minimal documentation and a silent truncation bug compared to the other providers. This PR brings it to parity with the MiniMax provider.

Changes

minitap/mobile_use/services/llm.py

• get_anthropic_llm: adds max_tokens=8192. LangChain's ChatAnthropic defaults max_tokens=1024, which silently truncates responses for long agent plans (screenshot analysis → action descriptions can easily exceed 1 K tokens). Every other provider either passes max_tokens=None (Google/Vertex) or inherits a higher limit from its own defaults; Anthropic needs an explicit value because the Anthropic API requires it. 8 192 is safe for all current Claude models and covers the longest expected agent outputs.

llm-config.defaults.jsonc

• Adds an "anthropic" preset (mirroring the "minimax" preset already present) so users can get started without hand-editing every agent node. Recommended model: claude-sonnet-4-6; fallback: claude-haiku-4-5-20251001.

README.md

• Expands the one-line Anthropic NOTE into a proper **Using Anthropic Claude:** section, exactly matching the MiniMax section's structure (numbered steps + available-models line), so users know to copy the preset and which models to expect.

Why

Without these changes, a user who follows the README to set ANTHROPIC_API_KEY and then manually sets "provider": "anthropic" in their config:

1. Has to guess which model names are valid.
2. Gets responses silently cut off at ~1 024 tokens when the executor or cortex agent produces a longer-than-expected plan.

Test plan

• llm.py change is a one-line max_tokens=8192 addition — no logic change, no tests required.
• llm-config.defaults.jsonc JSON is valid (verify with python -m json.tool after stripping comments, or your preferred JSONC linter).
• README diff renders correctly (standard Markdown; the **Using Anthropic Claude:** heading and numbered list follow the same pattern as the MiniMax section above).

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• rat-review

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: a0f83198-aa81-4c0d-916b-a5eb6e1be4ba

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}