openclaw - 💡(How to fix) Fix Android: LLM thinking/reasoning blocks rendered as separate chat messages [1 pull requests]

Description

In the OpenClaw Android app, the assistant's internal reasoning/thinking content (the thinking block emitted before the final response) is being rendered as separate, visible chat messages in the conversation view. This appears as one or more message bubbles containing raw JSON, internal notes, and partial tool call details.

Users should never see this content — it should either be hidden entirely or collapsed behind a "show thinking" toggle.

Example of what the user sees as separate messages:

• Internal reasoning text (e.g. "Manifest is healthy. The timeouts are likely...")
• Raw JSON from tool call parameters (e.g. job IDs, session keys, delivery configs)

Environment

• Device: Pixel 8a
• App: OpenClaw Android
• Model: manifest/gpt-5.5 (with thinking enabled on gateway side)

Expected behaviour

• Thinking/reasoning content should not appear as visible messages in the chat
• Only the final assistant response should be rendered

Actual behaviour

• Thinking blocks appear as separate message bubbles
• Raw tool call JSON is sometimes visible
• This leaks internal reasoning and API details to the user

Suggested fix

Filter out type: thinking message parts and tool-call details from the rendered conversation. Only render type: text content from assistant messages. If thinking visibility is desired, hide it behind a collapsible UI element.