fix: preserve paragraph breaks between text blocks in bridge responses

[wujieli0207]

Summary

• Fixed .join('') → .join('\n\n') in conversation engine's text block assembly, which was causing multi-paragraph responses to render as a single wall of text in Telegram

Root cause

When Claude's response includes tool calls (Read, Edit, Bash, etc.), the SSE stream splits the response into multiple text content blocks:

Text Block 1: "Let me check the code..."  → tool_use: Read
Text Block 2: "Found the issue. Here's the fix..."  → tool_use: Edit  
Text Block 3: "Done. The changes are..."

These blocks were joined with .join('') (empty string), producing:

Let me check the code...Found the issue. Here's the fix...Done. The changes are...

Now joined with .join('\n\n'):

Let me check the code...

Found the issue. Here's the fix...

Done. The changes are...

Files changed

File	Change
src/lib/bridge/conversation-engine.ts	.join('') → .join('\n\n') in both normal and error-recovery paths

Test plan

• Send a message that triggers tool usage (e.g. "read file X and explain it") → verify response paragraphs have proper line breaks in Telegram
• Send a simple message with no tool calls → verify single text block still renders correctly
• Verify the desktop UI is unaffected (it uses a different rendering path)

🤖 Generated with Claude Code

[op7418]

LGTM. Simple and correct fix.

.join('') → .join('\n\n') for text block assembly is the right call. When Claude's response spans multiple text blocks (interleaved with tool calls), joining without separators creates a wall of text. The \n\n separator preserves paragraph breaks.

Both the normal path and the error-recovery path are updated consistently. No side effects on the desktop UI (which uses a different rendering pipeline).