[AI] Add streaming responses for assistant (real-time output)

[Atharv777]

Summary

Implement streaming responses for AI so users can see output generated in real-time instead of waiting for full responses.

Why this matters

Streaming improves:

• perceived performance
• user experience
• responsiveness of the assistant

This is especially important for longer outputs like contract generation.

Scope

• Modify backend to support streaming responses
• Update frontend to consume streamed data
• Render tokens incrementally in chat UI
• Handle stream interruptions and errors

Acceptance Criteria

• AI responses appear progressively in UI
• No blocking until full response completes
• Errors during streaming are handled gracefully
• Works for both short and long responses

Files Involved

• server/start.py
• server/agent.py
• pages/app/index.jsx
• components/* (chat UI)

Difficulty

Medium

Labels: ai backend frontend enhancement