refactor(chat): extract reusable chat and chat_stream functions for MCP server

[devin-ai-integration]

Summary

This PR extracts reusable chat and chat_stream functions from the CLI into a new vlmrun/client/chat.py module. These functions can now be imported and used by the MCP server or any other client code.

The new module provides:

• chat() - Non-streaming chat completion
• chat_stream() - Streaming chat completion (yields chunks)
• collect_stream() - Helper to collect streaming chunks into a single response
• ChatResponse and ChatStreamChunk dataclasses for structured responses
• ChatError exception for chat-specific errors

The CLI has been refactored to use these core functions while retaining its presentation layer (Rich output, progress spinners, artifact download UI).

Usage example:

from vlmrun.client import VLMRun, chat, chat_stream
from pathlib import Path

client = VLMRun(api_key="your-api-key")
response = chat(client, "Describe this image", inputs=[Path("image.jpg")])
print(response.content)

Updates since last revision

• Fixed test failure on Python 3.10 caused by naming collision between the chat function and chat module. The test now uses sys.modules["vlmrun.client.chat"] to explicitly access the module for mocking.

Review & Testing Checklist for Human

• Verify CLI behavior is unchanged: Run vlmrun chat "Hello" and vlmrun chat "Describe this" -i image.jpg to confirm the CLI still works as expected with both streaming and non-streaming modes
• Test the new functions directly: Import chat and chat_stream from vlmrun.client and verify they work correctly for the MCP server use case
• Check error handling: Verify that invalid models, unsupported file types, and API errors are handled gracefully with clear error messages
• Review naming collision risk: The module vlmrun.client.chat and function chat share the same name - verify this doesn't cause import issues in your MCP server integration

Recommended test plan:

1. Run make test to verify all tests pass
2. Test CLI with various inputs: text-only, with images, with documents
3. Test both --no-stream and default streaming modes
4. Test JSON output mode with --json
5. Import and use chat() function directly in a Python script to verify it works for MCP server integration

Notes

• The CLI now shows "Uploading X file(s)" before upload starts instead of "Uploaded X file(s)" after completion
• collect_stream() uses DEFAULT_MODEL since the model isn't available from stream chunks
• Link to Devin run: https://app.devin.ai/sessions/181d605400e449f5a2a217f7ce682f01
• Requested by: Sudeep Pillai (@spillai)

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring