feat: Add streaming endpoint for real-time intervention plan generation

[keithesanks-prog]

Summary

This PR adds a new streaming endpoint that enables real-time token streaming for intervention plan generation using Server-Sent Events (SSE).

Changes

• ✅ Add POST /stream endpoint with Server-Sent Events (SSE) support
• ✅ Implement StreamingModel wrapper for Gemini streaming API
• ✅ Add build_prompt_and_model helper function in gateway.py
• ✅ Update CORS configuration to allow frontend origins (localhost:8501)
• ✅ Fix pydantic version conflict in requirements.txt (>=2.9.0)
• ✅ Add comprehensive documentation for streaming endpoint
• ✅ Support automatic filling of missing EMT score fields

API Details

Endpoint: POST /stream

Content-Type: application/json
Response: text/event-stream

Example Request:

{
  "scores": {
    "EMT1": [35.0, 40.0, 38.0]
  },
  "metadata": {
    "class_id": "A1",
    "deficient_area": "EMT1",
    "num_students": 25
  }
}

Example Response:

data: {"token": "..."}
data: {"token": "..."}
data: {"status": "complete"}

Testing

Test with curl:

curl -N -X POST http://localhost:8000/stream \
  -H "Content-Type: application/json" \
  -d '{"scores":{"EMT1":[35,40,38]}, "metadata":{"class_id":"A1","deficient_area":"EMT1","num_students":25}}'

Documentation

• Updated README.md with streaming endpoint documentation
• Updated DEPLOYMENT.md with testing instructions
• Updated project structure to show new API endpoints directory

Files Changed

• app/api/endpoints/stream.py (new)
• app/api/endpoints/__init__.py (new)
• app/llm/gateway.py (added StreamingModel and build_prompt_and_model)
• app/main.py (added router, updated CORS)
• requirements.txt (fixed pydantic version)
• README.md (added streaming documentation)
• DEPLOYMENT.md (added testing instructions)