feat: audit and purge legacy backend files, stub Phase 2 AI LLM modules, create comprehensive PAGEWISE_API_DOCUMENTATION.md, and wire DPS test seeding

backend/llm.py

@@ -1,55 +1,23 @@
-import os
-import json
-from openai import AsyncOpenAI
-from sqlmodel import Session, select
-from fastapi.responses import StreamingResponse
+"""
+llm.py — AI Tutor / LLM Streaming Interface
+STATUS: DEFERRED — Phase 2 (AI Coach LLM integration)
+
+This module will provide streaming LLM responses via the local Ollama server
+or a cloud inference endpoint. It is intentionally stubbed out in the current
+production build to avoid dependency on the openai SDK and a running LLM server.
+
+When ready to enable:
+1. Add `openai==1.14.0` back to requirements.txt
+2. Set LLM_BASE_URL, LLM_MODEL_NAME, OPENAI_API_KEY in backend/.env
+3. Replace this stub with the full implementation below.
+"""
 from uuid import UUID
+from typing import AsyncIterator
 
-import models
 
-# Point the standard AsyncOpenAI client to your local Ollama server
-client = AsyncOpenAI(
-    base_url=os.getenv("LLM_BASE_URL", "http://host.docker.internal:11434/v1"),
-    api_key=os.getenv("OPENAI_API_KEY", "ollama"), # Required by the SDK, but Ollama ignores it
-)
-
-async def stream_tutor_response(user_id: UUID, user_message: str, session: Session):
-    # 1. Fetch Context: Get the student's active weak areas
-    active_mistakes = session.exec(
-        select(models.Mistake)
-        .where(models.Mistake.user_id == user_id)
-        .where(models.Mistake.is_mastered == False)
-    ).all()
-    
-    weak_topics = list(set([m.topic for m in active_mistakes]))
-    context_string = ", ".join(weak_topics) if weak_topics else "General Syllabus"
-
-    # 2. Construct the highly-specific System Prompt
-    system_prompt = f"""
-    You are PragatiVeer's elite AI Tutor for JEE & NEET aspirants. 
-    Your goal is to guide students to the answer, not just give it to them.
-    
-    STUDENT CONTEXT:
-    This student currently struggles with the following topics: {context_string}.
-    Keep this in mind and relate explanations back to foundational concepts in these areas if applicable.
-    
-    RULES:
-    1. Respond in the student's preferred language (English or Hindi).
-    2. Use LaTeX wrapped in single $ for inline math and $$ for block equations.
-    3. Keep responses concise and focused on problem-solving steps.
+async def stream_tutor_response(user_id: UUID, user_message: str, session) -> AsyncIterator[str]:
     """
-
-    # 3. Call the LLM
-    response = await client.chat.completions.create(
-        model=os.getenv("LLM_MODEL_NAME", "gemma"),
-        messages=[
-            {"role": "system", "content": system_prompt},
-            {"role": "user", "content": user_message}
-        ],
-        stream=True,
-    )
-
-    # 4. Yield tokens as they arrive
-    async for chunk in response:
-        if chunk.choices[0].delta.content is not None:
-            yield chunk.choices[0].delta.content
+    Stub: yields a single message explaining AI is coming soon.
+    Replace with real LLM streaming call in Phase 2.
+    """
+    yield "🤖 AI Tutor is coming soon. This feature will be enabled once LLM infrastructure is provisioned."