Update app.py

app.py

@@ -1,7 +1,7 @@
 import os
 import requests
 import httpx
-from fastapi import FastAPI
+from fastapi import FastAPI, HTTPException, Security, Header
 from pydantic import BaseModel
 from typing import Optional
 from llama_cpp import Llama
@@ -9,17 +9,16 @@ from fastembed import TextEmbedding
 
 app = FastAPI()
 
-# Qdrant Configuration (unchanged)
 QDRANT_URL = os.environ["QDRANT_URL"].rstrip("/")
 QDRANT_API_KEY = os.environ["QDRANT_API_KEY"]
 COLLECTION = "well_vectors"
 
-# Groq API Configuration
-GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
-GROQ_API_URL = "https://api.groq.com/openai/v1/chat/completions"
-GROQ_MODEL = "llama-3.3-70b-versatile"  # Best for scientific reasoning
+SERVICE_API_KEY = os.environ.get("SERVICE_API_KEY")
+SERVICE_API_URL = "https://api.groq.com/openai/v1/chat/completions"
+SERVICE_MODEL = "llama-3.3-70b-versatile"
+
+EDYX_ACCESS_TOKEN = os.environ.get("EDYX_ACCESS_TOKEN")
 
-# Physics system prompt for Groq
 PHYSICS_SYSTEM_PROMPT = """You are an expert physics researcher and teacher.
 You are given retrieved scientific material from a physics knowledge base.
 Your job:
@@ -30,7 +29,6 @@ Your job:
 - Produce a clean, coherent, human-readable explanation
 Style: Clear, structured, graduate-level physics understanding."""
 
-# Local fallback model (only loaded when needed)
 local_llm = None
 
 def get_local_llm():
@@ -45,7 +43,6 @@ def get_local_llm():
         )
     return local_llm
 
-# Embedder (always needed for RAG search)
 embedder = TextEmbedding(
     model_name="BAAI/bge-large-en-v1.5",
 )
@@ -55,12 +52,16 @@ class QueryRequest(BaseModel):
     top_k: Optional[int] = 5
     max_tokens: Optional[int] = 512
 
+async def verify_token(x_edyx_token: str = Header(None)):
+    if EDYX_ACCESS_TOKEN and x_edyx_token != EDYX_ACCESS_TOKEN:
+        raise HTTPException(status_code=403, detail="Unauthorized: Invalid Access Token")
+    return x_edyx_token
+
 @app.get("/")
 def root():
-    return {"status": "edyx-phy running", "mode": "groq-primary"}
+    return {"status": "edyx-phy running", "mode": "accelerated-primary"}
 
 def search_qdrant(question: str, top_k: int):
-    """Search Qdrant for relevant physics context"""
     vector = [float(x) for x in next(embedder.embed(question))]
 
     r = requests.post(
@@ -93,28 +94,25 @@ def search_qdrant(question: str, top_k: int):
     context = "\n\n".join(collected)[:12000]
     return context, len(hits)
 
-async def call_groq_api(question: str, context: str, max_tokens: int):
-    """Try to get response from Groq API"""
-    if not GROQ_API_KEY:
-        raise Exception("GROQ_API_KEY not configured")
+async def call_service_api(question: str, context: str, max_tokens: int):
+    if not SERVICE_API_KEY:
+        raise Exception("Service API key not configured")
     
     user_prompt = f"""CONTEXT (retrieved evidence):
 {context}
-
 QUESTION:
 {question}
-
 Now produce a high-quality physics explanation that a serious learner would trust."""
 
     async with httpx.AsyncClient(timeout=60.0) as client:
         response = await client.post(
-            GROQ_API_URL,
+            SERVICE_API_URL,
             headers={
                 "Content-Type": "application/json",
-                "Authorization": f"Bearer {GROQ_API_KEY}"
+                "Authorization": f"Bearer {SERVICE_API_KEY}"
             },
             json={
-                "model": GROQ_MODEL,
+                "model": SERVICE_MODEL,
                 "messages": [
                     {"role": "system", "content": PHYSICS_SYSTEM_PROMPT},
                     {"role": "user", "content": user_prompt}
@@ -125,13 +123,12 @@ Now produce a high-quality physics explanation that a serious learner would trus
         )
         
         if response.status_code != 200:
-            raise Exception(f"Groq API error: {response.status_code} - {response.text}")
+            raise Exception(f"Service API error: {response.status_code} - {response.text}")
         
         data = response.json()
         return data["choices"][0]["message"]["content"]
 
 def call_local_model(question: str, context: str, max_tokens: int):
-    """Fallback to local llama model - YOUR ORIGINAL LOGIC"""
     llm = get_local_llm()
     
     prompt = f"""
@@ -145,7 +142,7 @@ This material may include:
 Your job:
 - Use the retrieved material as grounding evidence
 - Ignore irrelevant technical artifacts (paths, array shapes, file names)
-- If the retrieved information is incomplete or fragmented, use your general physics knowledge to complete the explanation
+- If the retrieved information is incomplete, use your physics knowledge to complete the explanation
 - Do NOT invent specific papers, experiments, or citations
 - Do NOT mention datasets, storage paths, or indexing systems
 - Produce a clean, coherent, human-readable explanation
@@ -172,7 +169,7 @@ Now produce a high-quality physics explanation that a serious learner would trus
     
     return out["choices"][0]["text"].strip()
 
-@app.post("/v1/query")
+@app.post("/v1/query", dependencies=[Security(verify_token)])
 async def query(req: QueryRequest):
 
     context, sources = search_qdrant(req.question, req.top_k)
@@ -184,14 +181,14 @@ async def query(req: QueryRequest):
         return {"answer": "No relevant scientific data found.", "sources_used": 0}
 
     try:
-        answer = await call_groq_api(req.question, context, req.max_tokens)
+        answer = await call_service_api(req.question, context, req.max_tokens)
         return {
             "answer": answer,
             "sources_used": sources,
             "source": "primary"
         }
     except Exception as e:
-        print(f"Groq API failed: {e}, falling back to local model...")
+        print(f"Service API failed: {e}, falling back to local model...")
     
 
     try: