Updated

app.py

@@ -1,11 +1,11 @@
 import os
 import logging
-from fastapi import FastAPI, Request
+from fastapi import FastAPI, Request, Header, HTTPException
 from fastapi.responses import JSONResponse
+from fastapi.middleware.cors import CORSMiddleware
+from huggingface_hub.utils import HfHubHTTPError
 from langchain.prompts import PromptTemplate
 from langchain_huggingface import HuggingFaceEndpoint
-from huggingface_hub.utils import HfHubHTTPError
-from langchain.schema import HumanMessage  # ✅ For conversational inputs
 from vector import query_vector
 
 # ==============================
@@ -14,6 +14,11 @@ from vector import query_vector
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger("AgriCopilot")
 
+# ==============================
+# Config
+# ==============================
+PROJECT_API_KEY = os.getenv("PROJECT_API_KEY", "super-secret-123")  # 🔑 Change in prod
+
 # ==============================
 # App Init
 # ==============================
@@ -29,96 +34,114 @@ async def root():
 @app.exception_handler(Exception)
 async def global_exception_handler(request: Request, exc: Exception):
     logger.error(f"Unhandled error: {exc}")
-    return JSONResponse(
-        status_code=500,
-        content={"error": str(exc)},
-    )
+    return JSONResponse(status_code=500, content={"error": str(exc)})
+
+# ==============================
+# Auth Helper
+# ==============================
+def check_auth(authorization: str | None):
+    if not PROJECT_API_KEY:
+        return
+    if not authorization or not authorization.startswith("Bearer "):
+        raise HTTPException(status_code=401, detail="Missing bearer token")
+    token = authorization.split(" ", 1)[1]
+    if token != PROJECT_API_KEY:
+        raise HTTPException(status_code=403, detail="Invalid token")
 
 # ==============================
-# MODELS PER ENDPOINT (Meta Models, Conversational)
+# HuggingFace Model Config
+# ==============================
+default_model = dict(
+    temperature=0.3,
+    top_p=0.9,
+    do_sample=True,
+    repetition_penalty=1.1,
+    max_new_tokens=1024
+)
+
+# ==============================
+# CHAINS
 # ==============================
 
 # 1. Crop Doctor
 crop_template = PromptTemplate(
     input_variables=["symptoms"],
-    template="You are an agricultural crop doctor. A farmer reports: {symptoms}. Diagnose the most likely disease and suggest treatments in simple farmer-friendly language."
-)
-crop_llm = HuggingFaceEndpoint(
-    repo_id="meta-llama/Llama-3.2-11B-Vision-Instruct",
-    task="conversational"
+    template="You are AgriCopilot, a multilingual AI assistant for farmers. "
+             "A farmer reports: {symptoms}. Diagnose the likely disease and suggest "
+             "clear, farmer-friendly treatments."
 )
+crop_llm = HuggingFaceEndpoint(repo_id="meta-llama/Llama-3.2-11B-Vision-Instruct", **default_model)
+crop_chain = crop_template | crop_llm
 
 # 2. Multilingual Chat
 chat_template = PromptTemplate(
     input_variables=["query"],
-    template="You are a multilingual AI assistant for farmers. Answer clearly in the same language as the user. Farmer says: {query}"
-)
-chat_llm = HuggingFaceEndpoint(
-    repo_id="meta-llama/Llama-3.1-8B-Instruct",
-    task="conversational"
+    template="You are AgriCopilot, a supportive multilingual AI guide. "
+             "Respond in the SAME language as the user. Farmer says: {query}"
 )
+chat_llm = HuggingFaceEndpoint(repo_id="meta-llama/Llama-3.1-8B-Instruct", **default_model)
+chat_chain = chat_template | chat_llm
 
 # 3. Disaster Summarizer
 disaster_template = PromptTemplate(
     input_variables=["report"],
-    template="You are an AI disaster assistant. Summarize the following report for farmers in simple steps: {report}"
-)
-disaster_llm = HuggingFaceEndpoint(
-    repo_id="meta-llama/Llama-3.1-8B-Instruct",
-    task="conversational"
+    template="You are AgriCopilot, an AI disaster assistant. "
+             "Summarize the following report into 3–5 short, actionable steps farmers can follow: {report}"
 )
+disaster_llm = HuggingFaceEndpoint(repo_id="meta-llama/Llama-3.1-8B-Instruct", **default_model)
+disaster_chain = disaster_template | disaster_llm
 
-# 4. Marketplace Recommendation
+# 4. Marketplace Recommender
 market_template = PromptTemplate(
     input_variables=["product"],
-    template="You are an agricultural marketplace recommender. Farmer wants to sell or buy: {product}. Suggest possible matches and advice."
-)
-market_llm = HuggingFaceEndpoint(
-    repo_id="meta-llama/Llama-3.1-8B-Instruct",
-    task="conversational"
+    template="You are AgriCopilot, an agricultural marketplace advisor. "
+             "Farmer wants to sell or buy: {product}. Suggest options, advice, and safe trade tips."
 )
+market_llm = HuggingFaceEndpoint(repo_id="meta-llama/Llama-3.1-8B-Instruct", **default_model)
+market_chain = market_template | market_llm
 
 # ==============================
-# ENDPOINTS
+# ENDPOINTS (with auth)
 # ==============================
 @app.post("/crop-doctor")
-async def crop_doctor(symptoms: str):
-    prompt = crop_template.format(symptoms=symptoms)
+async def crop_doctor(symptoms: str, authorization: str | None = Header(None)):
+    check_auth(authorization)
     try:
-        response = crop_llm.invoke([HumanMessage(content=prompt)])  # ✅ FIXED
+        response = crop_chain.invoke({"symptoms": symptoms})
         return {"diagnosis": str(response)}
     except HfHubHTTPError as e:
         return {"error": f"HuggingFace error: {str(e)}"}
 
 @app.post("/multilingual-chat")
-async def multilingual_chat(query: str):
-    prompt = chat_template.format(query=query)
+async def multilingual_chat(query: str, authorization: str | None = Header(None)):
+    check_auth(authorization)
     try:
-        response = chat_llm.invoke([HumanMessage(content=prompt)])  # ✅ FIXED
+        response = chat_chain.invoke({"query": query})
         return {"reply": str(response)}
     except HfHubHTTPError as e:
         return {"error": f"HuggingFace error: {str(e)}"}
 
 @app.post("/disaster-summarizer")
-async def disaster_summarizer(report: str):
-    prompt = disaster_template.format(report=report)
+async def disaster_summarizer(report: str, authorization: str | None = Header(None)):
+    check_auth(authorization)
     try:
-        response = disaster_llm.invoke([HumanMessage(content=prompt)])  # ✅ FIXED
+        response = disaster_chain.invoke({"report": report})
         return {"summary": str(response)}
     except HfHubHTTPError as e:
         return {"error": f"HuggingFace error: {str(e)}"}
 
 @app.post("/marketplace")
-async def marketplace(product: str):
-    prompt = market_template.format(product=product)
+async def marketplace(product: str, authorization: str | None = Header(None)):
+    check_auth(authorization)
     try:
-        response = market_llm.invoke([HumanMessage(content=prompt)])  # ✅ FIXED
+        response = market_chain.invoke({"product": product})
         return {"recommendation": str(response)}
     except HfHubHTTPError as e:
         return {"error": f"HuggingFace error: {str(e)}"}
 
 @app.post("/vector-search")
-async def vector_search(query: str):
+async def vector_search(query: str, authorization: str | None = Header(None)):
+    check_auth(authorization)
     try:
         results = query_vector(query)
         return {"results": results}