Updated

app.py

@@ -5,154 +5,165 @@ from fastapi.responses import JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
 from langchain.prompts import PromptTemplate
-from transformers import pipeline
+from langchain_huggingface import HuggingFaceEndpoint
+from huggingface_hub.utils import HfHubHTTPError
+from langchain.schema import HumanMessage
+from vector import query_vector
 
 # ==============================
-# Logging
+# Setup Logging
 # ==============================
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger("AgriCopilot")
 
 # ==============================
-# Auth
+# App Init
+# ==============================
+app = FastAPI(title="AgriCopilot")
+
+@app.get("/")
+async def root():
+    return {"status": "AgriCopilot AI Backend is working perfectly"}
+
+# ==============================
+# AUTH CONFIG
 # ==============================
 PROJECT_API_KEY = os.getenv("PROJECT_API_KEY")
 
 def check_auth(authorization: str | None):
+    """Validate Bearer token against PROJECT_API_KEY"""
     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 bearer token")
+        raise HTTPException(status_code=403, detail="Invalid token")
 
 # ==============================
-# FastAPI Init
+# Global Exception Handler
 # ==============================
-app = FastAPI(title="AgriCopilot")
-
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],  # ✅ change to frontend URL in prod
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
+@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)},
+    )
 
 # ==============================
-# Models via Transformers Pipelines
+# Request Models
 # ==============================
-logger.info("Loading models... this may take a while.")
+class CropRequest(BaseModel):
+    symptoms: str
 
-crop_pipeline = pipeline(
-    "text-generation",
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
-    torch_dtype="auto",
-    device_map="auto"
-)
+class ChatRequest(BaseModel):
+    query: str
 
-chat_pipeline = pipeline(
-    "text-generation",
-    model="meta-llama/Llama-3.1-8B-Instruct",
-    torch_dtype="auto",
-    device_map="auto"
-)
+class DisasterRequest(BaseModel):
+    report: str
 
-disaster_pipeline = pipeline(
-    "text-generation",
-    model="meta-llama/Llama-3.1-8B-Instruct",
-    torch_dtype="auto",
-    device_map="auto"
-)
+class MarketRequest(BaseModel):
+    product: str
 
-market_pipeline = pipeline(
-    "text-generation",
-    model="meta-llama/Llama-3.1-8B-Instruct",
-    torch_dtype="auto",
-    device_map="auto"
-)
+class VectorRequest(BaseModel):
+    query: str
 
 # ==============================
-# Prompt Templates
+# MODELS PER ENDPOINT (Meta Models, Conversational)
 # ==============================
+
+# 1. Crop Doctor
 crop_template = PromptTemplate(
     input_variables=["symptoms"],
-    template=(
-        "You are AgriCopilot, a multilingual AI crop doctor. "
-        "Farmer reports: {symptoms}. Diagnose the issue and suggest treatments "
-        "in simple farmer-friendly language."
-    )
+    template="You are AgriCopilot, a multilingual AI assistant created to support farmers. 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",   # ✅ FIXED
+    temperature=0.3,
+    top_p=0.9,
+    do_sample=True,
+    repetition_penalty=1.1,
+    max_new_tokens=1024
 )
 
+# 2. Multilingual Chat
 chat_template = PromptTemplate(
     input_variables=["query"],
-    template=(
-        "You are AgriCopilot, a supportive multilingual assistant for farmers. "
-        "Respond in the same language. Farmer says: {query}"
-    )
+    template="You are AgriCopilot, a supportive multilingual AI guide built for farmers. Farmer says: {query}"
+)
+chat_llm = HuggingFaceEndpoint(
+    repo_id="meta-llama/Llama-3.1-8B-Instruct",
+    task="conversational",   # ✅ FIXED
+    temperature=0.3,
+    top_p=0.9,
+    do_sample=True,
+    repetition_penalty=1.1,
+    max_new_tokens=1024
 )
 
+# 3. Disaster Summarizer
 disaster_template = PromptTemplate(
     input_variables=["report"],
-    template=(
-        "You are AgriCopilot, an AI disaster assistant. Summarize this report "
-        "into 3–5 clear steps farmers can follow. Report: {report}"
-    )
+    template="You are AgriCopilot, an AI disaster-response assistant. Summarize in simple steps: {report}"
+)
+disaster_llm = HuggingFaceEndpoint(
+    repo_id="meta-llama/Llama-3.1-8B-Instruct",
+    task="conversational",   # ✅ FIXED
+    temperature=0.3,
+    top_p=0.9,
+    do_sample=True,
+    repetition_penalty=1.1,
+    max_new_tokens=1024
 )
 
+# 4. Marketplace Recommendation
 market_template = PromptTemplate(
     input_variables=["product"],
-    template=(
-        "You are AgriCopilot, an AI marketplace advisor. Farmer wants to sell or buy: {product}. "
-        "Suggest matches, advice, and safe practices."
-    )
+    template="You are AgriCopilot, an AI agricultural marketplace advisor. Farmer wants to sell or buy: {product}. Suggest best options and advice."
+)
+market_llm = HuggingFaceEndpoint(
+    repo_id="meta-llama/Llama-3.1-8B-Instruct",
+    task="conversational",   # ✅ FIXED
+    temperature=0.3,
+    top_p=0.9,
+    do_sample=True,
+    repetition_penalty=1.1,
+    max_new_tokens=1024
 )
 
 # ==============================
-# Request Models
-# ==============================
-class CropRequest(BaseModel):
-    symptoms: str
-
-class ChatRequest(BaseModel):
-    query: str
-
-class DisasterRequest(BaseModel):
-    report: str
-
-class MarketRequest(BaseModel):
-    product: str
-
-# ==============================
-# Routes
+# ENDPOINTS
 # ==============================
-@app.get("/")
-async def root():
-    return {"status": "AgriCopilot AI Backend running with Transformers"}
-
 @app.post("/crop-doctor")
 async def crop_doctor(req: CropRequest, authorization: str | None = Header(None)):
     check_auth(authorization)
     prompt = crop_template.format(symptoms=req.symptoms)
-    response = crop_pipeline(prompt, max_new_tokens=512, do_sample=True)
-    return {"diagnosis": response[0]["generated_text"]}
+    response = crop_llm.invoke([HumanMessage(content=prompt)])
+    return {"diagnosis": str(response)}
 
 @app.post("/multilingual-chat")
 async def multilingual_chat(req: ChatRequest, authorization: str | None = Header(None)):
     check_auth(authorization)
     prompt = chat_template.format(query=req.query)
-    response = chat_pipeline(prompt, max_new_tokens=512, do_sample=True)
-    return {"reply": response[0]["generated_text"]}
+    response = chat_llm.invoke([HumanMessage(content=prompt)])
+    return {"reply": str(response)}
 
 @app.post("/disaster-summarizer")
 async def disaster_summarizer(req: DisasterRequest, authorization: str | None = Header(None)):
     check_auth(authorization)
     prompt = disaster_template.format(report=req.report)
-    response = disaster_pipeline(prompt, max_new_tokens=512, do_sample=True)
-    return {"summary": response[0]["generated_text"]}
+    response = disaster_llm.invoke([HumanMessage(content=prompt)])
+    return {"summary": str(response)}
 
 @app.post("/marketplace")
 async def marketplace(req: MarketRequest, authorization: str | None = Header(None)):
     check_auth(authorization)
     prompt = market_template.format(product=req.product)
-    response = market_pipeline(prompt, max_new_tokens=512, do_sample=True)
-    return {"recommendation": response[0]["generated_text"]}
+    response = market_llm.invoke([HumanMessage(content=prompt)])
+    return {"recommendation": str(response)}
+
+@app.post("/vector-search")
+async def vector_search(req: VectorRequest, authorization: str | None = Header(None)):
+    check_auth(authorization)
+    results = query_vector(req.query)
+    return {"results": results}