Updated

app.py

@@ -5,165 +5,154 @@ from fastapi.responses import JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
 from langchain.prompts import PromptTemplate
-from langchain_huggingface import HuggingFaceEndpoint
-from huggingface_hub.utils import HfHubHTTPError
-from langchain.schema import HumanMessage
-from vector import query_vector
+from transformers import pipeline
 
 # ==============================
-# Setup Logging
+# Logging
 # ==============================
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger("AgriCopilot")
 
 # ==============================
-# App Init
-# ==============================
-app = FastAPI(title="AgriCopilot")
-
-@app.get("/")
-async def root():
-    return {"status": "AgriCopilot AI Backend is working perfectly"}
-
-# ==============================
-# AUTH CONFIG
+# Auth
 # ==============================
 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 token")
+        raise HTTPException(status_code=403, detail="Invalid bearer token")
 
 # ==============================
-# Global Exception Handler
+# FastAPI Init
 # ==============================
-@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)},
-    )
+app = FastAPI(title="AgriCopilot")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],  # ✅ change to frontend URL in prod
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
 
 # ==============================
-# Request Models
+# Models via Transformers Pipelines
 # ==============================
-class CropRequest(BaseModel):
-    symptoms: str
+logger.info("Loading models... this may take a while.")
 
-class ChatRequest(BaseModel):
-    query: str
+crop_pipeline = pipeline(
+    "text-generation",
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    torch_dtype="auto",
+    device_map="auto"
+)
 
-class DisasterRequest(BaseModel):
-    report: str
+chat_pipeline = pipeline(
+    "text-generation",
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    torch_dtype="auto",
+    device_map="auto"
+)
 
-class MarketRequest(BaseModel):
-    product: str
+disaster_pipeline = pipeline(
+    "text-generation",
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    torch_dtype="auto",
+    device_map="auto"
+)
 
-class VectorRequest(BaseModel):
-    query: str
+market_pipeline = pipeline(
+    "text-generation",
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    torch_dtype="auto",
+    device_map="auto"
+)
 
 # ==============================
-# MODELS PER ENDPOINT (Meta Models, Conversational)
+# Prompt Templates
 # ==============================
-
-# 1. Crop Doctor
 crop_template = PromptTemplate(
     input_variables=["symptoms"],
-    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
+    template=(
+        "You are AgriCopilot, a multilingual AI crop doctor. "
+        "Farmer reports: {symptoms}. Diagnose the issue and suggest treatments "
+        "in simple farmer-friendly language."
+    )
 )
 
-# 2. Multilingual Chat
 chat_template = PromptTemplate(
     input_variables=["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
+    template=(
+        "You are AgriCopilot, a supportive multilingual assistant for farmers. "
+        "Respond in the same language. Farmer says: {query}"
+    )
 )
 
-# 3. Disaster Summarizer
 disaster_template = PromptTemplate(
     input_variables=["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
+    template=(
+        "You are AgriCopilot, an AI disaster assistant. Summarize this report "
+        "into 3–5 clear steps farmers can follow. Report: {report}"
+    )
 )
 
-# 4. Marketplace Recommendation
 market_template = PromptTemplate(
     input_variables=["product"],
-    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
+    template=(
+        "You are AgriCopilot, an AI marketplace advisor. Farmer wants to sell or buy: {product}. "
+        "Suggest matches, advice, and safe practices."
+    )
 )
 
 # ==============================
-# ENDPOINTS
+# Request Models
 # ==============================
+class CropRequest(BaseModel):
+    symptoms: str
+
+class ChatRequest(BaseModel):
+    query: str
+
+class DisasterRequest(BaseModel):
+    report: str
+
+class MarketRequest(BaseModel):
+    product: str
+
+# ==============================
+# Routes
+# ==============================
+@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_llm.invoke([HumanMessage(content=prompt)])
-    return {"diagnosis": str(response)}
+    response = crop_pipeline(prompt, max_new_tokens=512, do_sample=True)
+    return {"diagnosis": response[0]["generated_text"]}
 
 @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_llm.invoke([HumanMessage(content=prompt)])
-    return {"reply": str(response)}
+    response = chat_pipeline(prompt, max_new_tokens=512, do_sample=True)
+    return {"reply": response[0]["generated_text"]}
 
 @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_llm.invoke([HumanMessage(content=prompt)])
-    return {"summary": str(response)}
+    response = disaster_pipeline(prompt, max_new_tokens=512, do_sample=True)
+    return {"summary": response[0]["generated_text"]}
 
 @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_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}
+    response = market_pipeline(prompt, max_new_tokens=512, do_sample=True)
+    return {"recommendation": response[0]["generated_text"]}

requirements.txt

@@ -9,4 +9,8 @@ langchain
 langchain-huggingface
 kagglehub
 pandas
-datasets
+datasets
+transformers
+accelerate
+torch
+sentencepiece