Create app.py

app.py

@@ -0,0 +1,150 @@
+import os
+import base64
+import torch
+import numpy as np
+import faiss
+import json
+
+from fastapi import FastAPI
+from pydantic import BaseModel
+from contextlib import asynccontextmanager
+from huggingface_hub import snapshot_download
+from sentence_transformers import SentenceTransformer
+from PIL import Image
+from io import BytesIO
+
+from transformers import AutoProcessor, AutoModelForVision2Seq
+
+# ─────────────────────────────
+# CONFIG
+# ─────────────────────────────
+MODEL_REPO = "Rady10/Plant-Disease-Qwen3VL-2B"
+RAG_REPO = "Rady10/Agriculture-Rag-Data-Index"
+
+DEVICE = "cpu"
+
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+# ─────────────────────────────
+# GLOBALS
+# ─────────────────────────────
+model = None
+processor = None
+
+faiss_index = None
+rag_chunks = None
+embedder = None
+
+# ─────────────────────────────
+# FASTAPI APP
+# ─────────────────────────────
+app = FastAPI(title="🌿 Plant Disease Vision API")
+
+# ─────────────────────────────
+# LOAD MODELS ONCE
+# ─────────────────────────────
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+
+    global model, processor, faiss_index, rag_chunks, embedder
+
+    print("Loading vision model...")
+
+    processor = AutoProcessor.from_pretrained(
+        MODEL_REPO,
+        trust_remote_code=True
+    )
+
+    model = AutoModelForVision2Seq.from_pretrained(
+        MODEL_REPO,
+        torch_dtype=torch.float32,
+        device_map="cpu",
+        trust_remote_code=True
+    )
+
+    model.eval()
+
+    # ───── RAG (optional but included) ─────
+    print("Loading RAG...")
+
+    rag_dir = snapshot_download(
+        repo_id=RAG_REPO,
+        repo_type="dataset",
+        local_dir="./rag"
+    )
+
+    faiss_index = faiss.read_index(
+        os.path.join(rag_dir, "agro.index")
+    )
+
+    with open(os.path.join(rag_dir, "chunks.json"), "r", encoding="utf-8") as f:
+        rag_chunks = json.load(f)
+
+    embedder = SentenceTransformer(
+        "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
+    )
+
+    print("ALL LOADED")
+
+    yield
+
+app = FastAPI(lifespan=lifespan)
+
+# ─────────────────────────────
+# REQUEST MODEL
+# ─────────────────────────────
+class VisionRequest(BaseModel):
+    image: str   # base64
+    text: str = ""
+
+# ─────────────────────────────
+# IMAGE DECODER
+# ─────────────────────────────
+def decode_image(base64_str):
+    img_data = base64.b64decode(base64_str)
+    return Image.open(BytesIO(img_data)).convert("RGB")
+
+# ─────────────────────────────
+# GENERATION
+# ─────────────────────────────
+def generate(image, text):
+
+    if text.strip() == "":
+        text = "What disease is shown in this plant image?"
+
+    inputs = processor(
+        text=text,
+        images=image,
+        return_tensors="pt"
+    )
+
+    inputs = {k: v.to(model.device) for k, v in inputs.items()}
+
+    with torch.no_grad():
+        output = model.generate(
+            **inputs,
+            max_new_tokens=256,
+            temperature=0.7,
+            top_p=0.9
+        )
+
+    return processor.batch_decode(
+        output,
+        skip_special_tokens=True
+    )[0]
+
+# ─────────────────────────────
+# API ROUTES
+# ─────────────────────────────
+@app.get("/")
+def root():
+    return {"status": "vision api running"}
+
+@app.post("/analyze")
+def analyze(req: VisionRequest):
+
+    image = decode_image(req.image)
+
+    result = generate(image, req.text)
+
+    return {"response": result}