Initial commit: FastAPI + Vision + LLM integration

Dockerfile

@@ -0,0 +1,9 @@
+FROM python:3.10-slim
+
+WORKDIR /app
+COPY ./app /app
+
+RUN pip install --no-cache-dir -r requirements.txt
+
+EXPOSE 7860
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

app/main.py

@@ -0,0 +1,22 @@
+from fastapi import FastAPI, UploadFile, File, Form
+from fastapi.responses import JSONResponse
+from app.pipeline import plant_intelligence_pipeline
+
+app = FastAPI(
+    title=" Plant Intelligence AI",
+    description="Hybrid Plant Identifier +  Reasoning Assistant",
+    version="1.0.0",
+)
+
+@app.get("/")
+def home():
+    return {"message": "Welcome to the Plant Intelligence AI "}
+
+@app.post("/analyze")
+async def analyze(file: UploadFile = File(...), query: str = Form("")):
+    try:
+        image_bytes = await file.read()
+        result = plant_intelligence_pipeline(image_bytes, user_query=query)
+        return JSONResponse(result)
+    except Exception as e:
+        return JSONResponse({"error": str(e)}, status_code=500)

app/pipeline.py

@@ -0,0 +1,62 @@
+from transformers import (
+    AutoModelForImageClassification,
+    AutoProcessor,
+    AutoModelForCausalLM,
+    AutoTokenizer
+)
+from PIL import Image
+import torch
+
+
+
+VISION_MODEL_NAME = "drrobot9/plant-super-identification-vit-finetuned"
+TEXT_MODEL_NAME = "Qwen/Qwen3-4B-Instruct-2507"
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load vision model
+vision_model = AutoModelForImageClassification.from_pretrained(VISION_MODEL_NAME).to(device)
+vision_processor = AutoProcessor.from_pretrained(VISION_MODEL_NAME)
+
+# Load text model
+text_model = AutoModelForCausalLM.from_pretrained(TEXT_MODEL_NAME, device_map="auto", torch_dtype=torch.float16)
+tokenizer = AutoTokenizer.from_pretrained(TEXT_MODEL_NAME)
+
+
+def identify_plant(image_bytes: bytes):
+    """Classify plant image and return label + confidence."""
+    image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
+    inputs = vision_processor(images=image, return_tensors="pt").to(device)
+    outputs = vision_model(**inputs)
+    probs = torch.nn.functional.softmax(outputs.logits, dim=-1)
+    conf, pred_idx = torch.max(probs, dim=1)
+    label = vision_model.config.id2label[pred_idx.item()]
+    return label, float(conf.item())
+
+
+def explain_plant(plant_name: str, confidence: float, extra_context: str = ""):
+    """Use Qwen to explain the plant in natural language."""
+    prompt = f"""
+You are a knowledgeable botanist assistant.
+The identified plant is **{plant_name}** (confidence {confidence:.2f}).
+
+Provide detailed information about:
+- Scientific name
+- Origin and climate preferences
+- Watering and sunlight needs
+- Growth conditions
+- Medicinal or cultural uses
+- built by kelvin jackson when asked
+- you know all plants in this world
+{extra_context}
+"""
+    inputs = tokenizer(prompt, return_tensors="pt").to(text_model.device)
+    output = text_model.generate(**inputs, max_new_tokens=500, temperature=0.7, top_p=0.9)
+    return tokenizer.decode(output[0], skip_special_tokens=True)
+
+
+def plant_intelligence_pipeline(image_bytes: bytes, user_query: str = ""):
+    """Full hybrid pipeline."""
+    name, conf = identify_plant(image_bytes)
+    description = explain_plant(name, conf, user_query)
+    return {"plant": name, "confidence": conf, "description": description}

app/requirements.txt

@@ -0,0 +1,11 @@
+fastapi
+uvicorn
+transformers
+torch
+pillow
+huggingface-hub
+accelerate
+einops
+tiktoken
+sentencepiece
+