app.py

import requests
from fastapi import FastAPI
from pydantic import BaseModel

app = FastAPI()

OLLAMA_API_BASE = "http://localhost:11434"

class EmbeddingRequest(BaseModel):
    input: str
    model: str = "dengcao/qwen3-embedding-0.6b:q8_0"
    instruction: str = "Represent this sentence for retrieval: "  # Zgodnie z rekomendacją z dokumentacji
    normalize: bool = True  # Zalecane dla Qwen3
    options: dict = None

@app.post("/api/embeddings")
def get_embedding(request: EmbeddingRequest):
    full_prompt = request.instruction + request.input

    payload = {
        "model": request.model,
        "prompt": full_prompt,
        "options": request.options or {}
    }

    try:
        response = requests.post(f"{OLLAMA_API_BASE}/api/embeddings", json=payload)
        data = response.json()

        embedding = data.get("embedding", [])
        if request.normalize and embedding:
            # Opcjonalna normalizacja (dokumentacja sugeruje jej użycie)
            from math import sqrt
            norm = sqrt(sum(x**2 for x in embedding))
            if norm > 0:
                embedding = [x / norm for x in embedding]

        return {
            "object": "list",
            "data": [
                {
                    "object": "embedding",
                    "embedding": embedding,
                    "index": 0
                }
            ],
            "model": request.model,
            "usage": {
                "prompt_tokens": data.get("prompt_eval_count", 0),
                "total_tokens": data.get("prompt_eval_count", 0)
            }
        }
    except Exception as e:
        return {
            "error": str(e),
            "message": "Failed to generate embedding. Check if Ollama is running."
        }

@app.get("/")
def home():
    return {
        "status": "running",
        "model": "dengcao/qwen3-embedding-0.6b:q8_0",
        "endpoint": "/api/embeddings",
        "method": "POST",
        "body": {
            "input": "Your text here",
            "instruction": "Optional task-specific instruction"
        },
        "info": "Qwen3-Embedding-0.6B supports instructions for better performance."
    }