app.py

from fastapi import FastAPI
from pydantic import BaseModel
from typing import Optional

from llama_index.core import Document, ServiceContext
from llama_index.llms.openai import OpenAI
from llama_index.core.node_parser import SemanticSplitterNodeParser
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
import os

app = FastAPI()

# 🔹 Schéma d'entrée
class ChunkRequest(BaseModel):
    text: str
    source_id: Optional[str] = None
    titre: Optional[str] = None
    source: Optional[str] = None
    type: Optional[str] = None

# 🔹 Endpoint principal
@app.post("/chunk")
async def chunk_text(data: ChunkRequest):
    # Modèle LLM (OpenRouter - Llama 4 Maverick)
    llm = OpenAI(
        model="meta-llama/llama-4-maverick:free",
        api_base="https://openrouter.ai/api/v1",
        api_key=os.getenv("OPENROUTER_API_KEY")
    )

    # 🔹 Embedding open source gratuit
    embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

    # 🔹 Service Context avec LLM + embeddings
    service_context = ServiceContext.from_defaults(
        llm=llm,
        embed_model=embed_model
    )

    try:
        parser = SemanticSplitterNodeParser.from_defaults(service_context=service_context)
        nodes = parser.get_nodes_from_documents([Document(text=data.text)])

        return {
            "chunks": [node.text for node in nodes],
            "metadatas": [node.metadata for node in nodes],
            "source_id": data.source_id,
            "titre": data.titre,
            "source": data.source,
            "type": data.type
        }
    except Exception as e:
        return {"error": str(e)}