Update app.py
Browse files
app.py
CHANGED
|
@@ -1,16 +1,18 @@
|
|
| 1 |
import os
|
| 2 |
-
os.environ["TRANSFORMERS_CACHE"] = "./cache"
|
| 3 |
-
|
| 4 |
from fastapi import FastAPI
|
| 5 |
from pydantic import BaseModel
|
| 6 |
from typing import Dict
|
| 7 |
-
from
|
| 8 |
|
| 9 |
app = FastAPI()
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
|
| 12 |
-
#
|
| 13 |
-
|
| 14 |
|
| 15 |
class AgentInput(BaseModel):
|
| 16 |
prompt: str
|
|
@@ -18,6 +20,16 @@ class AgentInput(BaseModel):
|
|
| 18 |
|
| 19 |
@app.post("/analyze")
|
| 20 |
async def analyze(input: AgentInput):
|
| 21 |
-
full_prompt =
|
| 22 |
-
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
|
|
|
|
|
|
| 2 |
from fastapi import FastAPI
|
| 3 |
from pydantic import BaseModel
|
| 4 |
from typing import Dict
|
| 5 |
+
from huggingface_hub import InferenceClient
|
| 6 |
|
| 7 |
app = FastAPI()
|
| 8 |
|
| 9 |
+
# Ton token HF (défini dans Settings → Secrets)
|
| 10 |
+
HF_TOKEN = os.getenv("HF_API_TOKEN")
|
| 11 |
+
# Le modèle que tu veux utiliser en remote
|
| 12 |
+
HF_MODEL = "google/flan-t5-base"
|
| 13 |
|
| 14 |
+
# Client d’inférence Hugging Face
|
| 15 |
+
client = InferenceClient(token=HF_TOKEN)
|
| 16 |
|
| 17 |
class AgentInput(BaseModel):
|
| 18 |
prompt: str
|
|
|
|
| 20 |
|
| 21 |
@app.post("/analyze")
|
| 22 |
async def analyze(input: AgentInput):
|
| 23 |
+
full_prompt = (
|
| 24 |
+
f"Voici des données JSON :\n{input.data}\n\n"
|
| 25 |
+
f"Question : {input.prompt}\nRéponse :"
|
| 26 |
+
)
|
| 27 |
+
# Appel remote, pas de cache, pas de pipeline local
|
| 28 |
+
out = client.text_generation(
|
| 29 |
+
model=HF_MODEL,
|
| 30 |
+
inputs=full_prompt,
|
| 31 |
+
parameters={"max_new_tokens": 200}
|
| 32 |
+
)
|
| 33 |
+
# out est une liste de dicts
|
| 34 |
+
response = out[0].generated_text if hasattr(out[0], "generated_text") else out[0]["generated_text"]
|
| 35 |
+
return {"response": response}
|