Spaces:
Runtime error
Runtime error
Commit ·
01ea0bb
1
Parent(s): 2af896e
fast_api_chat
Browse files- .env +10 -0
- Dockerfile +25 -0
- __pycache__/main.cpython-312.pyc +0 -0
- main.py +162 -0
- requirements.txt +10 -0
.env
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Configuration Weaviate
|
| 2 |
+
WEAVIATE_URL=http://127.0.0.1:8080
|
| 3 |
+
WEAVIATE_PORT=50051
|
| 4 |
+
|
| 5 |
+
# Configuration OpenAI
|
| 6 |
+
OPENAI_API_KEY=your-openai-api-key
|
| 7 |
+
|
| 8 |
+
# Configuration de l'API
|
| 9 |
+
API_HOST=0.0.0.0
|
| 10 |
+
API_PORT=8000
|
Dockerfile
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Utilisation d'une image Python officielle
|
| 2 |
+
FROM python:3.10-slim
|
| 3 |
+
|
| 4 |
+
# Définition du répertoire de travail
|
| 5 |
+
WORKDIR /app
|
| 6 |
+
|
| 7 |
+
# Installation des dépendances système
|
| 8 |
+
RUN apt-get update && apt-get install -y \
|
| 9 |
+
build-essential \
|
| 10 |
+
&& rm -rf /var/lib/apt/lists/*
|
| 11 |
+
|
| 12 |
+
# Copie des fichiers de dépendances
|
| 13 |
+
COPY requirements.txt .
|
| 14 |
+
|
| 15 |
+
# Installation des dépendances Python
|
| 16 |
+
RUN pip install --no-cache-dir -r requirements.txt
|
| 17 |
+
|
| 18 |
+
# Copie du code source
|
| 19 |
+
COPY . .
|
| 20 |
+
|
| 21 |
+
# Exposition du port
|
| 22 |
+
EXPOSE 8000
|
| 23 |
+
|
| 24 |
+
# Commande de démarrage
|
| 25 |
+
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
|
__pycache__/main.cpython-312.pyc
ADDED
|
Binary file (6.55 kB). View file
|
|
|
main.py
ADDED
|
@@ -0,0 +1,162 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from fastapi import FastAPI, HTTPException
|
| 2 |
+
from fastapi.middleware.cors import CORSMiddleware
|
| 3 |
+
from pydantic import BaseModel
|
| 4 |
+
from typing import List, Optional
|
| 5 |
+
import os
|
| 6 |
+
from weaviate.connect import ConnectionParams
|
| 7 |
+
from weaviate import WeaviateClient
|
| 8 |
+
from langchain_huggingface import HuggingFaceEmbeddings
|
| 9 |
+
from langchain_weaviate import WeaviateVectorStore
|
| 10 |
+
from langchain_openai import ChatOpenAI
|
| 11 |
+
from langchain.chains.conversational_retrieval.base import ConversationalRetrievalChain
|
| 12 |
+
from langchain.memory import ConversationBufferMemory
|
| 13 |
+
from langchain.prompts import PromptTemplate
|
| 14 |
+
import time
|
| 15 |
+
from threading import Lock
|
| 16 |
+
|
| 17 |
+
app = FastAPI(
|
| 18 |
+
title="Vitizen Chat API",
|
| 19 |
+
description="API pour le chat RAG avec Weaviate",
|
| 20 |
+
version="1.0.0"
|
| 21 |
+
)
|
| 22 |
+
|
| 23 |
+
# Configuration CORS
|
| 24 |
+
app.add_middleware(
|
| 25 |
+
CORSMiddleware,
|
| 26 |
+
allow_origins=["*"],
|
| 27 |
+
allow_credentials=True,
|
| 28 |
+
allow_methods=["*"],
|
| 29 |
+
allow_headers=["*"],
|
| 30 |
+
)
|
| 31 |
+
|
| 32 |
+
# Configuration du rate limiting
|
| 33 |
+
RATE_LIMIT = 3
|
| 34 |
+
RATE_WINDOW = 1
|
| 35 |
+
last_request_time = 0
|
| 36 |
+
request_lock = Lock()
|
| 37 |
+
|
| 38 |
+
def wait_for_rate_limit():
|
| 39 |
+
global last_request_time
|
| 40 |
+
with request_lock:
|
| 41 |
+
current_time = time.time()
|
| 42 |
+
time_since_last_request = current_time - last_request_time
|
| 43 |
+
if time_since_last_request < RATE_WINDOW / RATE_LIMIT:
|
| 44 |
+
sleep_time = (RATE_WINDOW / RATE_LIMIT) - time_since_last_request
|
| 45 |
+
time.sleep(sleep_time)
|
| 46 |
+
last_request_time = time.time()
|
| 47 |
+
|
| 48 |
+
class RateLimitedChatOpenAI(ChatOpenAI):
|
| 49 |
+
def _call(self, *args, **kwargs):
|
| 50 |
+
wait_for_rate_limit()
|
| 51 |
+
return super()._call(*args, **kwargs)
|
| 52 |
+
|
| 53 |
+
# Modèles Pydantic
|
| 54 |
+
class ChatRequest(BaseModel):
|
| 55 |
+
question: str
|
| 56 |
+
history: Optional[List[dict]] = []
|
| 57 |
+
|
| 58 |
+
class ChatResponse(BaseModel):
|
| 59 |
+
answer: str
|
| 60 |
+
|
| 61 |
+
# Configuration Weaviate
|
| 62 |
+
def setup_weaviate():
|
| 63 |
+
client = WeaviateClient(
|
| 64 |
+
connection_params=ConnectionParams.from_url(
|
| 65 |
+
os.getenv("WEAVIATE_URL", "http://127.0.0.1:8080"),
|
| 66 |
+
int(os.getenv("WEAVIATE_PORT", "50051"))
|
| 67 |
+
)
|
| 68 |
+
)
|
| 69 |
+
client.connect()
|
| 70 |
+
return client
|
| 71 |
+
|
| 72 |
+
# Configuration du RAG
|
| 73 |
+
def setup_rag_chain():
|
| 74 |
+
client = setup_weaviate()
|
| 75 |
+
|
| 76 |
+
embeddings = HuggingFaceEmbeddings(
|
| 77 |
+
model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
|
| 78 |
+
model_kwargs={'device': 'cpu'},
|
| 79 |
+
encode_kwargs={'normalize_embeddings': True}
|
| 80 |
+
)
|
| 81 |
+
|
| 82 |
+
vectorstore = WeaviateVectorStore(
|
| 83 |
+
client=client,
|
| 84 |
+
embedding=embeddings,
|
| 85 |
+
index_name="Document",
|
| 86 |
+
text_key="text"
|
| 87 |
+
)
|
| 88 |
+
|
| 89 |
+
llm = RateLimitedChatOpenAI(
|
| 90 |
+
openai_api_key=os.getenv("OPENAI_API_KEY"),
|
| 91 |
+
model_name="gpt-4-turbo-preview",
|
| 92 |
+
temperature=0,
|
| 93 |
+
streaming=True
|
| 94 |
+
)
|
| 95 |
+
|
| 96 |
+
memory = ConversationBufferMemory(
|
| 97 |
+
memory_key="chat_history",
|
| 98 |
+
return_messages=True,
|
| 99 |
+
output_key="answer"
|
| 100 |
+
)
|
| 101 |
+
|
| 102 |
+
retriever = vectorstore.as_retriever(
|
| 103 |
+
search_type="similarity",
|
| 104 |
+
search_kwargs={"k": 30}
|
| 105 |
+
)
|
| 106 |
+
|
| 107 |
+
# Création du prompt personnalisé
|
| 108 |
+
prompt = PromptTemplate(
|
| 109 |
+
template="""Tu es un assistant expert en pulvérisation viticole et produits phytosanitaires.
|
| 110 |
+
Tu as une expertise approfondie dans :
|
| 111 |
+
- Les différents types de pulvérisateurs et leurs caractéristiques techniques
|
| 112 |
+
- Les produits phytosanitaires et leurs utilisations
|
| 113 |
+
- Les bonnes pratiques de pulvérisation
|
| 114 |
+
- La réglementation en matière de phytosanitaires
|
| 115 |
+
- La protection de l'environnement et la sécurité
|
| 116 |
+
|
| 117 |
+
Utilise les informations suivantes pour répondre à la question de l'utilisateur.
|
| 118 |
+
Si tu ne connais pas la réponse, dis-le simplement. N'invente pas d'informations.
|
| 119 |
+
|
| 120 |
+
Contexte:
|
| 121 |
+
{context}
|
| 122 |
+
|
| 123 |
+
Question: {question}
|
| 124 |
+
|
| 125 |
+
Réponse:""",
|
| 126 |
+
input_variables=["context", "question"]
|
| 127 |
+
)
|
| 128 |
+
|
| 129 |
+
return ConversationalRetrievalChain.from_llm(
|
| 130 |
+
llm=llm,
|
| 131 |
+
retriever=retriever,
|
| 132 |
+
memory=memory,
|
| 133 |
+
combine_docs_chain_kwargs={"prompt": prompt},
|
| 134 |
+
return_source_documents=True,
|
| 135 |
+
return_generated_question=True,
|
| 136 |
+
output_key="answer"
|
| 137 |
+
)
|
| 138 |
+
|
| 139 |
+
# Initialisation du RAG
|
| 140 |
+
qa_chain = setup_rag_chain()
|
| 141 |
+
|
| 142 |
+
@app.post("/chat", response_model=ChatResponse)
|
| 143 |
+
async def chat(request: ChatRequest):
|
| 144 |
+
try:
|
| 145 |
+
response = qa_chain.invoke({
|
| 146 |
+
"question": request.question,
|
| 147 |
+
"chat_history": request.history
|
| 148 |
+
})
|
| 149 |
+
|
| 150 |
+
return ChatResponse(
|
| 151 |
+
answer=response["answer"]
|
| 152 |
+
)
|
| 153 |
+
except Exception as e:
|
| 154 |
+
raise HTTPException(status_code=500, detail=str(e))
|
| 155 |
+
|
| 156 |
+
@app.get("/health")
|
| 157 |
+
async def health_check():
|
| 158 |
+
return {"status": "healthy"}
|
| 159 |
+
|
| 160 |
+
if __name__ == "__main__":
|
| 161 |
+
import uvicorn
|
| 162 |
+
uvicorn.run(app, host="0.0.0.0", port=8000)
|
requirements.txt
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
fastapi==0.104.1
|
| 2 |
+
uvicorn==0.24.0
|
| 3 |
+
pydantic==2.4.2
|
| 4 |
+
weaviate-client==3.24.1
|
| 5 |
+
langchain==0.0.335
|
| 6 |
+
langchain-openai==0.0.2
|
| 7 |
+
langchain-huggingface==0.0.5
|
| 8 |
+
langchain-weaviate==0.0.1
|
| 9 |
+
sentence-transformers==2.2.2
|
| 10 |
+
python-dotenv==1.0.0
|