Hidden token credentials

Author: adribaeza

README.md

@@ -100,6 +100,9 @@ Remove the Docker Image (if needed)
 docker rmi ghcr.io/adribaeza/llm-tinyllama-backend:latest
 
 
+
+
+
 Razones para enviar el historial completo:
 Contexto: El modelo necesita el historial completo para entender el contexto de la conversación y generar respuestas adecuadas.
 Coherencia: Mantener el historial ayuda a asegurar que las respuestas sean coherentes con las interacciones anteriores.
@@ -108,4 +111,10 @@ Memoria: Los modelos LLM no tienen memoria persistente entre solicitudes, por lo
 Cómo usar docker-compose
 Construir y ejecutar los servicios:
 Detener los servicios:
-Este archivo docker-compose.yml te permitirá levantar tanto el backend con FastAPI como el frontend con Streamlit de manera sencilla y reproducible.
+Este archivo docker-compose.yml te permitirá levantar tanto el backend con FastAPI como el frontend con Streamlit de manera sencilla y reproducible.
+
+kubectl create secret generic llm-service-token-secret --from-literal=SERVICE_TOKEN=myllservicetoken2024
+
+export SERVICE_TOKEN=myllservicetoken2024
+
+docker-compose up

backend/.env

@@ -0,0 +1 @@
+SERVICE_TOKEN=myllservicetoken2024

backend/Dockerfile

@@ -5,12 +5,15 @@ FROM python:3.10-slim
 WORKDIR /api
 
 # Copy the requirements file into the container at /api and install dependencies
-COPY ./requirements.txt /api/requirements.txt
+COPY requirements.txt /api/requirements.txt
 RUN pip install --no-cache-dir -r requirements.txt
 
 # Copy backend code into the container at /api
 COPY ./api /api
 
+# Copy .env file into the container at /api
+COPY .env /api
+
 # Expose the port the api runs in
 EXPOSE 8000
 

backend/api/main.py

@@ -9,12 +9,22 @@
 from fastapi import FastAPI, HTTPException, APIRouter, Depends
 from fastapi.security import OAuth2PasswordBearer
 from transformers import pipeline
-
 from starlette.middleware.cors import CORSMiddleware #Import the middleware
 import json
 from pydantic import BaseModel
 import asyncio
 import os
+from dotenv import load_dotenv
+
+# Load the environment variables from the .env file
+load_dotenv()
+
+# Static token for the API
+STATIC_TOKEN = os.getenv("SERVICE_TOKEN")
+
+# Verify that the SERVICE_TOKEN is defined in the environment variables
+if STATIC_TOKEN is None:
+    raise ValueError("The SERVICE_TOKEN environment variable is not defined")
 
 #Default LLM configuration values
 DEFAULT_MAX_NEW_TOKENS = 100
@@ -24,8 +34,8 @@
 DEFAULT_TOP_P = 0.6
 LLM_MODEL = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
 
-#instance logger
-logger = logging.getLogger(__name__)
+# Set up logging
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
 
 #Set default route for the API with prefix /api/v1
 api_router = APIRouter(prefix="/api/v1")
@@ -38,9 +48,6 @@
 # OAuth2 configuration
 oauth2_scheme = OAuth2PasswordBearer(tokenUrl="token")
 
-# Static token for the API
-STATIC_TOKEN = "myllservicetoken2024" #os.getenv("STATIC_TOKEN")
-
 def verify_token(token: str):
     if token != STATIC_TOKEN:
         raise HTTPException(
@@ -62,7 +69,7 @@ async def get_current_user(token: str = Depends(oauth2_scheme)):
     allow_headers=["*"],
 )
 
-logger.info('Adding v1 endpoints..')
+logging.info('Adding v1 endpoints..')
 
 # Load the model with the TinyLlama model
 pipe = pipeline("text-generation", model=LLM_MODEL, torch_dtype=torch.bfloat16, device_map="auto")

backend/k8s/deployment.yaml

@@ -16,4 +16,10 @@ spec:
       - name: chatllmapi
         image: ghcr.io/adribaeza/llm-tinyllama-backend:latest
         ports:
-        - containerPort: 8000
+        - containerPort: 8000
+        env:
+        - name: SERVICE_TOKEN
+          valueFrom:
+            secretKeyRef:
+              name: llm-service-token-secret
+              key: SERVICE_TOKEN

backend/requirements.txt

@@ -6,4 +6,5 @@ pydantic==2.8.2
 huggingface_hub==0.24.2
 accelerate==0.33.0
 pytest==8.3.2
-pytest-cov==5.0.0
+pytest-cov==5.0.0
+python-dotenv==1.0.1

dicker-compose-ghimages.yaml

@@ -8,6 +8,7 @@ services:
     environment:
       - MODULE_NAME=api.main
       - VARIABLE_NAME=api
+      - SERVICE_TOKEN=${SERVICE_TOKEN}
 
   frontend:
     image: ghcr.io/adribaeza/llm-tinyllama-frontend:latest
@@ -16,4 +17,9 @@ services:
     ports:
       - "8501:8501"
     depends_on:
-      - backend
+      - backend
+    environment:
+      - SERVICE_TOKEN=${SERVICE_TOKEN}
+
+env_file:
+  - .env

docker-compose.yml

@@ -9,6 +9,9 @@ services:
     environment:
       - MODULE_NAME=api.main
       - VARIABLE_NAME=api
+      - SERVICE_TOKEN=${SERVICE_TOKEN}
+    env_file:
+      - ./frontend/.env
 
   frontend:
     build:
@@ -18,4 +21,8 @@ services:
     ports:
       - "8501:8501"
     depends_on:
-      - backend
+      - backend
+    environment:
+      - SERVICE_TOKEN=${SERVICE_TOKEN}
+    env_file:
+      - ./backend/.env

frontend/.env

@@ -0,0 +1 @@
+SERVICE_TOKEN=myllservicetoken2024

frontend/Dockerfile

@@ -11,6 +11,9 @@ COPY . /app
 COPY requirements.txt /app/
 RUN pip install --no-cache-dir -r requirements.txt
 
+# Copy .env file into the container at /api
+COPY .env /api
+
 # Make port 8501 available to the world outside this container
 EXPOSE 8501
 

frontend/app/main.py

@@ -1,5 +1,14 @@
 import streamlit as st
 import requests, logging, os
+from dotenv import load_dotenv
+
+# Load the environment variables from the .env file
+load_dotenv()
+# Static token for the API
+STATIC_TOKEN = os.getenv("SERVICE_TOKEN")
+# Verify that the SERVICE_TOKEN is defined in the environment variables
+if STATIC_TOKEN is None:
+    raise ValueError("The SERVICE_TOKEN environment variable is not defined")
 
 # Set up logging
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
@@ -11,8 +20,6 @@
 DEFAULT_TOP_K = 50
 DEFAULT_TOP_P = 0.9
 
-DEFAULT_TOKEN = "myllservicetoken2024" #os.getenv("TINYLLAMA_API_TOKEN")
-
 
 # Función para limpiar el historial de mensajes
 def clear_chat():
@@ -83,7 +90,7 @@ def main():
             st.markdown(prompt)
 
         headers = {
-            "Authorization": f"Bearer {DEFAULT_TOKEN}",
+            "Authorization": f"Bearer {STATIC_TOKEN}",
             "Content-Type": "application/json"
         }
         # Construir el historial de la conversación

frontend/k8s/deployment.yaml

@@ -16,4 +16,10 @@ spec:
       - name: chatllm-frontend
         image: ghcr.io/adribaeza/llm-tinyllama-frontend:latest
         ports:
-        - containerPort: 8501
+        - containerPort: 8501
+        env:
+        - name: SERVICE_TOKEN
+          valueFrom:
+            secretKeyRef:
+              name: llm-service-token-secret
+              key: SERVICE_TOKEN

frontend/requirements.txt

@@ -1,3 +1,4 @@
 streamlit==1.37.0
 pytest==8.3.2
-pytest-cov==5.0.0
+pytest-cov==5.0.0
+python-dotenv==1.0.1