jamesf/Perplexica
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Backend GKE Deploy, access key for backend

Browse source 
- Configs and automation for deploying backend to GKE
- First steps to adding an optional token check for requests to backend
- First steps frontend sending optional token to backend when configured
This commit is contained in:
Hristo 2024-05-10 16:07:58 -04:00
parent 0fedaef537
commit e6c2042df6
17 changed files with 296 additions and 39 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
Makefile Normal file
View file

@ -0,0 +1,18 @@
.PHONY: run
run:
docker compose -f docker-compose.yaml up
.PHONY: rebuild-run
rebuild-run:
docker compose -f docker-compose.yaml up --build
.PHONY: run-app-only
run-app-only:
docker compose -f app-docker-compose.yaml up
.PHONY: rebuild-run-app-only
rebuild-run-app-only:
docker compose -f app-docker-compose.yaml up --build

20
README.md
View file

@ -106,6 +106,26 @@ You need to edit the ports accordingly.
[![Deploy to RepoCloud](https://d16t0pc4846x52.cloudfront.net/deploylobe.svg)](https://repocloud.io/details/?app_id=267)
## Deploy Perplexica backend to Google GKE
0: Install `docker` and `terraform` (Process specific to your system)
1a: Copy the `sample.env` file to `.env`
1b: Copy the `deploy/gcp/sample.env` file to `deploy/gcp/.env`
2a: Fillout desired LLM provider access keys etc. in `.env`
- Note: you will have to comeback and edit this file again once you have the address of the K8s backend deploy
2b: Fillout the GCP info in `deploy/gcp/.env`
3: Edit `GCP_REPO` to the correct docker image repo path if you are using something other than Container registry
4: Edit the `PREFIX` if you would like images and GKE entities to be prefixed with something else
5: In `deploy/gcp` run `make init` to initialize terraform
6: Follow the normal Preplexica configuration steps outlined in the project readme
7: Auth docker with the appropriate credential for repo Ex. for `gcr.io` -> `gcloud auth configure-docker`
8: In `deploy/gcp` run `make build-deplpy` to build and push the project images to the repo, create a GKE cluster and deploy the app
9: Once deployed successfully edit the `.env` file in the root project folder and update the `REMOTE_BACKEND_ADDRESS` with the remote k8s deployment address and port
10: In root project folder run `make rebuild-run-app-only`
If you configured everything correctly frontend app will run locally and provide you with a local url to open it.
Now you can run queries against the remotely deployed backend from your local machine. :celebrate:
## Upcoming Features
- [ ] Finalizing Copilot Mode

13
app-docker-compose.yaml Normal file
View file

@ -0,0 +1,13 @@
services:
perplexica-frontend:
build:
context: .
dockerfile: app.dockerfile
args:
- SUPER_SECRET_KEY=${SUPER_SECRET_KEY}
- NEXT_PUBLIC_API_URL=http://${REMOTE_BACKEND_ADDRESS}/api
- NEXT_PUBLIC_WS_URL=ws://${REMOTE_BACKEND_ADDRESS}
expose:
- 3000
ports:
- 3000:3000

3
app.dockerfile
View file

@ -2,8 +2,11 @@ FROM node:alpine
ARG NEXT_PUBLIC_WS_URL
ARG NEXT_PUBLIC_API_URL
ARG SUPER_SECRET_KEY
ENV NEXT_PUBLIC_WS_URL=${NEXT_PUBLIC_WS_URL}
ENV NEXT_PUBLIC_API_URL=${NEXT_PUBLIC_API_URL}
ENV SUPER_SECRET_KEY=${SUPER_SECRET_KEY}
WORKDIR /home/perplexica

35
deploy/gcp/Makefile
View file

@ -1,21 +1,14 @@
# USAGE:
# 0: Install `docker` and `terraform` (Process specific to your system)
# 1: Copy the sample.env file to .env
# 2: Fillout the GCP info in .env
# 3: Edit GCP_REPO to the correct docker image repo path if you are using something other than Container registry
# 4: Edit the PREFIX if you would like images and GKE entities to be prefixed with something else
# 5: Run `make init` to initialize terraform
# 6: Follow the normal Preplexica configuration steps outlined in the project readme
# 7: Run `make build-deplpy` to build and push the project images to the repo, create a GKE cluster and deploy the app
#
# NOTES/ WARNINGS:
# - The app endpoint is unsecured and exposed to the internet at large, you will need to implement your desired auth
# - No auto scaling is enabled for this cluster and deployments, edit the terraform files accordingly for that
# Adds all the deployment relevant sensitive information about project
include .env
# Adds secrets/ keys we have define for the project locally and deployment
include ../../.env
# Use `location-id-docker.pkg` for artifact registry Ex. west-1-docker.pkg
GCP_REPO=gcr.io
PREFIX=perplexica
SEARCH_PORT=8080
BACKEND_PORT=3001
SEARCH_IMAGE_TAG=$(GCP_REPO)/$(GCP_PROJECT_ID)/$(PREFIX)-searxng:latest
BACKEND_IMAGE_TAG=$(GCP_REPO)/$(GCP_PROJECT_ID)/$(PREFIX)-backend:latest
APP_IMAGE_TAG=$(GCP_REPO)/$(GCP_PROJECT_ID)/$(PREFIX)-app:latest
@ -38,8 +31,10 @@ show_config:
&& echo $(GCP_SERVICE_ACCOUNT_KEY_FILE) \
&& echo $(SEARCH_IMAGE_TAG) \
&& echo $(BACKEND_IMAGE_TAG) \
&& echo $(APP_IMAGE_TAG)
&& echo $(APP_IMAGE_TAG) \
&& echo $(SEARCH_PORT) \
&& echo $(BACKEND_PORT) \
&& echo $(OPENAI)
.PHONY: docker-build-push-searxng
docker-build-push-searxng:
@ -49,14 +44,15 @@ docker-build-push-searxng:
.PHONY: docker-build-push-backend
docker-build-push-backend:
cd ../../ && docker build -f ./backed.dockerfile -t $(BACKEND_IMAGE_TAG) . --platform="linux/amd64"
cd ../../ && docker build -f ./backend.dockerfile -t $(BACKEND_IMAGE_TAG) . --platform="linux/amd64"
docker push $(BACKEND_IMAGE_TAG)
.PHONY: docker-build-push-app
docker-build-push-app:
cd ../../ && docker build -f ./app.dockerfile -t $(APP_IMAGE_TAG) . --platform="linux/amd64"
docker push $(APP_IMAGE_TAG)
#
# cd ../../ && docker build -f ./app.dockerfile -t $(APP_IMAGE_TAG) . --platform="linux/amd64"
# docker push $(APP_IMAGE_TAG)
.PHONY: init
@ -73,6 +69,9 @@ deploy:
&& export TF_VAR_search_image=$(SEARCH_IMAGE_TAG) \
&& export TF_VAR_backend_image=$(BACKEND_IMAGE_TAG) \
&& export TF_VAR_app_image=$(APP_IMAGE_TAG) \
&& export TF_VAR_search_port=$(SEARCH_PORT) \
&& export TF_VAR_backend_port=$(BACKEND_PORT) \
&& export TF_VAR_open_ai=$(OPENAI) \
&& terraform apply

100
deploy/gcp/main.tf
View file

@ -35,6 +35,9 @@ provider "kubernetes" {
)
}
#####################################################################################################
# SearXNG - Search engine deployment and service
#####################################################################################################
resource "kubernetes_deployment" "searxng" {
metadata {
name = "searxng"
@ -60,7 +63,7 @@ resource "kubernetes_deployment" "searxng" {
image = var.search_image
name = "searxng-container"
port {
container_port = 8080
container_port = var.search_port
}
}
}
@ -80,14 +83,88 @@ resource "kubernetes_service" "searxng_service" {
}
port {
port = 8080
target_port = 8080
port = var.search_port
target_port = var.search_port
}
type = "LoadBalancer"
}
}
#####################################################################################################
# Perplexica - backend deployment and service
#####################################################################################################
resource "kubernetes_deployment" "backend" {
metadata {
name = "backend"
labels = {
app = "backend"
}
}
spec {
replicas = 1
selector {
match_labels = {
component = "backend"
}
}
template {
metadata {
labels = {
component = "backend"
}
}
spec {
container {
image = var.backend_image
name = "backend-container"
port {
container_port = var.backend_port
}
env {
# searxng service ip
name = "SEARXNG_API_URL"
value = "http://${kubernetes_service.searxng_service.status[0].load_balancer[0].ingress[0].ip}:${var.search_port}"
}
env {
# openai key
name = "OPENAI"
value = var.open_ai
}
env {
# port
name = "PORT"
value = var.backend_port
}
}
}
}
}
}
resource "kubernetes_service" "backend_service" {
metadata {
name = "backend-service"
namespace = "default"
}
spec {
selector = {
component = "backend"
}
port {
port = var.backend_port
target_port = var.backend_port
}
type = "LoadBalancer"
}
}
#####################################################################################################
# Variable and module definitions
#####################################################################################################
variable "project_id" {
description = "The ID of the project in which the resources will be deployed."
type = string
@ -113,7 +190,7 @@ variable "search_image" {
type = string
}
variable "backed_image" {
variable "backend_image" {
description = "Tag for the Perplexica backend image"
type = string
}
@ -123,6 +200,21 @@ variable "app_image" {
type = string
}
variable "open_ai" {
description = "OPENAI access key"
type = string
}
variable "search_port" {
description = "Port for searxng service"
type = number
}
variable "backend_port" {
description = "Port for backend service"
type = number
}
module "gke-cluster" {
source = "./gke-cluster"

12
docker-compose.yaml
View file

@ -15,7 +15,12 @@ services:
context: .
dockerfile: backend.dockerfile
args:
- SEARXNG_API_URL=http://searxng:8080
- SEARXNG_API_URL=null
environment:
SEARXNG_API_URL: "http://searxng:8080"
OPENAI: ${OPENAI}
GROQ: ${GROQ}
OLLAMA_API_URL: ${OLLAMA_API_URL}
depends_on:
- searxng
expose:
@ -30,8 +35,9 @@ services:
context: .
dockerfile: app.dockerfile
args:
- NEXT_PUBLIC_API_URL=http://127.0.0.1:3001/api
- NEXT_PUBLIC_WS_URL=ws://127.0.0.1:3001
- SUPER_SECRET_KEY=${SUPER_SECRET_KEY}
- NEXT_PUBLIC_API_URL=http://${REMOTE_BACKEND_ADDRESS}/api
- NEXT_PUBLIC_WS_URL=ws://${REMOTE_BACKEND_ADDRESS}
depends_on:
- perplexica-backend
expose:

20
sample.env Normal file
View file

@ -0,0 +1,20 @@
# Copy this file over to .env and fill in the desired config.
# .env will become available to docker compose and these values will be
# used when running docker compose up
# Edit to set OpenAI access key
OPENAI=ADD OPENAI KEY HERE
# Uncomment and edit to set GROQ access key
# GROQ: ${GROQ}
# Uncomment and edit to set OLLAMA Url
# OLLAMA_API_URL: ${OLLAMA_API_URL}
# Address and port of the remotely deployed Perplexica backend
REMOTE_BACKEND_ADDRESS=111.111.111.111:0000
# Uncomment and edit to configure backend to reject requests without token
# leave commented to have open access to all endpoints
# Secret key to "secure" backend
# SUPER_SECRET_KEY=THISISASUPERSECRETKEYSERIOUSLY

7
src/app.ts
View file

@ -3,7 +3,8 @@ import express from 'express';
import cors from 'cors';
import http from 'http';
import routes from './routes';
import { getPort } from './config';
import { requireAccessKey } from './auth';
import { getAccessKey, getPort } from './config';
import logger from './utils/logger';
const port = getPort();
@ -23,6 +24,10 @@ app.get('/api', (_, res) => {
res.status(200).json({ status: 'ok' });
});
if (getAccessKey()) {
app.all('*', requireAccessKey);
};
server.listen(port, () => {
logger.info(`Server is running on port ${port}`);
});

18
src/auth.ts Normal file
View file

@ -0,0 +1,18 @@
import {
getAccessKey,
} from '../config';
const requireAccessKey = (req, res, next) => {
const authHeader = req.headers.authorization;
if (authHeader) {
const token = authHeader.split(' ')[1];
if (token !== getAccessKey()) {
return res.sendStatus(403);
}
next();
} else {
res.sendStatus(401);
}
};

29
src/config.ts
View file

@ -8,6 +8,7 @@ interface Config {
GENERAL: {
PORT: number;
SIMILARITY_MEASURE: string;
SUPER_SECRET_KEY: string;
};
API_KEYS: {
OPENAI: string;
@ -28,18 +29,38 @@ const loadConfig = () =>
fs.readFileSync(path.join(__dirname, `../${configFileName}`), 'utf-8'),
) as any as Config;
const loadEnv = () => {
return {
GENERAL: {
PORT: Number(process.env.PORT),
SIMILARITY_MEASURE: process.env.SIMILARITY_MEASURE,
SUPER_SECRET_KEY: process.env.SUPER_SECRET_KEY
},
API_KEYS: {
OPENAI: process.env.OPENAI,
GROQ: process.env.GROQ
},
API_ENDPOINTS: {
SEARXNG: process.env.SEARXNG_API_URL,
OLLAMA: process.env.OLLAMA_API_URL
}
} as Config;
};
export const getPort = () => loadConfig().GENERAL.PORT;
export const getAccessKey = () => loadEnv().GENERAL.SUPER_SECRET_KEY || loadConfig().GENERAL.SUPER_SECRET_KEY;
export const getSimilarityMeasure = () =>
loadConfig().GENERAL.SIMILARITY_MEASURE;
export const getOpenaiApiKey = () => loadConfig().API_KEYS.OPENAI;
export const getOpenaiApiKey = () => loadEnv().API_KEYS.OPENAI || loadConfig().API_KEYS.OPENAI;
export const getGroqApiKey = () => loadConfig().API_KEYS.GROQ;
export const getGroqApiKey = () => loadEnv().API_KEYS.GROQ || loadConfig().API_KEYS.GROQ;
export const getSearxngApiEndpoint = () => loadConfig().API_ENDPOINTS.SEARXNG;
export const getSearxngApiEndpoint = () => loadEnv().API_ENDPOINTS.SEARXNG || loadConfig().API_ENDPOINTS.SEARXNG;
export const getOllamaApiEndpoint = () => loadConfig().API_ENDPOINTS.OLLAMA;
export const getOllamaApiEndpoint = () => loadEnv().API_ENDPOINTS.OLLAMA || loadConfig().API_ENDPOINTS.OLLAMA;
export const updateConfig = (config: RecursivePartial<Config>) => {
const currentConfig = loadConfig();

5
ui/components/ChatWindow.tsx
View file

@ -6,6 +6,7 @@ import Navbar from './Navbar';
import Chat from './Chat';
import EmptyChat from './EmptyChat';
import { toast } from 'sonner';
import { clientFetch } from '@/lib/utils';
export type Message = {
id: string;
@ -34,8 +35,8 @@ const useSocket = (url: string) => {
!embeddingModel ||
!embeddingModelProvider
) {
const providers = await fetch(
`${process.env.NEXT_PUBLIC_API_URL}/models`,
const providers = await clientFetch(
'/models',
{
headers: {
'Content-Type': 'application/json',

5
ui/components/SearchImages.tsx
View file

@ -4,6 +4,7 @@ import { useState } from 'react';
import Lightbox from 'yet-another-react-lightbox';
import 'yet-another-react-lightbox/styles.css';
import { Message } from './ChatWindow';
import { clientFetch } from '@/lib/utils';
type Image = {
url: string;
@ -33,8 +34,8 @@ const SearchImages = ({
const chatModelProvider = localStorage.getItem('chatModelProvider');
const chatModel = localStorage.getItem('chatModel');
const res = await fetch(
`${process.env.NEXT_PUBLIC_API_URL}/images`,
const res = await clientFetch(
'/images',
{
method: 'POST',
headers: {

5
ui/components/SearchVideos.tsx
View file

@ -4,6 +4,7 @@ import { useState } from 'react';
import Lightbox, { GenericSlide, VideoSlide } from 'yet-another-react-lightbox';
import 'yet-another-react-lightbox/styles.css';
import { Message } from './ChatWindow';
import { clientFetch } from '@/lib/utils';
type Video = {
url: string;
@ -46,8 +47,8 @@ const Searchvideos = ({
const chatModelProvider = localStorage.getItem('chatModelProvider');
const chatModel = localStorage.getItem('chatModel');
const res = await fetch(
`${process.env.NEXT_PUBLIC_API_URL}/videos`,
const res = await clientFetch(
'/videos',
{
method: 'POST',
headers: {

5
ui/components/SettingsDialog.tsx
View file

@ -1,6 +1,7 @@
import { Dialog, Transition } from '@headlessui/react';
import { CloudUpload, RefreshCcw, RefreshCw } from 'lucide-react';
import React, { Fragment, useEffect, useState } from 'react';
import { clientFetch } from '@/lib/utils';
interface SettingsType {
chatModelProviders: {
@ -42,7 +43,7 @@ const SettingsDialog = ({
if (isOpen) {
const fetchConfig = async () => {
setIsLoading(true);
const res = await fetch(`${process.env.NEXT_PUBLIC_API_URL}/config`, {
const res = await clientFetch('/config', {
headers: {
'Content-Type': 'application/json',
},
@ -102,7 +103,7 @@ const SettingsDialog = ({
setIsUpdating(true);
try {
await fetch(`${process.env.NEXT_PUBLIC_API_URL}/config`, {
await clientFetch('/config', {
method: 'POST',
headers: {
'Content-Type': 'application/json',

20
ui/lib/config.ts Normal file
View file

@ -0,0 +1,20 @@
interface Config {
GENERAL: {
SUPER_SECRET_KEY: string;
NEXT_PUBLIC_API_URL: string;
};
}
const loadEnv = () => {
return {
GENERAL: {
SUPER_SECRET_KEY: process.env.SUPER_SECRET_KEY!,
NEXT_PUBLIC_API_URL: process.env.NEXT_PUBLIC_API_URL!,
NEXT_PUBLIC_WS_URL: process.env.NEXT_PUBLIC_WS_URL!
},
} as Config;
};
export const getAccessKey = () => loadEnv().GENERAL.SUPER_SECRET_KEY;
export const getBackendURL = () => loadEnv().GENERAL.NEXT_PUBLIC_API_URL;

18
ui/lib/utils.ts
View file

@ -1,5 +1,6 @@
import clsx, { ClassValue } from 'clsx';
import { twMerge } from 'tailwind-merge';
import { getAccessKey, getBackendURL } from './config'
export const cn = (...classes: ClassValue[]) => twMerge(clsx(...classes));
@ -19,3 +20,20 @@ export const formatTimeDifference = (date1: Date, date2: Date): string => {
else
return `${Math.floor(diffInSeconds / 31536000)} year${Math.floor(diffInSeconds / 31536000) !== 1 ? 's' : ''}`;
};
export const clientFetch = async (path: string, payload: any): Promise<any> => {
let headers = payload.headers;
const url = `${getBackendURL()}${path}`;
const secret_token = getAccessKey();
if (secret_token) {
if (headers == null) {
headers = {};
};
headers['Authorization'] = `Bearer ${secret_token}`;
payload.headers = headers;
};
return await fetch(url, payload);
};