mirror of https://github.com/danny-avila/LibreChat.git synced 2025-12-16 16:30:15 +01:00

* fix .env.example

comment out DALLE_API_KEY by default and remove `user_provided` as default value since it is not supported

* 📚 docs update

2023-12-07 22:44:26 -05:00

5.5 KiB

Raw Blame History

PandoraNext Deployment Guide

If you're looking to use the ChatGPT Endpoint in LibreChat, setting up a reverse proxy is a essential. PandoraNext offers a robust solution for this purpose. This guide will walk you through deploying PandoraNext to enable the CHATGPT_REVERSE_PROXY for use with LibreChat.

Using this method you will only be able to use text-davinci-002-render-sha with PandoraNext in LibreChat. Other models offered with the plus subscription do not work.

You can use it locally in docker or deploy it onthe web for remote access.

Deploy Online by Duplicating Hugging Face Space

To deploy PandoraNext online by duplicating the Hugging Face Space, follow these steps:

Get your PandoraNext license id here: PandoraNext Dashboard

Configure config.json: Edit the following config.json. Specify your license_id and proxy_api_prefix. For the proxy_api_prefix, use at least 8 characters, avoid characters that can't be used in a URL and make sure it's unique.

Here's the config.json for your reference:

{
"bind": "0.0.0.0:8181",
"tls": {
    "enabled": false,
    "cert_file": "",
    "key_file": ""
},
"timeout": 600,
"proxy_url": "",
"license_id": "",
"public_share": false,
"site_password": "",
"setup_password": "",
"server_tokens": true,
"proxy_api_prefix": "",
"isolated_conv_title": "*",
"captcha": {
    "provider": "",
    "site_key": "",
    "site_secret": "",
    "site_login": false,
    "setup_login": false,
    "oai_username": false,
    "oai_password": false
},
"whitelist": null
}

Hugging Face Space: Visit the PandoraNext LibreChat Space on Hugging Face.
Duplicate the Space: Utilize the available options to duplicate or fork the space into your own Hugging Face account.

Fill the required secrets When asked for the SECRETS,

for CONFIG_JSON use the whole content of the config.json you just modified,
for TOKENS_JSON use the following default token.json:

{
  "test-1": {
      "token": "access token / session token / refresh token",
      "shared": true,
      "show_user_info": false
  },
  "test-2": {
      "token": "access token / session token / refresh token",
      "shared": true,
      "show_user_info": true,
      "plus": true
  },
  "test2": {
      "token": "access token / session token / refresh token / share token",
      "password": "12345"
  }
}

Configure LibreChat: In the .env file (or secrets settings if you host LibreChat on Hugging Face), set the CHATGPT_REVERSE_PROXY variable using the following format:
```
CHATGPT_REVERSE_PROXY=http://your_server_domain.com/your_proxy_api_prefix_here/backend-api/conversation
```
- Replace your_server_domain.com with the domain of your deployed space.
  - you can use this format: https://username-pandoranext.hf.space (replace username with your Huggingface username)
- Replace your_proxy_api_prefix_here with the proxy_api_prefix you have set in your config.json.
- The resulting URL should look similar to: https://username-pandoranext.hf.space/your_proxy_api_prefix_here/backend-api/conversation

Deploy Locally Using Docker

For local deployment using Docker, the steps are as follows:

Clone or Download the Repository: Get the latest release from the PandoraNext GitHub repository.
```
git clone https://github.com/pandora-next/deploy.git
```
Get your PandoraNext license id here: PandoraNext Dashboard

Configure config.json: Within the cloned repository, in the data folder, edit config.json. Specify your license_id and proxy_api_prefix. For the proxy_api_prefix, use at least 8 characters, avoid characters that can't be used in a URL and make sure it's unique.