fgpt

A free reverse proxy and CLI tool for OpenAI GPT-3.5-turbo

It allows you to use the GPT-3.5 API without needing to sign up for an API key or pay for usage.

? OpenAI GPT-3.5-turbo is free to use, without any account or API key
? DON'T USE IN PRODUCTION, ONLY FOR PERSONAL USE/TESTING

Features

• REPL mode, you can input questions and get answers interactively
• Reverse proxy mode, you can use the OpenAI OpenAPI with a local server
• CLI mode, with shell pipe, file input, code output, etc.
• ? Support https proxy and socks5 proxy

Download precompiled binary

• Linux x64 executable binary
• Mac M1/M2 executable binary
• Windows (Coming soon)
• Or via Docker
• Or build from source (see below, cargo is required)

  cargo install fgpt

How to use

# To get answers from GPT-3.5
fgpt "How to get a domain's MX record on linux shell?"

# Output plain code -c/--code
fgpt -c "Write python code to reverse a string"

# With pipe
cat README.md | fgpt "summarize for reddit post"

# With stdin
fgpt "Convert the follow csv data to json, without any description" < contacts.csv

# With file -f/--file
fgpt -f contacts.csv  "Convert the follow csv data to json, without any description"

# REPL mode
fgpt
>> Write a javascript code to reverse a string
...

With http proxy

If you are unable to connect , you can try using a proxy. HTTP and SOCKS5 proxies are supported. For example:

# 1. pass the proxy address by -p/--proxy
fgpt -p 'socks5://127.0.0.1:9080' "Linux command to list files in a directory"

# 2. pass the proxy address by environment variable
export HTTPS_PROXY='socks5://127.0.0.1:9080'
fgpt "Linux command to list files in a directory"

# 3. use alias to set the proxy address
alias fgpt='fgpt -p "socks5://127.0.0.1:9080"'
fgpt "Linux command to list files in a directory"

Dump stats

fgpt --stats "Linux command to list files in a directory"

Use by docker

docker run -it --rm shenjinti/fgpt "Linux command to list files in a directory"

How to use Reverse Proxy

Offering free self-hosted API access to ChatGPT. This is useful if you want to use the OpenAI API without needing to sign up for an API key.

1. Start the server

fgpt -s 127.0.0.1:4090

Your local server will now be running and accessible at: http://127.0.0.1:4090/v1/chat/completions

2. Example Usage with OpenAI Libraries

import openai
import sys
openai.api_key = 'nothing'
openai.base_url = "http://127.0.0.1:4090/v1/"

completion = openai.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[
        {"role": "user", "content": "Write a javascript simple code"},
    ],
    stream=True,
)

for chunk in completion:
    print(chunk.choices[0].delta.content, end='')
    sys.stdout.flush()

print()

or test with curl:

curl -X POST -H "Content-Type: application/json" -d '{"model":"gpt-3.5-turbo",
"messages":[{"role":"user","content":"Write a javascript simple code"}], 
"stream":true}' 
http://127.0.0.1:4090/v1/chat/completions

curl -X POST -H "Content-Type: application/json" -d '{"model":"gpt-3.5-turbo",
"messages":[{"role":"user","content":"Write a javascript simple code"}]}' 
http://127.0.0.1:4090/v1/chat/completions