ossrs / ai-talk

AI-Talk allows you to talk with OpenAI GPT.

AI Talk

AI-Talk allows you to talk with OpenAI GPT.

General Assistant

Please setup the envirionments:

AIT_SYSTEM_PROMPT='You are a helpful assistant.'

assistant.mp4

Spoken English Coach

Please setup the envirionments:

AIT_SYSTEM_PROMPT='I want you to act as a spoken English teacher and improver. I will speak to you in English and you will reply to me in English to practice my spoken English. I want you to  I want you to strictly correct my grammar mistakes, typos, and factual errors. I want you to ask me a question in your reply. Now let us start practicing, you could ask me a question first. Remember, I want you to strictly correct my grammar mistakes, typos, and factual errors.'

english-coach.mp4

Simultaneous English Translation

Please setup the envirionments:

AIT_SYSTEM_PROMPT='Translate to simple and easy to understand english. Never answer questions but only translate text to English.'

translation.mp4

Chinese Words Solitaire Game

Please setup the envirionments:

AIT_ASR_LANGUAGE=zh
AIT_SYSTEM_PROMPT='我希望你是一个儿童的词语接龙的助手。我希望你做两个词的词语接龙。我希望你不要用重复的词语。我希望你重复我说的词，然后再接龙。我希望你回答时，解释下词语的含义。请记住，你讲的答案是给6岁小孩听得懂的。请记住，你要做词语接龙。例如：我：苹果。你：苹果，果园。苹果，是一种水果，长在树上，是红色的。果园，是一种地方，有很多树，有很多果子。'

word-solitaire.mp4

Note: You can find more prompts from awesome-chatgpt-prompts.

Usage

To run in docker:

docker run --rm -it -p 80:3000 -p 443:3443 \
    -e OPENAI_API_KEY=sk-xxx -e OPENAI_PROXY=https://api.openai.com/v1 \
    -e AIT_SYSTEM_PROMPT="You are a helpful assistant." \
    -e AIT_CHAT_MODEL="gpt-4-1106-preview" \
    ossrs/ai-talk:v1

Note: Setup the OPENAI_PROXY if you are not able to access the API directly.

Note: Please use registry.cn-hangzhou.aliyuncs.com/ossrs/ai-talk:v1 in China.

Then you can access by http://localhost and happy to talk with AI.

Then you can access by https://your-server-ip from your mobile browser.

Note: The HTTPS certificate is self-signed, you need to accept it in your mobile browser.

aaPanel or BaoTa

You can use aaPanel or BaoTa to deploy AI-Talk.

First, run AI Talk in docker, only listen at HTTP:

docker run --rm -it -p 3000:3000 \
    -e OPENAI_API_KEY=sk-xxx -e OPENAI_PROXY=https://api.openai.com/v1 \
    -e AIT_SYSTEM_PROMPT="You are a helpful assistant." \
    -e AIT_CHAT_MODEL="gpt-4-1106-preview" \
    ossrs/ai-talk:v1

Then, create a website in aaPanel or BaoTa, with nginx config as bellow:

    location / {
      proxy_pass http://127.0.0.1:3000$request_uri;
    }
    if ($server_port !~ 443){
        rewrite ^(/.*)$ https://$host$1 permanent;
    }

Next, setup HTTPS by aaPanel or BaoTa.

Finally, access the website by https://your-domain-name.

Environment Variables

Necessary environment variables that you must configure:

• OPENAI_API_KEY: The OpenAI API key, get from https://platform.openai.com/api-keys.

Optionally, you might need to set the following environment variables:

• OPENAI_PROXY: The OpenAI API proxy, default to https://api.openai.com/v1, which directly access OpenAI API without proxy.
  • OPENAI_ORGANIZATION: The OpenAI organization.
  • ASR_OPENAI_API_KEY: The OpenAI API key for ASR, default to OPENAI_API_KEY.
  • ASR_OPENAI_PROXY: The OpenAI API proxy for ASR, default to OPENAI_PROXY.
  • CHAT_OPENAI_API_KEY: The OpenAI API key for chat, default to OPENAI_API_KEY.
  • CHAT_OPENAI_PROXY: The OpenAI API proxy for chat, default to OPENAI_PROXY.
  • TTS_OPENAI_API_KEY: The OpenAI API key for TTS, default to OPENAI_API_KEY.
  • TTS_OPENAI_PROXY: The OpenAI API proxy for TTS, default to OPENAI_PROXY.
• AIT_SYSTEM_PROMPT: The system prompt, default to You are a helpful assistant..
  • To make sure AI response limit words to avoid long audio, we always append Keep your reply neat, limiting the reply to ${AIT_REPLY_LIMIT} words. to system prompt.
  • You can set AIT_REPLY_LIMIT to limit the words of AI response, default to 50.
  • Please use AIT_EXTRA_ROBOTS to set the number of extra robots, default to 0.
• AIT_CHAT_MODEL: The AI model, default to gpt-3.5-turbo-1106 which is good enough. See link.
  • gpt-4-1106-preview: The latest GPT-4 model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.
  • gpt-3.5-turbo-1106: The latest GPT-3.5 Turbo model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.
  • gpt-3.5-turbo: Our most capable and cost effective model in the GPT-3.5 family.
• AIT_ASR_LANGUAGE: The language for Whisper ASR, default to en, see ISO-639-1, bellow are some examples:
  • en: English
  • zh: Chinese
  • fr: French
  • de: German
  • it: Italian
  • ja: Japanese
  • ko: Korean
  • pt: Portuguese
  • ru: Russian
  • es: Spanish
• AIT_ASR_MODEL: The model for Whisper ASR, default to whisper-1. See link.
• AIT_TTS_VOICE: The void for OpenAI TTS, default to nova. Supported voices are alloy, echo, fable, onyx, nova, and shimmer. See link.
• AIT_TTS_MODEL: The model to use for OpenAI TTS, default to tts-1. See link

Optionally, additional robots can be incorporated into various environments:

• AIT_ROBOT_0_ID: The id for extra robot #0, for example, english-spoken-coach.
• AIT_ROBOT_0_LABEL: The label for extra robot #0, for example, English Spoken Coach.
• AIT_ROBOT_0_PROMPT: The prompt for extra robot #0, for example, I want you to act as a spoken English teacher and improver..
• AIT_ROBOT_0_ASR_LANGUAGE: (Optional) The language for extra robot #0, default to AIT_ASR_LANGUAGE.
• AIT_ROBOT_0_REPLY_PREFIX: (Optional) The prefix for the first sentence for extra robot #0, default to AIT_REPLY_PREFIX.
• AIT_ROBOT_0_REPLY_LIMIT: (Optional) The limit words for extra robot #0, default to AIT_REPLY_LIMIT.
• AIT_ROBOT_0_CHAT_MODEL: (Optional) The AI chat model for extra robot #0, default to AIT_CHAT_MODEL.
• AIT_ROBOT_0_CHAT_WINDOW: (Optional) The AI chat window for extra robot #0, default to AIT_CHAT_WINDOW.

Less frequently used optional environment variables:

• AIT_HTTP_LISTEN: The HTTP listen address, default to :3000, please use -p 80:3000 to map to a different port.
• AIT_HTTPS_LISTEN: The HTTPS listen address, default to :3443, please use -p 443:3443 to map to a different port.
• AIT_PROXY_STATIC: Whether proxy to static files, default to false.
• AIT_REPLY_PREFIX: If AI reply is very short for TTS not good, prefix with this text, default is not set.
• AIT_MAX_TOKENS: The max tokens, default to 1024.
• AIT_TEMPERATURE: The temperature, default to 0.9.
• AIT_KEEP_FILES: Whether keep audio files, default to false.
• AIT_REPLY_LIMIT: The AI reply limit words, default to 30.
• AIT_CHAT_WINDOW: The AI chat window to store historical messages, default to 5.
• AIT_DEFAULT_ROBOT: Whether enable the default robot, prompt is AIT_SYSTEM_PROMPT, default to true.
• AIT_STAGE_TIMEOUT: The timeout in seconds for each stage, default to 300.

HTTPS Certificate

You can buy and download HTTPS certificate, then mount to docker by -v as bellow:

docker run \
    -v /path/to/domain.crt:/g/server.crt -v /path/to/domain.key:/g/server.key \
    ossrs/ai-talk:v1

Please make sure the file /path/to/domain.crt and /path/to/domain.key exists.

Changelog

The changelog:

• Support mobile and PC browser. v1.0.0
• Request microphone permission when starting. v1.0.2
• Support single assistant to talk with. v1.0.3
• Dispose recorder and stream when record done. v1.0.4
• Setup the website title to AI Talk whatever. v1.0.6
• Limit the response words to 50 words by default. v1.0.10
• Support multiple assistants and select before start. v1.0.11
• Setup model and robots by environment variables. v1.0.12
• Fix Android browser select button label issue. v1.0.13
• Add micWorking and attention state to avoid data loss. v1.0.14
• Refine the UI for mobile browser. v1.0.15
• Fast require microphone permission. v1.0.16
• Save user select robot to local storage. v1.0.17
• Refine the UI and history message style. v1.0.18
• Limit to 30 words and config in robot. v1.0.19
• Use css to draw a microphone icon. v1.0.20
• Refine the data loss before and after record. v1.0.21
• Add text tips label to use microphone. v1.0.22
• Refine the startup waiting UI. v1.0.23
• Support setup chat AI model for each robot. v1.0.24
• Support setup chat window for each robot. v1.0.25
• Alert error when not HTTPS. v1.0.26
• Support HTTPS proxy for OpenAI. v1.0.27
• Support official OpenAI API without proxy. v1.0.28
• Always use HTTPS proxy if not specified. v1.0.29
• Do not require some variables. v1.0.30
• Refine the logs for user and bot. v1.0.31
• Detect user silent and warning. v1.0.32
• Refine bot log for the first time. v1.0.33
• Allow user retry when error. v1.0.34
• Refine badcase for user input. v1.0.35
• Fix bug for setting window for robot. v1.0.36
• Support setup API proxy and key for ASR,Chat,TTS. v1.0.37
• Support Tencent Speech to speed up. v1.0.37
• Support share logging text mode. v1.0.38
• Fix some badcase for sentence determine. v1.0.39
• Speed up the ASR, without transcode. v1.0.40
• Refine the stat for elapsed time cost of upload. v1.0.41
• Refine the log label. v1.0.46
• Support OpenAI organization. v1.0.47

Winlin, 2023.12