Craig Processor

The purpose of this project should be to do a number of things We're taking the url of a recorded voice call using the craig bot which you can find here

https://craig.chat/

Then we want to do a number of possible things either in combination of separately

• get craigbot recording metadata/summary

• download craigbot files (single/multi track with different formats)

• convert+/combine craigbot files downloaded to opus (create compression for voices) or mp3

• build a transcription (by downloading individual stream files), transcribing each one and then combining these transcriptions

• building a summary of this transcription using chatgpt or some other AI

• discord actions/once offs

  • post these elements/final results to a channel or edit a message with it

• discord full intetration or listening on triggers talking directly to the bot to ask it to give us this by giving it a craigbot url right clicking on a disord message that craigbot posted of the recording and having our discord bot post one of the elements above (such as a summary/audo file/transcription/summary) in the same channel

  after inviting the bot to a group chat or discord server, /slash command with a url of craigbot download url to post something (like a recording or a summary or transcription) in a certain channel

  both these tasks will take some time and we may have multiple jobs running at the same time so a status of what jobs are running and their progress

• web server finally a web server interface along with an api that we can see processing jobs, request jobs, and configure any hooks on messages to auto actions and so on

Each component should be able to be run separatly or in combination for example

• we have the zip or flac files already, combine them and transcribe them and summarise them via local command running
• we have just the url, do the above with a command
• we have hte url, just download hte files
• on discord, download the files and generate the file and post it etc etc

Configuration and examples

This tool looks for config.json in the current working directory by default. You can override with --config /path/to/config.json.

If config.json is missing and you didn't explicitly pass --config, that is fine unless an action requires secrets (for example using the OpenAI API for transcription, or posting to Discord). In that case the command will error and indicate you must either pass the required keys as CLI flags or provide a config.json with the required fields.

Example config.json (see config.example.json):

{
    "openai": { "api_key": "..." },
    "discord": { "bot_token": "...", "channel_aliases": { "default": "123...", "reports": "456..." }, "webhook_aliases": { "primary": "https://...", "mirror": "https://..." } }
}

Discord tips

You can post results back to Discord in two ways:

• Webhook: set discord.webhook_url in config.json or pass --post-discord-webhook on the CLI.
• Bot token: set discord.bot_token in config.json, export DISCORD_BOT_TOKEN in your environment, or pass --post-discord-bot-token on the CLI.

Channel selection supports aliases only. Add discord.channel_aliases to config.json as a mapping of alias -> channel_id (strings). When posting with a bot token you may pass either a channel alias via --post-discord-channel (alias required) or select an alias configured in config.json.

Resolution priority for bot token/channel is:

1. CLI flags (--post-discord-bot-token, --post-discord-channel)
2. config.json (discord.bot_token, discord.channel_aliases)
3. Environment variable DISCORD_BOT_TOKEN (for bot token only)

If no webhook aliases nor bot token+channel alias are configured, the post action will skip posting and print a helpful message.

Webhooks use aliases only (recommended). Add discord.webhook_aliases to config.json as a mapping of alias -> webhook_url. On the CLI you may pass --post-discord-webhook with either a single alias or a comma-separated list of aliases/URLs. Each resolved webhook will be posted to in turn. If you pass a raw URL it will be used directly, but prefer aliases for portability.

Examples

Minimal: download and transcribe per-track (defaults)

./craigify.py process -i "<URL>" --actions download,transcribe --download-file-type flac --transcribe-mode tracks

Full example: download, create final opus, transcribe, summarize and post to Discord webhook

./craigify.py process -i "<URL>" --actions download,post,transcribe,summarize \
    --download-file-type flac --download-mix individual --download-final-format opus --download-opus-bitrate 32k \
    --transcribe-mode tracks --transcribe-backend faster_whisper --transcribe-model medium \
    --summarize-style points --post-discord-webhook "https://discord.com/api/webhooks/..."

so we'll want modular classes/files that can be run with

if __name__ == "__main__":
	main()

but also run from other scripts as we combine things together