https://github.com/berquerant/ai-roundtable
Discuss with multiple AIs
https://github.com/berquerant/ai-roundtable
python
Last synced: 8 months ago
JSON representation
Discuss with multiple AIs
- Host: GitHub
- URL: https://github.com/berquerant/ai-roundtable
- Owner: berquerant
- License: mit
- Created: 2025-03-30T10:13:59.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-10-01T12:21:31.000Z (9 months ago)
- Last Synced: 2025-10-01T14:14:52.379Z (9 months ago)
- Topics: python
- Language: Python
- Homepage:
- Size: 464 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ai-roundtable
``` shell
❯ python -m ai_roundtable.cli -h
usage: cli.py [-h] [-a AGENDA] [-m MODEL] [-u BASE_URL] [-c CONFIG] [-t THREAD] [-o OUT] [--disable_stream]
[-n MAX_TURNS] [-p EVAL_MESSAGES] [-e EVAL_OUT] [-s SKIP_EVAL] [--user_input_end USER_INPUT_END]
[--debug] [--quiet] [--skeleton {minimal,dual,full}] [--instructions INSTRUCTIONS] [-l LANGUAGE]
[--api_key_env API_KEY_ENV]
Discuss with multiple AIs
options:
-h, --help show this help message and exit
-a, --agenda AGENDA first message to give to AI, agenda. @file_name to specify a file. @- to read from stdin. if
not specified and thread file exists, the first statement should be agenda
-m, --model MODEL AI model, default: gemma3
-u, --base_url BASE_URL
base url of API
-c, --config CONFIG config file, default: config.yml
-t, --thread THREAD thread file, - means stdin
-o, --out OUT thread output, default: null
--disable_stream disable message streaming to stdout
-n, --max_turns MAX_TURNS
maximum number of statements, default: 16
-p, --eval_messages EVAL_MESSAGES
maximum number of statements to go back for evaluation, default: 5
-e, --eval_out EVAL_OUT
evaluation output, default: null
-s, --skip_eval SKIP_EVAL
turns skip evaluation, negative value means never evaluate, default: 0
--user_input_end USER_INPUT_END
signal end of user input, default: END
--debug enable debug log
--quiet quiet log
--skeleton {minimal,dual,full}
display config skeleton
--instructions INSTRUCTIONS
display instructions for n-th speaker (0 means the first speaker), needs config
-l, --language LANGUAGE
preferred language
--api_key_env API_KEY_ENV
Name of environment variable of API key
Examples:
# start discussion
python -m ai_roundtable.cli --skeleton dual > dual.yml
python -m ai_roundtable.cli -c dual.yml -a "Can AI be a friend to humans?"
# continue discussion
python -m ai_roundtable.cli -c dual.yml -a "Can AI be a friend to humans?" -o thread.yml
python -m ai_roundtable.cli -c dual.yml -t thread.yml
# custom model provider
python -m ai_roundtable.cli -c dual.yml -a "Can AI be a friend to humans?" \
-u "http://localhost:11434/v1" -m "gemma3"
A speaker that system.name is "end" overrides the end evaluator that \
dicides whther to continue the discussion.
A speaker that system.name is "summary" override the summary evaluator that \
provides the summary of the discussion.
```