solanaceae_llama-cpp-web/README.md

55 lines
1.4 KiB
Markdown

# llama.cpp webserver based api + RPBot
used llama embedded webserver api for easier portability
## LlamaCppWeb config options
- `LlamaCppWeb`, `server`
- type: `string`
- host + port
- default: `localhost:8080`
## RPBot config options
TODO: move rpbot to own repo
- `RPBot`, `system_prompt`(, opt contact ID)
- type: `string`
- System prompt that is prefixed
- can contain specific formatters:
- `{self_name}` username for specified chat
- `{online_users}` coma seperated list of online users
- default: `Transcript of a group chat, where {self_name} talks to online strangers.\n{self_name} is creative and curious. {self_name} is writing with precision, but also with occasional typos.\n`
- `RPBot`, `min_messages`(, opt contact ID)
- type: `int`
- TODO: implement per ID
- default: `4`
- `RPBot`, `max_cont_messages`(, opt contact ID)
- type: `int`
- TODO: implement per ID
- default: '4'
- `RPBot`, `max_interactive_delay`(, opt contact ID)
- type: `float`
- TODO: implement per ID
- default: `4.0`
## example config
```json
{
"LlamaCppWeb": {
"server": "localhost:8080"
},
"RPBot": {
"min_messages": 1,
"system_prompt": "A group chat with an artificial intelligence assistant called {self_name}. The assistant gives helpful and detailed answers to the chat participants's questions. The assistant has it's own opinions. {self_name} only ever answers when explicitly asked.\n"
}
}
}
```