🤖 A Telegram bot that integrates with OpenWeb UI's OpenAI compatible APIs to provide chat functionality.
## Configuration
The configuration is loaded from environment variables, here are the environment variables you can set:
### OpenAI Endpoint Options
-`OPENAI_ENDPOINT`: The base OpenAI API compatible endpoint. example: `http://localhost:3000/ollama/v1/`. http & https both work.
-`OPENAI_API_KEY`: The `Bearer` token API Key, example: `sk-12345667890abcdefghijkl`
### Model Options
-`MODEL`: The model to use. Default is `llama3:instruct`.
-`MODEL_TWEAK_LEVEL`: The level of tweaking to apply to the model. Set to `advanced` to make Penalty tweaks take effect, else - none are provided to API. Default is `minimal`, i.e. no Penalty parameters.
### Model Tweaks
-`MAX_TOKENS`: The maximum number of tokens that the model can generate. Default is `1024`.
-`TEMPERATURE`: Controls the randomness of the model's output. Higher values make the output more random. Default is `0.8`.
-`REPEAT_PENALTY`: Penalty for repeating the same token. Default is `1.2`.
-`CONTEXT_LENGTH`: The maximum number of tokens in the context. Default is `8192`.
-`PRESENCE_PENALTY`: Penalty for using tokens that are not in the context. Default is `1.5`.
-`FREQUENCY_PENALTY`: Penalty for using tokens that are used frequently. Default is `1.0`.
### Bot Configuration
-`API_TOKEN`: The Telegram Bot API token.
## Usage
Only private chat is tested as of now.
### Chatting
If you send messages without a reply, bot treats that as starting a new chat / thread. To continue a chat / thread, reply to the last (or previous) message you want to pick up the conversation from.
### Commands
The bot supports two commands:
1. To set system prompt, use `/reset <system prompt>`.
- The default system prompt is `You are a friendly assistant`.
- Once you set a custom system prompt, it will remain set until you either change it or bot is restarted.
2. To regenerate a response, reply to the message you want to regenerate from and send `/resend`
- Once the bot is restarted, the conversation history is lost and thread can't be continued.
