Is it possible to add local LLM support

[marchyuker]

Looking to have a local LLM instance running for enhanted messages analysis using open-source models.

[umputun]

To use a local LLM with tg-spam:

1. Set up a local LLM server with OpenAI-compatible API (like Ollama, LM Studio, or any server exposing OpenAI-compatible endpoints)
2. Configure tg-spam to use your local endpoint: --openai.apibase="http://localhost:YOUR_PORT" and --openai.token="YOUR_KEY_OR_DUMMY_VALUE"
3. Optionally adjust model parameters: --openai.model="YOUR_LOCAL_MODEL_NAME"

No need to modify the code - tg-spam already supports custom API endpoints through the existing OpenAI integration.

[marchyuker]

Which small local model would you recommend to use on cpu with support of english, ukrainian and russian languages?