AI Agents & Local LLMs
RTILA X allows you to use 100% private Local LLMs or cloud providers (via OpenRouter) to generate automation scripts, reason about page layouts, and execute AI Skills.
Use this category to discuss model performance, prompt engineering, RAG (Knowledge Base) settings, and custom .md AI Skills.
When asking for help here, please provide:
- AI Provider & Model — Are you using OpenRouter (e.g.,
claude-3.5-sonnet) or a Local LLM (e.g.,qwen3-8b.Q4_K_M.gguf)? - Hardware Specs (if Local) — CPU, RAM, and GPU (if applicable).
- The Prompt / Skill — What did you ask the AI to do, or what does your custom Skill Markdown look like?
- The Issue — Did the AI hallucinate a selector? Did the local server fail to start? Include the exact error logs.
Quick Reference:
| Issue | Likely Cause | Fix |
|---|---|---|
| “Context length exceeded” | Page DOM is too large | Increase Context Size in Preferences (e.g., 100K) or use a cloud model with a larger window |
| Local LLM is very slow | CPU-only inference | Use a smaller quantized model (Q4_K_M) or enable GPU offloading |
| AI generates bad selectors | Stale DOM context | Make sure the AI is using the analyze_target_page tool to get fresh DOM data |
| Local server fails to start | Port conflict or missing model | Check the logs in ~/.rtila-x/logs/llama-server.log |