docs/plans/local_usage_checklist.md

• Integrated open-source AI models end-to-end without relying on external APIs (Llama-2 7B, Mistral 7B, GPT-J 6B, or smaller variants of Stable Diffusion (like SD 1.5) for image tasks. Make sure to download or convert these models into a format that suits GPU (FP16 or int4).)

• Leveraged local RTX 2060 to enable features beyond API limits (longer conversations, more generations, multiple image generations, complex multi-step reasoning)

• Wrapped local service as an MCP-compatible tool via a Gradio app, allowing agent to call the local GPU–powered tool

• Documented that all AI components run locally on the RTX 2060, eliminating dependence on proprietary services or sponsor credits