Gaming PC becomes local AI

Short summary

Developer converted gaming PC into local AI lab using Vision Language Models to eliminate ~1,500-token cost per screenshot. Built VLM API with Ollama and qwen2.5-vl model on RTX 4070, achieving 8-second inference locally while preserving privacy.

• •Identified token waste: screenshots costing ~1,500 tokens each to cloud APIs like Claude
• •Built local VLM solution using Ollama and qwen2.5-vl:7b model on gaming PC with RTX 4070
• •Open-sourced VLM Local Parser API; now using same approach for drone-based plant identification

Generated with AI, which can make mistakes.