Welcome to PatentForgeLocal v0.1.0 — Fully Local Patent Analysis

[scottconverse]

Hey everyone! PatentForgeLocal is now open source.

What it does: PatentForgeLocal is a fully local fork of PatentForge that replaces cloud AI with on-device inference using Ollama + Gemma 4. You describe your invention, and it runs a complete patent analysis pipeline -- feasibility research, prior art search, claim drafting, compliance checking, and application generation -- entirely on your machine. Your inventions never leave your computer.

What it does not do: This is a research tool, not a legal service. The author is not a lawyer, the AI is not a lawyer, and none of the output is legal advice. It is designed to help you prepare for a meeting with a real patent attorney -- not replace one.

Current status (v0.1.0):

• Fully local AI -- Ollama + Gemma 4 26B runs on your hardware
• 6-stage feasibility analysis
• Prior art search via USPTO Open Data Portal
• 3-agent claim drafting pipeline (Planner, Writer, Examiner)
• 4-check compliance validation with MPEP citations
• 5-agent application generator with Word export
• GPU/NPU hardware auto-detection for AMD Ryzen platforms
• 1,229 automated tests across all services

Requirements: Windows, macOS, or Linux. 32 GB RAM minimum (64 GB recommended). Ollama installed locally.

If you are trying it out, I would love to hear how it goes. File bugs as GitHub issues, but use Discussions for questions, ideas, and general chat.