Yujin Forge -- User manual

Voice-first development assistant for building React + NAC-3 apps. Designed so a caregiver or a kid learning to code can operate it without touching the terminal.

Welcome

Yujin Forge does three simple things:

Read documents to you out loud (PDF, Word, HTML, MD, EPUB, CSV, Excel, source code, txt, RTF).
Take a design specification and generate the application code automatically.
Chat with an AI assistant that knows your project and can read the manifest, search the SPEC, edit components, make commits.

Installation

Your institution's technician should already have Yujin Forge installed. If you have to install it yourself:

Install Node.js 18.17 or newer from nodejs.org.
Open a terminal and run:
```
npm install -g @yujin/forge-cli@alpha
```
Verify with yf --version.
Run yf doctor to check the environment.

First run

From your project folder:

yf chat

Your browser opens with the Yujin panel. Click the Keys button (top right) and load the API keys your technician gave you. Each key shows ONCE when you enter it; afterwards you can only replace it, not read it back.

Voice commands

Click the microphone button and say one of these commands. Yujin understands Spanish, English, Portuguese, French, German, Italian, Japanese, Chinese, Hindi and Arabic.

What you say	What happens
"Open the filename"	Opens the document.
"Next" / "Continue"	Advances to the next block.
"Search word"	Searches and reads the first match.
"Bookmark here as name"	Sets a bookmark.
"Jump to bookmark name"	Returns to the bookmark.
"Recap" / "Summary"	Reread the last blocks.
"Open the manual"	Opens this manual in your language.

Specification ingest

Drag a spec file (PDF, Word, HTML, markdown, etc) onto the panel. A purple screen with a kanji appears. Drop the file on it.

Yujin:

Reads the document.
Asks Claude to extract requirements, entities, endpoints, components and a summary.
Asks Claude to produce a plan of files to create.
Shows you the plan. Say "approve" if it looks right.
Creates the files in your project folder.
If something went wrong, say "rollback" and Yujin undoes the whole thing.

Troubleshooting

The microphone doesn't work: The browser asks for permission the first time. If you rejected it by accident, click the lock next to the URL and change the permission.
It doesn't understand me: Try another voice provider. If you only have Google loaded, also add the OpenAI key (slot openai) in the Keys modal; Whisper handles accents better.
The voice sounds bad: Yujin speaks with Google's voice by default. For a more natural voice, load the ElevenLabs key in the elevenlabs slot.
I want to start over: Close Yujin. Delete the ~/.yujin-forge/ folder (Linux/Mac) or %USERPROFILE%\.yujin-forge\ folder (Windows). Run yf chat again and reload the keys.

Support

For support, email pablo@rpaforce.com.

Technical docs: docs/HITO_1_ARCHITECTURE.md + docs/USAGE.md + docs/RECAP_CLAUDE_COSTS.md in the repository.