Skip to content

Home

Jump to bottom
Nick Bradshaw edited this page Feb 11, 2026 · 3 revisions

Here is a complete rewrite of the Wiki page.

I have structured this to be the "Deep Dive" companion to the README. It removes all references to the obsolete file-based configuration (like prompt.txt) and focuses on the new Database/WebUI-driven workflow.

Complete Installation & Setup Guide

This guide covers the end-to-end process of setting up Scribble, from configuring your server to running your first session.

🐳 Phase 1: Server Preparation

Scribble runs inside a Docker container. This ensures all dependencies (ffmpeg, python libraries, system tools) are isolated and don't mess with your host OS.

1. Prerequisites

You need a machine (Linux, Windows, or Mac) with:

2. Directory Setup

Create a dedicated folder for Scribble. This will hold your configuration and your data.

mkdir scribble
cd scribble

3. The Compose File

Create a docker-compose.yaml file in this folder. Choose the version that matches your hardware.

Option A: NVIDIA GPU (Recommended) Best performance. Transcription is significantly faster.

services:
  scribble:
    container_name: scribble
    image: ghcr.io/goose-ws/scribble:cuda-latest
    ports:
      - 13131:13131
    environment:
      - TZ=America/New_York  # Set your timezone
      - PUID=1000            # User ID to own files
      - PGID=1000            # Group ID to own files
    volumes:
      - "./data:/data"       # Persistent storage for DB and audio
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities: [gpu]
    restart: unless-stopped

Option B: CPU Only Universal compatibility, but slower transcription speeds.

services:
  scribble:
    container_name: scribble
    image: ghcr.io/goose-ws/scribble:cpu-latest
    ports:
      - 13131:13131
    environment:
      - TZ=America/New_York
      - PUID=1000
      - PGID=1000
    volumes:
      - "./data:/data"
    restart: unless-stopped

4. Launch

Start the container in the background:

docker compose up -d

🔐 Phase 2: Initial Login & Security

Scribble does not have a default password. On the very first run, it generates a secure, random temporary password.

  1. Find the Password: Check the container logs.
docker logs scribble
  1. Locate the Banner: Scroll until you see a block that looks like this:
==================================================
FIRST RUN - TEMPORARY PASSWORD: xxxxx-xxxxx-xxxxx
==================================================
  1. Login: Go to http://YOUR_SERVER_IP:13131 and log in with that token.
  2. Secure It: Immediately go to the Settings page and set your own permanent WebUI Password.

⚙️ Phase 3: System Configuration

Before processing files, you need to tell Scribble how to listen and think. Go to the Settings tab.

1. Whisper (The Ears)

This handles the Speech-to-Text.

  • Model Size:

  • small: Fast, good for clear audio. Low RAM usage.

  • medium: The "Goldilocks" zone. Good balance.

  • large-v3: Best accuracy, but requires ~4GB-8GB VRAM (or lots of RAM on CPU).

  • Threads:

  • GPU Users: Leave this at 0.

  • CPU Users: Set this to the number of physical cores you want to dedicate (e.g., 4 or 8).

2. LLM Provider (The Brain)

Scribble needs an AI to read the transcript and write the recap. Select one provider:

Enter your API Key and the Model Name (e.g., gemini-2.0-flash or claude-3-5-sonnet-20240620) and save.

⚔️ Phase 4: Campaign Setup

Unlike older versions of Scribble, you now manage Campaigns in the database. This allows you to run multiple different games (e.g., a Tuesday D&D game and a Friday Sci-Fi game) with different settings.

  1. Go to the Campaigns tab (or Settings).
  2. Create a New Campaign.
  3. Name: The display name (e.g., "Curse of Strahd").
  4. Discord Webhook: (See Phase 5 below).
  5. System Prompt: This is where you teach the AI how to write.

  • Prompt Variables: You can use these special tags in your prompt, and Scribble will fill them in automatically:

  • ${campaignName}

  • ${sessionNumber}

  • ${sessionDate}

  • Example Prompt:

You are the chronicler for the campaign "${campaignName}". Summarize the following transcript from Session {sessionDate}). The players are: Alice (Cleric), Bob (Wizard)... Keep the tone dark and mysterious.

💬 Phase 5: Discord Integration

Scribble relies on Craig for recording and Webhooks for posting.

1. Recording (Craig)

  1. Invite Craig to your server.
  2. Join a voice channel and type /join to start recording.
  3. When finished, Craig will DM you a download link.
  4. IMPORTANT: Download the FLAC (Multi-track) format. This gives you a .zip file. Do not unzip it.

2. Posting (Webhooks)

  1. Create a Forum Channel in your Discord server (e.g., #session-recaps). Forum channels are best because Scribble creates a new thread for every session.
  2. Right-click the channel → Edit Channel → Integrations → Webhooks.
  3. Click New Webhook, name it "Scribble", and copy the Webhook URL.
  4. Paste this URL into your Campaign Settings in Scribble.

🚀 Phase 6: Workflow (Running a Session)

  1. Upload:
  • Go to the Scribble Dashboard.
  • Click Upload.
  • Select the Campaign.
  • Session Number: Check the auto-detected number (e.g., Session 5). You can override this if you are uploading old sessions out of order.
  • Drop the Craig .zip file and submit.
  1. Processing:
  • Scribble will extract the audio and transcribe each speaker individually (to ensure high accuracy).
  • It will then merge them into a master transcript.
  • Finally, it sends the transcript to your LLM to generate the recap.
  1. Result:
  • The recap is automatically posted to Discord.
  • You can view, edit, or regenerate the recap from the Session Detail page in the Web UI.
  • You can also download the full transcript or the recap as PDF/Text files.

Clone this wiki locally