ogres.app is a free and open-source virtual tabletop that you can run in your browser and use to play with your friends. It aims to be a very lightweight alternative to some of the more comprehensive tools available today. Its limited core feature-set is intended to help dungeon masters quickly setup encounters and adventures with only the most important necessities.

Scene Management

• Create, switch between, and delete multiple scenes
• Upload background map images (up to 10 MB)
• Configurable grid with snapping, size, and origin controls
• Toggle grid display, object outlines, and dark mode per scene

Tokens

• Place, move, resize, rotate, and delete tokens
• Set token size categories (Tiny → Gargantuan), label, and image
• Mark tokens as player characters or NPCs
• Hide/reveal tokens (host-only), lock tokens against accidental movement
• Apply light sources and aura radii to tokens
• Track hit points and mark tokens as dead/defeated
• Apply D&D 5e status conditions (Blinded, Charmed, Frightened, Prone, etc.)

Combat & Initiative

• Initiative tracker with turn order, round counter, and HP tracking
• Roll and input initiative values; advance turns automatically

Drawing & Measurement Tools

• Draw circles, rectangles, polygons, cones, and lines on the scene
• Ruler for measuring distances between any two points
• Customizable shape colors and fill patterns
• Optional grid-alignment for drawn shapes

Fog of War

• Draw custom mask areas to hide parts of the map
• Toggle, delete, or batch-reveal/hide masks
• Three lighting states: Revealed, Dimmed, Hidden

Notes & Props

• Place GM notes with icons (bookmark, dice, door, location, fire, skull, question)
• Control per-note visibility (host-only or public)
• Place and manage environmental props/objects from an image library

Multiplayer Sessions

• One-click host/join with shareable room URLs — no sign-up required
• Real-time sync of scene state, token movement, and fog of war
• See connected players and share cursor positions

Player Profiles & Assets

• Customize name, color, description, and avatar image per player
• Shared image library for tokens, maps, and props with thumbnail previews

Data Management

• All data saved locally in the browser (IndexedDB) — nothing sent to a server
• Export full backup to file and restore from backup
• Reset application data at any time

Accessibility & UX

• Keyboard shortcuts for common actions
• Right-click context menus, drag-and-drop from galleries
• Copy/cut/paste objects; multi-select via shift-click or drag selection
• Responsive design suitable for phones and tablets
• Easy to adapt for any tabletop game system

AI Dungeon Master (experimental)

• LLM-powered DM that reads the live board and controls NPC/monster tokens
• Spawns, moves, and removes tokens; rolls initiative; updates HP
• Narrates events to all players in real time via the DM Narration panel
• Two-way chat — ask the DM questions and get in-character responses
• Voice narration via Kokoro TTS (British male narrator by default)
• Backends: Ollama (local/private), Grok xAI (cloud), LangGraph sidecar (multi-step agentic)
• LangGraph mode adds assess → plan → validate → reflect-retry reasoning loop

git clone https://github.com/goyanx/ogresvtt.git
cd ogresvtt
npm install
mkdir logs

Start frontend (required):

npx shadow-cljs watch app

Start backend (optional, only for multiplayer sessions):

clojure -M -m ogres.server.core 5000

Open http://localhost:8080.

ollama pull qwen2.5:14b-instruct-q4_K_M
$env:OLLAMA_ORIGINS="*"
ollama serve

In OgresVTT AI DM panel:

• Backend: Ollama (local, direct) (or LangGraph sidecar + Sidecar LLM backend = Ollama)
• Ollama endpoint: http://localhost:11434
• Model: qwen2.5:14b-instruct-q4_K_M (or another tool-capable model)

copy .env.local.example .env.local
cd ai_dm
pip install -r requirements.txt
uvicorn ai_dm.main:app --port 8765 --reload

Then in OgresVTT AI DM panel:

• Backend: LangGraph sidecar (multi-step)
• Sidecar URL: http://localhost:8765
• Sidecar LLM backend: Ollama or Grok

If you use Sidecar LLM backend = Grok, you can configure .env.local for defaults (file can be in repo root or ai_dm/, or your current working directory when launching uvicorn):

XAI_API_KEY=xai-...
GROK_MODEL=grok-3-mini

Supported env keys:

• Default sidecar backend: AI_DM_DEFAULT_BACKEND (ollama or grok)
• Grok API key: XAI_API_KEY or GROK_API_KEY or AI_DM_GROK_API_KEY
• Grok model: GROK_MODEL or XAI_MODEL or AI_DM_GROK_MODEL
• Ollama endpoint: AI_DM_OLLAMA_ENDPOINT or OLLAMA_ENDPOINT
• Ollama model: AI_DM_OLLAMA_MODEL or OLLAMA_MODEL
• Query loop rounds: AI_DM_MAX_QUERY_ROUNDS (default 4)
• Validation retries: AI_DM_MAX_RETRIES (default 2)

1. Open the ✦ AI DM panel.
2. Enable AI DM.
3. Configure backend/model/scenario.
4. Click Run turn to test.
5. Enable Auto-approve + set interval for autonomous turns.
6. Use T (Narration) panel to view output or chat with the DM.

Powered by Kokoro-82M.

pip install kokoro soundfile numpy

Windows phonemization may require espeak-ng: espeak-ng latest releases

• AI DM Error: Grok API key not set
  • Direct mode: set key in AI DM panel.
  • LangGraph mode: set key in panel or .env.local.
• LangGraph cannot connect
  • Ensure sidecar is running on http://localhost:8765.
• No multiplayer sync
  • Ensure Clojure backend is running on port 5000.

docker compose up -d

For full configuration options see the wiki docs.

Interested in helping fix bugs or extending features? Look for issues labeled beginner friendly and comment that you'd like to work on it.

The sidecar now initializes a local SQLite database for compendium RAG and campaign runtime memory (characters, stats, combat events, NPC relationships, map triggers, rulings).

Default database path:

• ai_dm/data/dm.sqlite

Optional env vars:

• AI_DM_DB_PATH — override sqlite file path
• AI_DM_ADMIN_ALLOW_WRITE=true — allow write SQL in admin console

Open admin UI:

• http://localhost:8765/dm-admin

Admin APIs:

• GET /dm-admin/api/tables
• GET /dm-admin/api/table/{table_name}?limit=100&offset=0
• POST /dm-admin/api/query
• GET /dm-admin/api/maps?limit=200
• POST /dm-admin/api/maps/upsert

Ingest handpicked DnD manuals (Markdown/Text) into RAG tables:

python -m ai_dm.ingest_compendium \
  --source-title "DnD 5.5e Manual" \
  --edition "5.5e" \
  --doc-title "PHB" \
  --file C:\path\to\manual.md

After ingestion, the agent can call retrieve_rules and get_monster_stats during planning.

See docs/AI_DM_DATA_ADMIN.md for schema/admin details.

LangGraph DM can now emit show_map action tool calls to request scene/map switching in the client, backed by map_scenes config in SQLite.

Use the standalone CLI importer to ingest marker_single output into SQLite RAG and campaign tables. This script is independent from sidecar runtime code (it does not import or modify ai_dm modules).

Script path:

• scripts/marker_import_grok.py

Input:

• folder produced by marker_single (reads .md / .txt)

Output:

• RAG rows in comp_* tables
• NPC + map provisioning in camp_*, npc_*, map_* tables via Grok function/tool calls
• extraction artifact file: <marker-dir>/ogres_import_extracted.json

Required env for Grok:

• XAI_API_KEY (or GROK_API_KEY / AI_DM_GROK_API_KEY)
• Optional model override: AI_DM_GROK_MODEL (or GROK_MODEL / XAI_MODEL)
• Optional DB override: AI_DM_DB_PATH

Example:

python scripts/marker_import_grok.py `
  --marker-dir "C:\Users\goyan\Documents\Code\ogresvtt-testcampaigns\vecna_nest_of_eldritch_eye" `
  --source-title "Vecna: Nest of the Eldritch Eye" `
  --edition "5.5e"

Useful flags:

• --dry-run (no DB writes)
• --skip-grok (RAG ingest only)
• --dotenv <path> (use non-default env file)
• --max-files N (cap number of processed files)
• --legacy-json-mode (fallback: non-tool-call extraction path)

See docs/AI_DM_DATA_ADMIN.md for DB domains and tooling context. Quickstart: docs/MARKER_IMPORTER_QUICKSTART.md