Ontology Dashboard

An ontology-driven operations dashboard with multi-LLM chat, Zed-inspired layout, and zero database.

What is this?

A single-page operations dashboard that reads your workspace files (Markdown, JSONL, JSON) and renders them as live sections — tasks, projects, pipeline, git log, financials. It includes a multi-LLM chat panel with tool use, so the AI can read files, search, and operate on your workspace.

Everything is config-driven. One JSON file defines your sections, parsers, tools, and layout. No code changes needed to customize.

Key Features

• Zed-inspired 3-column layout — File tree | Overview | Chat, all resizable with drag handles
• 7 LLM providers, 21+ models — Anthropic, OpenAI, Google, Moonshot, xAI, Inception, OpenRouter. One adapter pattern, ~50-100 LOC each
• Tool use loop — AI can read files, list directories, search content (up to 10 tool iterations per message)
• Ontology-driven config — dashboard.config.json defines everything: sections, parsers, tools, layout
• Zed One Dark theme — Customizable via theme.config.json
• File browser with inline editor — Browse, preview, and edit files directly
• Git integration — Commit changes with auto-classified JSONL write-back
• Zero database — Pure filesystem. Markdown in, dashboard out
• Optional auth — JWT-based login, or run without auth for local use

Quick Start

git clone https://github.com/fstech-digital/ontology-dashboard.git
cd ontology-dashboard
npm install
cp .env.example .env
# Add your Anthropic API key to .env
npm run dev

Open http://localhost:3333. That's it.

The dashboard ships with demo data in demo/ so you can see it working immediately.

Configuration

dashboard.config.json

{
  "instance": {
    "name": "My Company",
    "title": "Ontology Dashboard"
  },
  "sections": [
    { "id": "welcome", "enabled": true },
    { "id": "heartbeat", "enabled": true },
    { "id": "projects", "enabled": true },
    { "id": "pipeline", "enabled": true },
    { "id": "gitlog", "enabled": true }
  ],
  "parsers": {
    "heartbeat": { "file": "HEARTBEAT.md" },
    "projects": { "dir": "projects" }
  },
  "tools": ["read_file", "list_directory", "ontology_search"]
}

Environment Variables

Variable	Required	Description
DASHBOARD_ANTHROPIC_KEY	Yes	Anthropic API key
DASHBOARD_OPENAI_KEY	No	OpenAI API key
DASHBOARD_GOOGLE_KEY	No	Google Gemini API key
DASHBOARD_MOONSHOT_KEY	No	Moonshot (Kimi) API key
DASHBOARD_XAI_KEY	No	xAI (Grok) API key
DASHBOARD_INCEPTION_KEY	No	Inception (Mercury) API key
DASHBOARD_OPENROUTER_KEY	No	OpenRouter API key
DASHBOARD_USER / DASHBOARD_PASS	No	Enable login auth
ONTOLOGY_PATH	No	Override workspace root

Architecture

┌─────────────────────────────────────────────────────┐
│                    Top Bar (48px)                    │
├──────────┬──────────────────────┬───────────────────┤
│          │                      │                   │
│  File    │     Overview         │    Chat Panel     │
│  Tree    │     (sections)       │    (SSE stream)   │
│  300px   │     flex: 1          │    380px          │
│          │                      │                   │
│  drag ↔  │  ┌──────────────┐    │  ┌─────────────┐  │
│          │  │ WelcomeCard  │    │  │ Model       │  │
│          │  │ Heartbeat    │    │  │ Selector    │  │
│          │  │ Projects     │    │  │             │  │
│          │  │ Pipeline     │    │  │ Messages    │  │
│          │  │ GitLog       │    │  │ + Tools     │  │
│          │  │ ...          │    │  │             │  │
│          │  └──────────────┘    │  └─────────────┘  │
└──────────┴──────────────────────┴───────────────────┘

Multi-LLM Adapter Pattern

Each provider is a thin adapter (~50-100 LOC) that normalizes to SSE:

chat.js (router, ~65 LOC)
  └─► getAdapter(providerId)
        ├─► anthropic.js   (native SDK stream)
        ├─► openai.js      (OpenAI SDK, also used by Moonshot/xAI/Inception)
        ├─► google.js      (@google/generative-ai SDK)
        ├─► moonshot.js    (OpenAI SDK + baseURL)
        ├─► xai.js         (OpenAI SDK + baseURL)
        ├─► inception.js   (OpenAI SDK + baseURL)
        └─► openrouter.js  (OpenAI SDK + baseURL)

All adapters emit the same SSE events: token, tool_use_start, tool_result, usage, done, error.

Section Registry

Sections are registered in src/lib/section-registry.js. To add a section:

1. Create the component in src/components/sections/
2. Register it with a unique ID
3. Add the ID to dashboard.config.json

Parsers

Parsers in src/lib/parsers.js read workspace files and extract structured data:

• parseHeartbeat — Counts [ ] and [x] checkboxes from Markdown
• parseSpecProgress — Tracks project completion from _specs.md files
• parseComercial — Aggregates JSONL logs into pipeline view
• parseCrons — Extracts automation tables from Markdown
• parsePinAgents — Reads agent hierarchy from Markdown tables
• aggregateFinanceiro — Sums JSONL financial records by month

Stack

Layer	Tech
Framework	Next.js 13.5 (Pages Router)
UI	MUI 5, Emotion
Charts	ApexCharts
Icons	Iconify (Tabler)
Chat	Server-Sent Events (SSE)
Auth	JWT (jsonwebtoken)
LLMs	Anthropic SDK, OpenAI SDK, Google GenAI SDK

Data Format

The dashboard expects your workspace to contain standard files:

• HEARTBEAT.md — Task list with - [ ] checkboxes
• projects/<name>/_specs.md — Project specs with checkbox progress
• pins/<name>.md — Operational pins with agent/cron tables
• logs/comercial/*.jsonl — Commercial activity logs

See the demo/ directory for examples.

Contributing

See CONTRIBUTING.md for architecture overview and development guide.

License

MIT — see LICENSE.

---

Built by FSTech Digital. Powered by the Operational Ontology framework.