Orange Concierge - Sovereign Client Operations Copilot

Orange Concierge is a self-hosted AI copilot for Bitcoin consultants. It turns client notes and meeting transcripts into evidence-backed recommendations and review-ready action plans.

Before any model or embedding call, it scans for wallet seeds, private keys, API keys, recovery codes, and similar prohibited secrets.

What It Does

• Accepts consultant notes and client meeting transcripts.
• Rejects prohibited secrets before any model or embedding call.
• Extracts structured facts with supporting evidence.
• Retrieves relevant internal guidance through RAG.
• Generates cited recommendations for human review.
• Enforces role-based approval and audit rules in application code.

Stack

• Application: Next.js 16, React 19, TypeScript
• Data: PostgreSQL, pgvector, Drizzle ORM
• Auth: Better Auth
• AI: Ollama, Gemini
• Testing: Gherkin + Vitest-cucumber, Playwright, Testcontainers
• Infrastructure: Docker Compose, Nginx
• CI: GitHub Actions
• Workspace: pnpm workspaces

Project Structure

orange-concierge/
├── apps/
│   └── operations/              # Next.js operations app
├── packages/
│   ├── core/                    # Framework-free use cases, policies, domain rules, and ports
│   ├── infrastructure/          # Database, auth, audit, scanner, knowledge, and composition
│   └── ai/                      # Ollama/Gemini model adapters
│
├── knowledge/                   # Seeded Markdown playbooks for retrieval and citations
├── docs/                        # Project architecture deep dive
├── docker-compose*.yml          # Local/self-hosted service definitions
├── nginx.conf                   # Reverse proxy and rate-limit template
└── pnpm-workspace.yaml          # Workspace package boundaries

More details on the architecture: docs/architecture.md.

Run Locally With Ollama

Copy the development environment file:

cp .env.development.example .env

And run the following commands.

pnpm install
docker compose up -d db ollama ollama-pull
pnpm db:migrate
pnpm db:seed
pnpm knowledge:index
pnpm dev

Then Open: http://localhost:3000.

ollama-pull fetches the configured chat model plus nomic-embed-text for pnpm knowledge:index.

Run Locally With Gemini Cloud

Copy the development environment file and install dependencies:

cp .env.development.example .env
pnpm install

Edit your .env file:

AI_PROVIDER=gemini
AI_MODEL=gemini-1.5-flash
AI_API_KEY=your-gemini-api-key
AI_BASE_URL=https://generativelanguage.googleapis.com/v1beta

Then run Postgres and boot the app:

docker compose up -d db
pnpm db:migrate
pnpm db:seed
pnpm knowledge:index
pnpm dev

And Open: http://localhost:3000

Gemini extraction uses AI_MODEL; knowledge indexing uses text-embedding-004.

Useful Commands

pnpm typecheck
pnpm build
pnpm test:core
pnpm test:e2e
pnpm knowledge:index