AI-assisted coordination dashboard for Canadian local food networks.
Connecting local farmers, market vendors, and Canadian buyers through intelligent supply and demand matching — powered by Backboard.io AI agents.

Farmesh is a lightweight coordination platform built to strengthen Canadian local food systems. It helps nearby farms and businesses coordinate produce supply and demand more efficiently — reducing food waste, helping Canadian farmers sell more of their harvest, and making it easier for local businesses to source fresh, nearby food.

Not a marketplace. Farmesh is an AI coordination layer with transparent reasoning — every match shows why it was proposed, what preferences were used, and what tradeoffs were made.

• Free-text or structured input — farmers and buyers can submit in plain language; AI agents parse and normalize it
• AI parsing preview — before saving, users see the structured interpretation of their input
• Exact & split matching — one vendor or two-vendor split fulfillment with scored proposals
• Product family substitutions — e.g. "salad greens" → baby greens, spinach, mixed greens, arugula
• Local-first matching — prioritizes nearby Canadian suppliers by province, city, and delivery radius
• Buyer memory — the Demand Interpretation Agent remembers preferences (e.g. organic preference, substitution policy)
• Transparent AI reasoning — every match shows parsed data, match rationale, memory used, locality logic, and tradeoffs
• Live coordination — status transitions from PROPOSED → CONFIRMED → FULFILLED with buyer/vendor notifications
• Admin timeline — visualizes every AI agent action and system event in real time

All AI orchestration is handled by Backboard.io agents:

This is a monorepo with separate frontend and backend packages.

farmesh/
├── frontend/                   # Next.js application
│   ├── app/                    # Next.js App Router — pages and API routes
│   │   ├── actions/            # Server actions
│   │   ├── api/                # API route handlers
│   │   ├── auth/               # Auth pages
│   │   ├── buyer/              # Buyer dashboard pages
│   │   ├── farmer/             # Farmer dashboard pages
│   │   ├── layout.tsx          # Root layout
│   │   └── page.tsx            # Landing / home page
│   ├── components/             # Shared UI components
│   ├── data/                   # Seed data (Canadian vendors, buyers, listings, requests)
│   ├── lib/                    # Utilities and service clients
│   │   ├── auth.ts             # Auth helpers
│   │   ├── db.ts               # Database access layer
│   │   ├── listings.ts         # Listing helpers
│   │   ├── matchingPipeline.ts # AI matching logic
│   │   ├── requests.ts         # Request helpers
│   │   ├── supabase.ts         # Supabase client
│   │   └── supabase/           # Supabase SSR helpers
│   ├── types/                  # TypeScript type definitions
│   ├── middleware.ts            # Role-based route protection (Supabase SSR)
│   ├── next.config.ts
│   ├── postcss.config.mjs
│   └── eslint.config.mjs
├── backend/                    # Backend services (AI agents, data, SQL)
│   ├── src/
│   │   └── agents/             # AI agent implementations
│   ├── data/                   # Backend seed / reference data
│   ├── sql/                    # SQL migrations and schema
│   └── README.md
├── package.json                # Root workspace package
└── package-lock.json

Listing statuses: OPEN → PARTIALLY_MATCHED → MATCHED → EXPIRED

Request statuses: OPEN → PARTIALLY_FILLED → FULFILLED → CANCELLED

Match statuses: PROPOSED → AWAITING_CONFIRMATION → CONFIRMED → REJECTED → FULFILLED

An Ontario restaurant requests 100 lbs of salad greens, organic preferred.

1. Buyer submits free-text request → Intake Agent parses it
2. Normalization Agent standardizes product, unit, and region
3. Demand Interpretation Agent reads buyer memory: organic preference: high, allows substitutions: yes
4. Matching Agent finds:
   • 🥬 Green Valley Farm (Ontario) — 60 lbs mixed baby greens
   • 🌿 Riverbend Produce (Ontario) — 40 lbs certified organic spinach
   • Proposes a split match from two nearby Ontario vendors with full explanation
5. Buyer confirms in the Buyer Dashboard
6. Coordination Agent updates all statuses, notifies both vendors
7. Admin Dashboard shows the live match, agent actions, and memory signals

Vendors (Ontario)

• Green Valley Farm
• Riverbend Produce
• Maple Leaf Organics

Buyers

• Toronto Bistro
• Waterloo Corner Grocer
• Hamilton Community Food Hub

git clone https://github.com/MizuPanda/farmesh.git
cd farmesh

npm install

Create a .env.local file inside the frontend/ directory:

# Supabase
NEXT_PUBLIC_SUPABASE_URL=your_supabase_project_url
NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY=your_supabase_anon_key

# Backboard.io
BACKBOARD_API_KEY=your_backboard_api_key
BACKBOARD_WORKFLOW_URL=your_backboard_workflow_endpoint

See the Keys document for full details on obtaining these values.

cd frontend
npm run dev

Open http://localhost:3000 in your browser.

• Currency: CAD
• Supported units: lb, kg, bunch, case, tray, basket
• Supported provinces: All Canadian provinces and territories
• Matching priority: Local and nearby suppliers first — province → city → delivery radius
• Buyer types: Independent restaurants, small grocery stores, food co-ops, cafes, caterers, community organizations
• Vendor types: Farms, market growers, greenhouse operators, orchard vendors, local produce sellers
• Seasonal produce examples: Leafy greens, root vegetables, berries, apples, herbs, greenhouse vegetables, onions

Private repository — all rights reserved.