Skylark Drones - Monday.com BI Agent

A conversational Business Intelligence agent for founders and executives. It answers natural-language questions over live Monday.com data (Deals pipeline and Work Orders) with no cached responses.

Tech Stack

• Next.js 14 (App Router) - full-stack React with co-located API routes
• TypeScript - end-to-end type safety across tool schemas and normalized data
• OpenAI GPT-4o-mini - tool-calling loop for agentic BI queries
• Monday.com GraphQL API v2 - live data, no caching, exponential backoff on rate limits
• Tailwind CSS - dark-theme UI
• Vercel - zero-config deployment

Setup

1. Install dependencies

npm install

2. Configure environment variables

cp .env.local.example .env.local

Fill in .env.local:

MONDAY_API_TOKEN=your_monday_personal_api_token
DEALS_BOARD_ID=your_deals_board_numeric_id
WORK_ORDERS_BOARD_ID=your_work_orders_board_numeric_id
OPENAI_API_KEY=your_openai_api_key

3. Get your Monday.com API token

1. Log in to Monday.com, click your avatar, then open Administration.
2. Go to API and generate a Personal API Token.
3. Paste it as MONDAY_API_TOKEN.

4. Get board IDs

Open each board in Monday.com. The numeric ID is in the URL: monday.com/boards/1234567890

5. Run locally

npm run dev

Open http://localhost:3000

Deploy to Vercel

1. Push to GitHub.
2. Import the repo at vercel.com.
3. Add the 4 environment variables in Vercel project settings.
4. Deploy (Vercel auto-detects Next.js).

Architecture

User query
  |
  v
Next.js API Route (/api/chat)
  |
  +-> OpenAI tool-calling loop (max 8 iterations)
  |     |
  |     +-> get_board_items    -> Monday.com GraphQL (live)
  |     +-> search_board_items -> Monday.com GraphQL (live)
  |     +-> get_board_columns  -> Monday.com GraphQL (live)
  |     +-> get_board_groups   -> Monday.com GraphQL (live)
  |
  +-> lib/normalize.ts -> cleans INR currency, dates, statuses, nulls
  +-> lib/bi.ts        -> deterministic pipeline metrics + filter extraction
  |
  +-> SSE stream -> ChatInterface.tsx (traces + answer)

Sectors Supported

Mining, Powerline, Renewables, Railways, Construction, DSP, Aviation, Security and Surveillance, Tender, Manufacturing

Data Quality Handling

• INR currency parsing (masked values in crores/lakhs)
• ~52% of open deals have missing values and are surfaced as explicit caveats
• Date normalization across ISO, DD/MM/YYYY, MM-DD-YYYY formats
• Status/stage normalization (case-insensitive, trimmed)
• Null value counting reported in every tool trace