Skip to content

PROCESS.md

Latest commit

 

History

History
125 lines (90 loc) · 5.33 KB

PROCESS.md

File metadata and controls

125 lines (90 loc) · 5.33 KB

Project Structure

boston-dashboard/
├── backend/                  # Node.js Express backend
│   ├── server-with-mongodb.js # MongoDB-enabled server
│   ├── src                    # src code files 
│   ├── dist                  # java script compiled files (necessary!)
│   ├── .env                  # Environment variables
│   └── package.json          # Backend dependencies
│   └── package-lock.json          # Backend dependencies

├── frontend/                 # React TypeScript frontend
│   ├── src/
│   │   ├── components/       # React components
│   │   ├── services/         # API services
│   │   ├── types/            # TypeScript interfaces
│   │   └── App.tsx           # Main application component
│   ├── package.json          # Frontend dependencies
│   └── tsconfig.json         # TypeScript configuration
├── start-servers.js          # Script to start all servers (backend and frontend, assums mongoDB running on localhost:27017)

Design

architecture

Prompt Design and Considerations

System Prompt

You are an AI assistant for a Boston.ai logs dashboard. You help users analyze log data by converting natural language questions into MongoDB queries.

The MongoDB collection is called "logs", and documents follow this structure: { "_id": "ObjectId", "country_code": "US", "currency_code": "USD", "progress": { "SWITCH_INDEX": true, "TOTAL_RECORDS_IN_FEED": 16493, "TOTAL_JOBS_FAIL_INDEXED": 1521, "TOTAL_JOBS_IN_FEED": 13705, "TOTAL_JOBS_SENT_TO_ENRICH": 20, "TOTAL_JOBS_DONT_HAVE_METADATA": 2540, "TOTAL_JOBS_DONT_HAVE_METADATA_V2": 2568, "TOTAL_JOBS_SENT_TO_INDEX": 13686 }, "status": "completed", "timestamp": "2025-07-11T05:16:20.626Z", "transactionSourceName": "Deal4", "noCoordinatesCount": 160, "recordCount": 11118, "uniqueRefNumberCount": 9253 }

You can:

  1. Perform filters, aggregations, and calculations
  2. Use $match, $group, $project, etc.
  3. Include ISO 8601 format dates
  4. Always wrap aggregation pipelines inside an object with "aggregate": "logs", and "pipeline": [ ... ]

Only return a single valid JSON string that can be parsed using JSON.parse(). No extra text, no Markdown, no code blocks.

User Prompt

Generate a MongoDB query to answer this question: "..."

Return ONLY the query as a single valid JSON object that can be parsed with JSON.parse().

If the query requires aggregation, structure it like this: { "aggregate": "logs", "pipeline": [ { ... }, { ... } ] }

Do not include code blocks or explanations. Use ISO 8601 strings for dates. Ensure the JSON uses double quotes for all keys and values.

Considerations taken into account

Consideration Purpose
System Role Defined Establishes GPT-4's identity as a MongoDB-aware assistant
Full Field Schema Provided Ensures field names and structure match the expected documents
Strict JSON Output Required Guarantees compatibility with JSON.parse() in JavaScript
Aggregation Format Enforced Prevents incomplete pipelines or incorrect MongoDB structures
No Markdown/Code Block Output Avoids formatting issues from triple backticks or explanatory text
ISO 8601 Date Format Ensures compatibility with MongoDB date queries and JavaScript parsing
Double Quotes on Keys/Values Ensures valid JSON and avoids common syntax issues

Key Design Choices in Full-Stack Applications

(TypeScript + Node.js + MongoDB)

Design Choice Description
Separation Between Frontend and Backend Keep frontend (React) and backend (Node.js) separate in both code and deployment, allowing independent scaling and maintenance.
Environment-Based Config Use .env files or environment variables for managing different configurations (development, staging, production). Keeps sensitive data secure and environment-specific.
Stateless Backend Design Design APIs to be stateless when possible, making them easier to scale horizontally.
Centralized Error Handling Use global error handlers (middleware in Express) to standardize error responses.
Typed Contracts (TypeScript) Use TypeScript for strong typing and shared types between frontend and backend to reduce bugs and improve developer experience.
Secure MongoDB Access Avoid exposing the database directly; always go through controlled backend APIs. Use roles and environment-based URI configs.
Code Linting & Formatting Enforce consistent code style using tools like ESLint, Prettier, and Husky.

AI Tools

  • ChatGPT/Gemini for helping in generating src code.
  • OpenHands for helping in some features (rendering charts in the chat)