System Requirements Document for retro-satquery

1. Introduction

The project "SatQuery X" is a production-ready Agentic Vision-Language Assistant developed for ISRO's Problem Statement 26167 – SatQuery AI. The system is designed to analyze satellite imagery using natural language, supporting various image types and formats. It aims to provide advanced AI-driven analysis and reporting capabilities for satellite data.

2. System Overview

SatQuery X is a sophisticated AI system that processes satellite imagery through natural language queries. It supports single optical/multispectral images, single SAR images, optical + SAR image pairs, and bi-temporal image pairs in GeoTIFF, TIFF, and PNG/JPEG formats (the latter for benchmark demos only). The system automatically selects the appropriate AI workflow based on user queries and includes a novel GeoReason Engine for enhanced analysis.

2a. Product Interpretation and Delivery Boundary

The system is delivered as a web application with a responsive frontend and a robust backend. It includes a set of REST APIs for image upload, validation, planning, analysis, and reporting. The application requires user authentication for accessing protected features like image uploads, analysis, and report generation. The project adheres to ISRO's requirements and constraints, focusing on advanced AI and data visualization.

2b. Source Content Inventory

Not applicable as no content source directive was provided.

2c. Page Content and Component Coverage

Landing

• Information/State: Introduction to SatQuery X, overview of capabilities.
• Primary Actions: Navigate to Sign Up or Login.
• Components: Hero section with generative art, brief system description.

Sign Up

• Information/State: User registration form.
• Primary Actions: Submit registration details.
• Components: Input fields for user information, submit button.

Login

• Information/State: User login form.
• Primary Actions: Authenticate user credentials.
• Components: Input fields for username and password, login button.

Upload

• Information/State: Image upload interface.
• Primary Actions: Upload satellite images.
• Components: File upload control, format guidelines.

Validation

• Information/State: Image validation results.
• Primary Actions: Proceed with validated images.
• Components: Validation status, error messages for incompatible images.

Dashboard

• Information/State: Analysis tools and results.
• Primary Actions: Interact with AI chat, view confidence meter, download reports.
• Components: Satellite map, layer control, image viewer, AI chat, confidence meter, detected objects, execution summary, models used, download report button.

Report

• Information/State: Generated PDF reports.
• Primary Actions: View and download reports.
• Components: Report list, download links.

Settings

• Information/State: User preferences and application settings.
• Primary Actions: Update settings.
• Components: Preference controls, save button.

3. Functional Requirements

• As an Analyst, I should be able to upload satellite images in supported formats to initiate analysis.
  Provenance: explicit
  Lifecycle: Upload page, image validation, analysis initiation.
  Acceptance: Successful upload and validation of images.

• As an Analyst, I should receive validation feedback on uploaded images to ensure compatibility.
  Provenance: explicit
  Lifecycle: Validation page, feedback on compatibility.
  Acceptance: Display of validation results and error messages for incompatible images.

• As an Analyst, I should be able to query the system using natural language to perform specific analyses on satellite images.
  Provenance: explicit
  Lifecycle: Dashboard interaction, AI workflow selection.
  Acceptance: Correct AI workflow execution based on query.

• As an Analyst, I should receive a downloadable PDF report summarizing the analysis results.
  Provenance: explicit
  Lifecycle: Report generation, download availability.
  Acceptance: Generation and download of a professional PDF report.

• As an Analyst, I should be able to view and interact with analysis results on a dashboard.
  Provenance: explicit
  Lifecycle: Dashboard display, user interaction.
  Acceptance: Display of analysis results and interactive components.

4. User Personas

• Analyst: Responsible for uploading satellite imagery, submitting natural-language analysis queries, reviewing results, and downloading reports.

5. Core User Flows

1. User Registration and Login

   • Sign Up: User accesses the Sign Up page, enters registration details, and submits the form.
   • Login: User accesses the Login page, enters credentials, and gains access to the application.

2. Image Upload and Validation

   • Upload: User navigates to the Upload page, selects images, and uploads them.
   • Validation: System validates images for compatibility, displaying results on the Validation page.

3. Analysis and Interaction

   • Dashboard: User accesses the Dashboard, submits queries, and interacts with analysis tools.
   • AI Workflow Execution: System selects and executes the appropriate AI workflow based on the query.

4. Report Generation and Download

   • Report: User accesses the Report page, views available reports, and downloads them.

6. Visuals Colors and Theme

• Muse: Refik Anadol
• Palette:
  • Background: #000000
  • Surface: #111111
  • Text: #FFFFFF
  • Primary: #00FFFF
  • Accent: #FF00FF
  • Muted: #333333
• Typography:
  • Headings: Space Grotesk, Bold, uppercase, tight tracking
  • Body: Archivo
  • Scale: 1.414 modular, 48/34/24/18/14
• Shape Language: Full-bleed generative canvas hero with minimal overlay UI, thin luminous strokes.
• Layout: Full-width layouts with floating panels, minimal UI chrome, and generous spacing.

7. Signature Design Concept

A full-bleed generative canvas on the Landing page showcases fluid particle animations in electric cyan and magenta, with minimal overlay UI elements. This design emphasizes the high-tech, data-centric nature of SatQuery X, aligning with the project's focus on advanced AI and data visualization.

8. Interaction Model & Motion Direction

• Interaction Model: Animated
• Motion Tempo: Cinematic
• Hero Dimensionality: webgl
• Landing Hero Motion Brief: A full-bleed generative canvas visualizes data in real-time as flowing particles over a dark background. The motion is continuous and slow, with data-driven transitions and scroll-linked morphing.

9. Non-Functional Requirements

• Performance: The system must handle large satellite images efficiently and provide quick responses to user queries.
• Scalability: The architecture should support scaling to accommodate increased user load and data processing demands.
• Security: User data and uploaded images must be securely stored and processed, with appropriate authentication and authorization mechanisms.

10. Tech Stack

• Frontend: React, Next.js, TypeScript, Tailwind CSS, Leaflet Map
• Backend: Python, FastAPI, Async APIs
• AI Models: Llama 3.1, RemoteCLIP, GroundingDINO, ChangeCLIP, CDVQA, SEN12MS Encoder
• GIS: Rasterio, GDAL, GeoPandas
• Database: PostgreSQL, PostGIS
• Deployment: Docker, Ubuntu Linux

11. Assumptions and Constraints

• Assumptions:
  • Users have basic knowledge of satellite imagery and analysis.
  • The system will be used primarily by analysts within ISRO or related organizations.
• Constraints:
  • PNG/JPEG formats are only for benchmark demos.
  • Incompatible images must be rejected after validation.
  • Planner reasoning must not be exposed; only the execution trace may be output.

12. Glossary

• AI Workflow: A sequence of AI model executions tailored to a specific user query.
• GeoReason Engine: A module responsible for cross-modal verification and evidence fusion.
• VQA: Visual Question Answering, a task where the system answers questions about images.
• CRS: Coordinate Reference System, a coordinate-based system used to locate geographical entities.