AeroPaper is an intelligent, real-time academic evaluation and question paper generation engine built using Next.js and Node.js. From securely archiving massive textbook materials to automatically processing flat handwritten notes, it eliminates token bottlenecks and hallucinated data patterns using automated asynchronous queues, dynamic multimodal vision handling, and strict AI-driven prompt grounding.
Compressed Binary Storage Layer
Compresses raw multi-megabyte PDF and document allocation buffers natively on the fly using zlib stream gzip pipelines right before saving into MongoDB, cutting cloud database space footprints by up to 70%.
On-Demand Memory-Isolated PDF Streaming
Enables teachers to scroll through chapters and view actual textbook pages smoothly in-app using high-performance client-side fetch buffers that dynamically decode Base64 data back into standard web iframe Blob object URLs.
Granular Text Extraction Engine
Extracts plaintext layers from readable standard .pdf, .txt, and .md files, using an optimized character-budget boundary guard that caps prompt grounding text blocks to safely stay under free-tier API rate caps.
Destructive Purge Controls
Features a fully integrated state-driven confirmation modal architecture that handles index removal requests safely with zero dependency on basic web browser prompt boxes.
Academic Domain Tag Filters
Sorts cataloged textbook manuals into responsive color-coded tag clusters (e.g., emerald for Science, blue for Mathematics) paired with interactive text search inputs for instant workspace filtering.
Strict Blueprint Configuration Alignment
Compiles complete assessment grids matching unique multi-section weight rules (e.g., Section A: 5 MCQs at 1 mark each, Section B: 3 Short Answers at 3 marks each) matching blueprint constraints flawlessly.
Automated Multimodal Sniffing Layer
Bypasses blank text parser data failures by automatically parsing flat scanned image PDFs or handwritten lecture photos as raw Base64 files directly into the AI's multimodal vision matrix for precise data anchoring.
Context-Anchored Zero-Hallucination Guardrails
Enforces a strict context-grounding framework that restricts the generation loop to exclusively use formulas, physics concepts, and definitions extracted from the linked grounding file.
Complete Grading Solutions Mapping
Generates a step-by-step master answer grid and resolution breakdown paired with corresponding question numbers instantly for immediate teacher verification.
Asynchronous BullMQ Task Processor
Offloads heavy content generation tasks from the express routing core into a separate background Redis pipeline thread, avoiding connection drops during long generation sequences.
Glassmorphic Interactive UI Controls
Designed card dashboards wrapped in smooth hover-state animations (backdrop-blur-md) that seamlessly slide specific functional buttons (View, Revise, Delete) into focus on the frontend canvas.
White-Label Printable Previews
Formats compiled exam JSON objects straight into clean, browser-ready print layouts that teachers can download as physical sheets with a single click.