AI Engineering & Architecture Workspace

Sere AI

A high-performance academic and architectural AI workspace powered by a LangGraph cyclic state-machine engine, 9 self-healing subagents, two-tier cognitive memory, and high-fidelity visual rendering.

Project Overview

Sere AI is an ultra-lean, high-performance academic and architectural assistant engineered on a stateful cyclic LangGraph engine, coordinating 9 specialized subagents and 16 modular tools across a Planner → Generator → Verifier architecture with self-correction.

Built with a strict Zero-Fluff policy that delivers immediate, high-density structured answers with interactive Excalidraw whiteboards, verified Mermaid diagrams, Vega-Lite charts, LaTeX math, and responsive HTML components.

Implements a Two-Tier Cognitive Memory architecture: an auto-compressed background Cognitive Profile synthesizing multi-session user context, paired with a persistent Personal Identity store for zero prompt-window bloat.

Key Features

LangGraph StateGraph cyclic engine with START → agent ↔ tools → END autonomous loop
9 dedicated self-healing subagents with Planner → Generator → Verifier validation pipelines
16 modular tools covering Excalidraw, Mermaid, PPT decks, Quizzes, Syllabus Notes, and Assignment Solver
Two-Tier Cognitive Memory (Compressed Profile Biography + Persistent Personal Identity)
Interactive visualization suite: Excalidraw whiteboards, Mermaid diagrams, Vega-Lite charts, KaTeX LaTeX math
Streamdown custom rendering pipeline with live iframe previews and 16:9 slide decks
Zero-Fluff response engineering bypassing conversational preambles for immediate answers
Razorpay-powered INR credit economy with weekly auto-resets, referral bonuses, and atomic transaction locks