Architecture-First Web IDE & Extension with Continuous AST Constraint Enforcement, Real-Time Risk Scoring, AWS Cost Governance, and Autonomous SRE Self-Healing.
Created by Team HackerStreet Rebels for the AWS AI for Bharat Hackathon (in partnership with Hack2Skill & YourStory).
Architects: Harshita Nagesh (@hash066), Sumanta Prusty (@sammy-20-wqz), and Aakash Nair (@Sky-Code06) Β· Team Leader: Laavanya Gupta
Modern AI-assisted code editors accelerate coding speed but lack architectural and operational context. Developers inadvertently create architectural drift, violate distributed system boundaries, generate unbudgeted cloud infrastructure, and push fragile "happy-path" code that fails under production scale.
DevForge bridges this "speed vs. stability" gap by making system architecture a first-class, active citizen across the entire software development lifecycle:
- Design Before Code: Teams define their system blueprint, layer boundaries, and performance budgets upfront.
- Real-Time AST Enforcement: As code is written, Tree-sitter parsers validate implementations against the approved architecture within 2 seconds.
- AWS Cost Whisperer: Live billing estimation via the AWS Price List API enforces interactive budget gates before code deployment.
-
Shift-Left Resiliency & SRE: Dual operating modes (Developer & Student) feature Chaos Monkey failure injection, automated LeetCode pattern mappers (
$O(n^2)$ loop detection), adversarial "Argue With Me" reviews, and an autonomous Kubernetes self-healing engine.
graph TD
%% Styling
classDef client fill:#e1f5fe,stroke:#01579b,stroke-width:2px;
classDef api fill:#fff9c4,stroke:#fbc02d,stroke-width:2px;
classDef compute fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px;
classDef engine fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px;
classDef ai fill:#ffebee,stroke:#c62828,stroke-width:2px;
classDef storage fill:#fff3e0,stroke:#ef6c00,stroke-width:2px;
subgraph Client_Layer ["Client Interfaces (Local & Browser)"]
direction TB
Ext["DevForge Extension<br/>(Kiro / VS Code)"]:::client
WebIDE["DevForge Web IDE<br/>(Monaco + React)"]:::client
Dash["Incident Command Center<br/>(Next.js 15)"]:::client
CLI["DevForge CLI / KubePilot<br/>(Rust & Python)"]:::client
end
subgraph Gateway_Layer ["API & Control Plane"]
ControlPlane["FastAPI Unified Control Plane<br/>(/v1/diagnose, /v1/k8s, WebSocket)"]:::api
APIGateway["AWS API Gateway<br/>(REST & WebSocket)"]:::api
end
subgraph Lambda_Layer ["Serverless Microservices (AWS Lambda)"]
direction TB
fnAnalyze["analyze_code"]:::compute
fnDrift["check_drift"]:::compute
fnRisk["calculate_risk"]:::compute
fnCost["estimate_cost"]:::compute
fnSecurity["security_scan"]:::compute
fnMentor["mentor_console"]:::compute
fnPatterns["detect_patterns"]:::compute
fnQuiz["generate_quiz"]:::compute
fnScale["predict_scale"]:::compute
end
subgraph Analysis_Engine ["Multi-Language AST Analysis Engine"]
direction TB
PyParser["Python AST Parser"]:::engine
JSParser["JS / TS AST Parser"]:::engine
RustParser["Rust Parser"]:::engine
SolParser["Solidity Parser"]:::engine
DepGraph["Dependency Graph & Cycle Detector"]:::engine
end
subgraph AI_Layer ["AI Orchestration Layer"]
Orchestrator["Multi-Agent Orchestrator"]:::ai
BedrockClient["AWS Bedrock Wrapper"]:::ai
Claude["Claude 3.5 Sonnet / Amazon Nova"]:::ai
Prompts["Agent Roles: Analyzer, Adversary, Critic, Mentor"]:::ai
end
subgraph SRE_Layer ["Autonomous Kubernetes SRE"]
Operator["K8s Operator & Detectors<br/>(CrashLoopBackOff, OOMKilled, Drift)"]:::compute
Remediator["Autonomous Remediator & Policy Gates"]:::compute
end
%% Wiring
Ext & WebIDE -->|Live Code & Blueprints| APIGateway
Dash & CLI -->|Incident & Control Streams| ControlPlane
APIGateway --> Lambda_Layer
Lambda_Layer --> Analysis_Engine
Lambda_Layer --> Orchestrator
ControlPlane --> SRE_Layer
Orchestrator --> BedrockClient --> Claude
DevForge/
βββ apps/
β βββ extension/ # DevForge Kiro / VS Code IDE Extension (React webviews, AST hooks)
β βββ web-ide/ # Standalone React Web-based Architecture IDE (Monaco Editor, Blueprint Canvas)
β βββ dashboard/ # Next.js 15 Live Incident & Health Command Center (WebSocket telemetry)
β βββ marketing/ # Next.js 15 Marketing & Product Showcase Site
βββ backend/
β βββ analysis-engine/ # Multi-language Tree-sitter AST & constraint rule engine (Py, JS, Rust, Sol)
β βββ ai-agents/ # Bedrock AI orchestrators, agent prompts (Analyzer, Adversary, Critic, Mentor)
β βββ lambdas/ # AWS Serverless Lambda functions (all 10 analysis & governance microservices)
βββ services/
β βββ control-plane/ # FastAPI unified control plane, AI RCA engine & persistence
β βββ cli/ # High-performance Rust CLI (`devforge wtf`, cluster inspection)
βββ operators/
β βββ k8s/ # Autonomous Kubernetes operator & incident remediation detectors
βββ kubepilot/ # Autonomous terminal remediation tool (stdlib Python CLI)
βββ packages/
β βββ core/ # Shared TypeScript models, contracts, and API clients
β βββ tokens/ # Design tokens, color palettes, and Tailwind theme presets
β βββ ui/ # Shared UI component library
βββ deploy/ # Helm charts, Kubernetes manifests, and cloud deployment scripts
βββ infra/
β βββ cdk/ # AWS CDK App Runner & serverless deployment infrastructure
βββ demo/ # Live failure simulations, broken workloads, and demo walk-through scripts
βββ sample-project/ # Microservices reference sample repository for drift testing
βββ release/ # Pre-packaged VS Code Extension VSIX binaries (`devforge-0.0.1.vsix`)
βββ docs/
β βββ specifications/ # Complete PRD (`requirements.md`), System Design (`design.md`), Tasks (`tasks.md`)
β βββ presentations/ # AWS AI for Bharat Hackathon official submission decks (PDFs)
β βββ skills/ # SRE diagnostic playbooks and incident response skills
β βββ PERFORMANCE_REPORT.md # Sub-2s AST validation & benchmark reports generated by Kiro
β βββ DEPLOY.md # End-to-end cloud and local deployment guides
βββ design.md # 2,000+ line Comprehensive System Architecture Design
βββ requirements.md # Product Requirements Document (FR-1 through FR-13)
βββ tasks.md # Phased implementation roadmap & task tracking
βββ PERFORMANCE_REPORT.md # Benchmarking analysis
βββ package.json # Turborepo root package definition
βββ pnpm-workspace.yaml # PNPM workspace definition
βββ turbo.json # High-performance build and test pipeline
- Live Blueprint Designer: Define service nodes, microservice boundaries, data stores, and protocols (gRPC, REST, WebSocket).
- Continuous AST Constraint Enforcement: Detects unauthorized inter-service calls, unencrypted communication, and data access bypasses directly at the editor keystroke level.
- Interactive Webview Modals: Includes
LiveMap,QuizModal,SecurityGateModal,RiskScoreCard, andArchitectureReport.
Synthesizes static code analysis, dependency cycles, and deployment metrics into an objective 0β10 risk score across four critical pillars:
-
Scalability Risk: Bottlenecks against declared throughput targets (
$O(n^2)$ loops, thread pool exhaustion). - Overengineering Risk: Identifies premature microservices and unnecessary complexity for low-traffic architectures.
- Security Compliance: Flags exposed endpoints, missing auth guards, and hardcoded credentials.
- Architectural Drift: Quantifies deviation between the design blueprint and actual implementation.
- Integrated with the AWS Price List API to dynamically compute projected monthly cloud spend based on declared resources (Lambda, ECS Fargate, DynamoDB, RDS, ElastiCache).
- Cost Gate System: Halts accidental over-provisioning by triggering interactive warnings if architectural modifications exceed user-defined monthly budget limits.
- Developer Mode (Production Teams):
- Scale Collapse Predictor: Simulates high concurrency to predict database pool exhaustion or message queue backpressure.
- Chaos Monkey Agent: Simulates network latency, node crashes, and dead-letter queues.
- Smart Contract & ML Pipeline Scanners: Domain-specific checks for Solidity reentrancy and ML data leakage.
- Student Mode (Education & Upskilling):
- LeetCode Pattern Mapper: Maps inefficient algorithms to relevant LeetCode problems for deliberate practice.
- Comprehension Validator: Triggers 1-question pop quizzes to ensure developers understand AI-generated code.
- Argue With Me & Anti-Vibe Critic: Adversarial AI that challenges architectural choices and flags code smells.
- Detects in-cluster incidents (
CrashLoopBackOff,OOMKilled,ImagePullBackOff, unschedulable pods, configuration drift). - Performs AI Root Cause Analysis (RCA) via AWS Bedrock / Claude 3.5 Sonnet / OpenAI.
- Applies least-privilege self-healing remediations under configurable risk and policy gates.
Extracted from the official validation suite (PERFORMANCE_REPORT.md):
| Metric | Target Specification | Measured Performance | Result |
|---|---|---|---|
| Real-Time AST Validation Latency |
|
π’ PASS (4.7x faster) | |
| Drift Detection Accuracy |
|
π’ PASS | |
| Risk Scoring Computation |
|
π’ PASS (4.2x faster) | |
| Cost Whisperer Price Sync |
|
π’ PASS | |
| Codebase Scalability Tested | Verified up to |
π’ PASS |
- Node.js:
v20.0.0or later - PNPM:
v9.0.0or later (npm i -g pnpm) - Python:
v3.11orv3.12 - Rust / Cargo: (Optional, for building the native CLI in
services/cli)
# Install all monorepo dependencies across apps and packages
pnpm install# Start all web apps concurrently (Dashboard, Marketing, Web IDE)
pnpm dev
# Or run individual apps:
pnpm --filter @devforge/dashboard dev # Next.js 15 Incident Dashboard
pnpm --filter @devforge/web-ide dev # Standalone Browser Web IDE
pnpm --filter @devforge/marketing dev # Marketing Sitecd services/control-plane
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\Activate.ps1
pip install -e .
python -m uvicorn app.main:app --reload --port 8000# Typecheck all TypeScript workspaces
pnpm typecheck
# Run monorepo unit and integration tests
pnpm test
# Build all production bundles
pnpm buildThe extension is pre-packaged and located in the release/ directory:
code --install-extension release/devforge-0.0.1.vsix- π Product Requirements Document (
requirements.md) β Exhaustive specifications for FR-1 through FR-13. - ποΈ System Architecture & Design Document (
design.md) β Complete 2,000+ line technical design and component models. - π οΈ Implementation Roadmap (
tasks.md) β Phased task breakdown and engineering milestones. - π Performance Report (
PERFORMANCE_REPORT.md) β Real-time AST validation latency & accuracy benchmarks. - βοΈ Deployment Guide (
docs/DEPLOY.md) β Cloud deployment instructions for AWS App Runner, ECS, and EKS. - π Hackathon Presentation Decks:
- Hackathon: AWS AI for Bharat Hackathon
- Innovation Partner: Hack2Skill
- Media Partner: YourStory
- Team: HackerStreet Rebels
- Architects:
- Harshita Nagesh (@hash066)
- Sumanta Prusty (@sammy-20-wqz)
- Aakash Nair (@Sky-Code06)
- Team Leader: Laavanya Gupta