SudarshanAI | Knowledge Base
Engine Docs.
Everything you need to architect, simulate, and ship high-fidelity cloud infrastructure.
What is SudarshanAI?
SudarshanAI is an AI-powered cloud architecture engine that converts plain English system requirements into complete, production-ready infrastructure blueprints. Unlike general-purpose AI chat tools, SudarshanAI applies mathematical constraints (M/M/1 queueing theory) and cloud provider pricing APIs to derive architecturally grounded, cost-validated specifications — not guesses.
A typical blueprint includes: compute SKU selection and sizing, database engine and version, caching configuration, VPC and networking layout, IAM role structure, monthly cost estimate, P99 latency forecast, and OWASP security gap analysis.
The Three Modes
Terminal Mode
Describe your system requirements in plain English. The engine generates an optimized production-ready architecture with cost estimates and latency simulation.
Launch Terminal →Compiler Mode
You define hard constraints: exact QPS limits, latency ceilings, budget bounds. The engine mathematically derives the architecture that formally satisfies every constraint.
Launch Compiler →Deep Scanner Mode
Point the engine at any GitHub repository. It reverse-engineers the actual architecture via AST analysis, identifies security flaws, and recommends improvements.
Launch Deep Scanner →Tier System
| Feature | Free | Starter | Pro | Enterprise |
|---|---|---|---|---|
| Monthly blueprints | 5 | 25 | Unlimited | Unlimited |
| Compiler mode | ✓ | ✓ | ✓ | ✓ |
| Deep Scanner | ✓ | ✓ | ✓ | ✓ |
| Archive history | 7 days | 30 days | Unlimited | Unlimited |
| Share links | ✓ | ✓ | ✓ | ✓ |
| Private repo scan | — | — | — | ✓ |
| White-label | — | — | — | ✓ |
| Custom SLA | — | — | — | ✓ |
Architecture Insights
Deterministic Sizing
Our engine doesn't guess instance sizes. It calculates absolute minimum resource requirements using M/M/1 queueing models based on your declared Requests Per Second (RPS) and latency targets.
Cost Forecasting
The engine pulls live public pricing data for compute, storage, and networking transfer to generate a highly accurate Monthly Recurring Cost (MRC) estimate for your exact geographic deployment.
API Quotas and Engine Limits
To maintain deterministic performance, the SudarshanAI engine enforces strict boundary constraints on input parameters during generation.
| Parameter | Allowed Range | Description |
|---|---|---|
| Max QPS | 1 - 100,000 | Maximum acceptable queries per second baseline. |
| Object Storage | 0 - 10,000 GB | Initial Blob/S3 storage provisioning block. |
| Relational Dataset | 0.1 - 5,000 GB | Hard limit on single-instance RDBMS projection. |
| Expected Uptime | 99.0% - 99.999% | Determines cross-region failover replication triggers. |
| Repo Scan Depth | Max 1,000 Files | Scanner AST tree parse limit per invocation. |
Security & Data Privacy
How we handle your data
SudarshanAI is designed around a Zero-Retention intelligence model for business logic.
- Project descriptions are never used to train foundational AI models.
- Deep Scanner analyzes ASTs purely in memory and immediately drops the repo buffer. We do not cache your source code.
- Blueprints stored in Archive mode are encrypted at rest using AES-256 in our PostgreSQL instance.
- You can permanently delete any Blueprint from the Archive view, instantly destroying the generated JSON record.
Frequently Asked Questions
How does SudarshanAI prevent hallucinations?
SudarshanAI uses mathematically grounded M/M/1 queueing theory and deterministic SKU derivation. Rather than generating plausible-sounding architecture, it calculates required compute, storage, and networking from your specified QPS, latency SLA, and data volume constraints.
Is there a free tier?
Yes. Guest users can run one blueprint per mode (Terminal, Compiler, Scanner) without creating an account. Free account holders receive 5 blueprint generations per month, reset on the 1st of each month.
What cloud providers are supported?
AWS (full support), GCP (full support), and Azure (full support). Multi-cloud architectures spanning multiple providers in a single blueprint are supported in Compiler mode.
How accurate are the cost estimates?
Cost estimates are derived from current cloud provider public pricing data. Because cloud pricing changes and your actual usage patterns may vary (reserved instances, savings plans, data transfer costs), treat estimates as directionally accurate planning figures — not billing predictions.
Can I use SudarshanAI for system design interview prep?
Absolutely. The generated blueprints are excellent references for understanding how production systems are structured at scale. The architectural decisions, component choices, and failure mode discussions are directly applicable to system design interview questions.
Does the Deep Scanner work with private repositories?
Currently, the Deep Scanner works with public GitHub repositories. Private repository support requires a Pro or Enterprise account and GitHub App installation. This is available in the Enterprise tier.
What is the difference between Terminal and Compiler mode?
Terminal mode generates a production-ready architecture optimized for your requirements — think of it as the AI's best recommendation. Compiler mode applies strict mathematical constraints — you specify exact QPS limits, latency ceilings, and budget bounds, and the engine derives the architecture that formally satisfies those constraints.
How do blueprint share links work?
Share links generate a permanent, view-only public URL for any blueprint. Anyone with the link can view the full diagram, specifications, and cost table — no account required. Share links cannot be modified by viewers and do not expire.