Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/paramveer-cyber/Deployaar/llms.txt

Use this file to discover all available pages before exploring further.

Deployaar is a full-stack SaaS platform that eliminates the gap between a product idea and shipped code. You type a one-line feature request — “Add dark mode toggle to settings page” — and Deployaar’s AI pipeline handles everything else: validating the request, generating a structured product spec, spawning a sandboxed coding agent that reads your repository and writes the implementation, opening a pull request via your GitHub App, and running a depth-branched AI code review before a human ever touches a keyboard. It is built for teams that want AI to do the heavy lifting of routine feature work while keeping humans in control of the final merge decision.

The Pipeline

From intake gate to merged PR — seven automated steps powered by Inngest workflows and E2B sandboxed execution.

Coding Agent

An @inngest/agent-kit agent running inside an E2B sandbox with terminal, file-read, and file-write tools. Every tool call is checkpointed via step.run().

PR Review

Depth-branched AI code review at three levels — shallow, standard, and deep — giving you the right signal-to-noise ratio for every change.

GitHub Integration

A GitHub App (via Octokit) handles installations, branch creation, commits, pull request authoring, and webhook events from your repositories.

Key Features

Deployaar covers the entire software delivery lifecycle in a single platform. The features below are the building blocks of that pipeline.

End-to-End AI Pipeline

One feature request enters; a reviewed pull request exits. Every intermediate step — spec, tasks, code, PR, review — is handled by AI agents without manual handoffs.

Duplicate & Education Detection Gate

Before any work begins, the clarification service runs an AI similarity check against recent titles in the same project and rejects near-duplicates or “how do I…” education questions that aren’t real feature requests.

AI-Generated PRD

The prd service expands an accepted one-line request into a structured Product Requirements Document containing goals, scope, and acceptance criteria — all generated by the configured AI provider.

Sandboxed Coding Agent

The coding-agent service spins up an E2B sandbox and runs an @inngest/agent-kit agent. The agent loop is bounded by stopWhen: [isStepCount(MAX_ITERATIONS)] — never an unbounded loop.

Depth-Branched AI PR Review

Choose shallow (quick lint-level pass), standard (logic + style + tests), or deep (full repo-context analysis with cross-file impact) when triggering a review on any pull request.

GitHub App Integration

Deployaar authenticates as a GitHub App via Octokit and manages app installations, repository listings, branch creation, commits, and pull request lifecycle — all without user OAuth tokens for repository access.

Step-Checkpointed Inngest Workflows

Every workflow step — including individual tool calls inside the coding agent — is wrapped in step.run(). This means a Render restart or a transient failure mid-task does not re-run completed steps.

Multi-Provider AI

A single resolveAiModel() abstraction (via the Vercel AI SDK) routes to Anthropic, OpenAI, Google Generative AI, or DeepSeek. Switch providers and models with two environment variables.

Razorpay Billing

Per-plan usage limits are enforced by the billing service, backed by Razorpay for payment processing and webhook-driven subscription management.

Strict Four-Layer Architecture

Every feature is implemented through exactly four layers — DB queries → services → tRPC routes → frontend hooks — with no layer allowed to skip the one below it.

Tech Stack

Deployaar is a Turborepo monorepo with a clear separation of concerns across every layer of the stack.
LayerTechnology
Monorepo toolingTurborepo · pnpm workspaces
FrontendNext.js 16 (App Router) · React 19 · Tailwind CSS
API layertRPC · trpc-to-openapi · Scalar API reference
HTTP serverExpress 5
Authbetter-auth (Drizzle adapter)
DatabasePostgreSQL (Neon, serverless) · Drizzle ORM
Background jobsInngest (step-checkpointed workflows)
Sandboxed executionE2B
Agent framework@inngest/agent-kit
AI SDKVercel AI SDK (Anthropic · OpenAI · Google Generative AI · DeepSeek)
GitHub integrationGitHub App via Octokit
PaymentsRazorpay
UI componentsshadcn/ui
Deployment — webVercel (deployaar.paramveer.xyz)
Deployment — APIRender (deployaar.onrender.com)

Built for the ShipFlow AI Hackathon

Deployaar was built for the ShipFlow AI Hackathon (Jun 20–30, 2026). It is a fully functional, production-deployed platform — not a demo — with a live Vercel frontend, a Render-hosted Express API, and a Neon Postgres database.
The hackathon brief asked for an AI-powered workflow that meaningfully reduces manual engineering work. Deployaar answers that by automating the entire first-draft loop: from a plain-English request to a diff that a human engineer can review and merge in minutes rather than hours. The architecture is designed to be extended — swap AI providers, add new review depths, or wire up additional agent tools — without touching the core four-layer structure.
Ready to run Deployaar locally? Head to the Quickstart to be up and running in under ten minutes.

Build docs developers (and LLMs) love