Core Concepts

Compiler Pipeline

How NovaServe's 4-stage compiler transforms TypeScript AST into deterministic cloud deployments via Nova IR.

NovaServe treats cloud infrastructure as a compilation target rather than a set of imperatively invoked API scripts. The compiler transforms your TypeScript source code into deterministic cloud deployments through a 4-stage pipeline.

Stage 1: AST Parser & Static Analysis

Input: TypeScript application source file (App.ts)
Mechanism: The parser scans the Abstract Syntax Tree using TypeScript compiler APIs, extracting all static invocations of NovaServe primitives (defineApp, api, storage, queue)
Output: Unlinked AST nodes representing application handlers and resource declarations
Performance: ~0.04 seconds for typical projects

Stage 2: Dependency Graph Engine & IAM Inference

Input: Parsed AST nodes
Mechanism:
1. Constructs a Directed Acyclic Graph (DAG) of resource dependencies
2. Analyzes handler function bodies for method invocations (e.g., uploads.put())
3. Synthesizes least-privilege IAM policies, mapping method calls to exact provider API actions without wildcards (s3:PutObject)
Output: Resource DAG + synthesized IAM policy JSON

Stage 3: Nova IR Generation

Input: Validated resource DAG & IAM specs
Mechanism: Serializes the graph into Nova Intermediate Representation (Nova IR) — a normalized, provider-neutral JSON format. Calculates a SHA-256 cryptographic checksum of the entire graph state.
Output: Nova IR JSON payload + SHA-256 state lock file (.nova/state.json)

Stage 4: Target Cloud Emitter

Input: Nova IR payload + target provider selection
Mechanism: Translates provider-neutral IR declarations into target-specific manifests (AWS Cloud Control API, Cloudflare API, or Docker build context)
Output: Deterministically provisioned cloud resources + updated state lock

Compilation Example

Given this application code:

App.ts

import { defineApp, api, storage } from "novaserve";

export const app = defineApp({ name: "my-app" });
export const uploads = storage("user-uploads");

export const upload = api.post("/upload", async (req) => {
  const data = await req.arrayBuffer();
  await uploads.put("file.bin", data);
  return { status: "uploaded" };
});

The compiler infers this IAM policy automatically:

Synthesized IAM Policy

{
  "Effect": "Allow",
  "Action": ["s3:PutObject"],
  "Resource": "arn:aws:s3:::user-uploads/*"
}

Why a Compiler Approach Matters

Eliminates duplicate config — resources are declared alongside handlers
Zero-trust security by default — compiler-driven IAM ensures exact permissions
Multi-cloud portability — Nova IR decouples logic from vendor APIs

Edit this page on GitHub

Stage 1: AST Parser & Static Analysis

Input: TypeScript application source file (App.ts)

Mechanism: The parser scans the Abstract Syntax Tree using TypeScript compiler APIs, extracting all static invocations of NovaServe primitives (defineApp, api, storage, queue)

Output: Unlinked AST nodes representing application handlers and resource declarations

Performance: ~0.04 seconds for typical projects

Stage 2: Dependency Graph Engine & IAM Inference

Input: Parsed AST nodes

Mechanism:

Constructs a Directed Acyclic Graph (DAG) of resource dependencies
Analyzes handler function bodies for method invocations (e.g., uploads.put())
Synthesizes least-privilege IAM policies, mapping method calls to exact provider API actions without wildcards (s3:PutObject)

Output: Resource DAG + synthesized IAM policy JSON

Stage 3: Nova IR Generation

Input: Validated resource DAG & IAM specs

Mechanism: Serializes the graph into Nova Intermediate Representation (Nova IR) — a normalized, provider-neutral JSON format. Calculates a SHA-256 cryptographic checksum of the entire graph state.

Output: Nova IR JSON payload + SHA-256 state lock file (.nova/state.json)

Compilation Example

Given this application code:

App.ts

import { defineApp, api, storage } from "novaserve";

export const app = defineApp({ name: "my-app" });
export const uploads = storage("user-uploads");

export const upload = api.post("/upload", async (req) => {
  const data = await req.arrayBuffer();
  await uploads.put("file.bin", data);
  return { status: "uploaded" };
});

The compiler infers this IAM policy automatically:

Synthesized IAM Policy

{
  "Effect": "Allow",
  "Action": ["s3:PutObject"],
  "Resource": "arn:aws:s3:::user-uploads/*"
}

Stage 1: AST Parser & Static Analysis

Stage 2: Dependency Graph Engine & IAM Inference

Stage 3: Nova IR Generation

Stage 4: Target Cloud Emitter

Compilation Example

Why a Compiler Approach Matters

Related

Stage 1: AST Parser & Static Analysis

Stage 2: Dependency Graph Engine & IAM Inference

Stage 3: Nova IR Generation

Stage 4: Target Cloud Emitter

Compilation Example

Why a Compiler Approach Matters

Related