Overview

DocCompile is a professional artifact compilation platform in progress. It started as a docs-as-code publisher and is evolving toward a processing layer that compiles messy source material into professional artifacts.

The original product formula:

Markdown
+ Mermaid
+ architecture icons
+ images
+ professional themes
-> PDF

The current product formula:

Source in. Contract satisfied. Artifact out.

The longer formulation: compile messy source material into professional artifacts that satisfy explicit contracts, remain grounded in evidence, and stay consistent across every output.

DocCompile should not compete for the place where documents live. It should compete for the moment where messy source material becomes a correct professional artifact.

Product evolution

1. Docs-as-code publisher. Portable Markdown, diagrams-as-code, professional themes, and PDF export, rendered locally in the browser.
2. SaaS contour. Authentication, admin control plane, credits, billing, and server-side AI transformations on AWS.
3. Strategic pivot. Deprioritize proprietary cloud workspace / document storage. Own the transformation, not the document.
4. Target Compiler platform. A Compiler is a transformation contract. A universal Harness executes that contract without per-artifact hardcoded orchestration.

Implemented vs target

Implemented

The working system includes a local-first Markdown renderer with Mermaid, architecture icons, themes, images, print layout, and PDF export; an AWS-deployed SPA and serverless backend; Cognito identity and an admin control plane; runtime / versioned prompts; credit-metered AI transformations; and a commercial contour around Paddle.

Working AI transformations currently exist for at least ADR, requirements (FR/NFR, business rules, constraints), and Resume / CV. The exact Compiler list, publication lifecycle, and billing mode (Live vs sandbox) should be confirmed against the product repository.

Target / planned

The next architecture is a universal Transformation Harness driven by Artifact Contracts, evidence / canonical facts, customizable output templates, Situations (multi-artifact packs), and later Patch Compile. These are not presented here as deployed behavior.

Conceptual flow

flowchart TB
    src["Messy source"]
    compiler["Compiler"]
    engine["Transformation engine"]
    validate["Validation / repair"]
    artifact["Professional artifact"]
    sinks["Markdown / YAML / Mermaid / PDF / Git"]

    src --> compiler
    compiler --> engine
    engine --> validate
    validate --> artifact
    artifact --> sinks

Validation, repair, Git-native sinks, and contract-driven compilation are the target Harness path. The implemented path today is local rendering plus per-transformation AI jobs.

Main actors

Actor	Role
Author / professional user	Writes or pastes Markdown locally; exports PDF; optionally requests an AI compilation
Admin / Super Admin	Runtime configuration, prompt versions, transformation observability, Compiler publication
Compiler	Transformation contract for one class of professional artifact
Harness	Target generic orchestration that executes a Compiler contract
LLM provider	Replaceable model implementation, not the product moat

Major capabilities

Capability	Status
GFM / Markdown rendering	Implemented
Mermaid and architecture icons	Implemented
Professional themes, print layout, PDF	Implemented
Local-first rendering; content stays in the browser by default	Implemented
Cognito authentication and RBAC	Implemented
Runtime / versioned prompts	Implemented
Credit-metered AI transformations	Implemented
Paddle billing contour	Implemented; Live vs sandbox to be confirmed
Transformation history / admin observability	Implemented; depth to be confirmed
Universal Harness / Artifact Contract	Target
Evidence / Canonical Fact Model	Target
Situations and Patch Compile	Target

Technology stack

Layer	Choice	Role
UI	Client-side SPA	Local rendering, editor, export
Delivery	CloudFront + private S3 origin / OAC	Static frontend
API	API Gateway + Lambda	Authenticated server-side operations
State	DynamoDB	Jobs, credits, runtime config, prompts
Identity	Amazon Cognito	Registration, authentication, roles
Infra	Terraform	Reproducible AWS environment
CI/CD	GitHub Actions + OIDC	Deploy without long-lived cloud credentials
Intelligence	External LLM API	Replaceable generation component
Billing	Paddle	Merchant-of-Record payments

SQS, WebSocket result notification, Step Functions, and ECS Fargate appear in the architecture direction; only currently deployed services should be treated as implemented.