Codepurify

Semantic metadata inference engine + template compiler for generating architecture artifacts from typed domain configs.

---

What is Codepurify?

• semantic metadata DSL
• normalized manifest compiler
• inference engine
• template-driven code generator

You define facts about your domain.

Codepurify infers:

• query capabilities
• mutation semantics
• relation groups
• workflows
• validation groups
• reusable template contexts

Templates decide final architecture output.

---

Core Philosophy

Configs define facts
        ↓
Inference engine derives semantics
        ↓
Normalized context is generated
        ↓
Handlebars templates render output

Codepurify does not hardcode:

• NestJS
• TypeORM
• FastAPI
• GraphQL
• React
• REST
• DTO patterns
• folder structures

All architecture styles are implemented through templates.

---

Features

• Strongly typed TypeScript configs
• Semantic metadata inference
• Handlebars-based generation
• Framework agnostic
• Runtime metadata compilation
• Query/mutation capability inference
• Typed enum transitions/workflows
• Semantic validation rule AST
• Relation graph inference
• JSON-safe normalized manifests
• Extensible template ecosystem

---

Example Use Cases

Generate:

• DTOs
• ORM entities
• repositories
• GraphQL schemas
• Zod schemas
• Pydantic models
• React forms
• OpenAPI specs
• validation layers
• constants
• metadata registries
• query builders
• admin panels
• SDKs

from a single semantic source of truth.

---

Installation

npm install codepurify

---

Quick Example

Entity Config

import {
  EntityConfigBase,
  stringField,
  enumField,
  query,
  mutation,
  transition,
} from "codepurify";

export default class UserEntityConfig extends EntityConfigBase {
  key = "user";

  fields = this.defineFields({
    email: stringField({
      length: 255,

      query: query().select().defaultSelect().search().sort().build(),

      mutation: mutation().apiWritable().build(),
    }),

    status: enumField(["active", "suspended", "deleted"] as const, {
      default: "active",
    }),
  });

  transitions = [
    transition({
      field: () => this.fields.status,

      initial: this.fields.status.values.active,

      terminal: [this.fields.status.values.deleted],

      transitions: {
        [this.fields.status.values.active]: [
          this.fields.status.values.suspended,
          this.fields.status.values.deleted,
        ],

        [this.fields.status.values.suspended]: [
          this.fields.status.values.active,
          this.fields.status.values.deleted,
        ],

        [this.fields.status.values.deleted]: [],
      },
    }),
  ];

  templates = [
    "dto.create",
    "dto.update",
    "typeorm.entity",
    "schema.zod",
  ] as const;
}

---

Semantic Inference

Codepurify automatically infers semantic groups from metadata.

You never manually define groups.

For example:

query().select().defaultSelect().search().build();

automatically contributes the field into:

entity.fields.query.select
entity.fields.query.default_select
entity.fields.query.search

Likewise:

mutation().apiWritable().immutableAfterCreate().build();

automatically contributes the field into:

entity.fields.mutation.api_create
entity.fields.mutation.immutable_after_create

---

Handlebars Templates

Codepurify uses Handlebars for rendering generated artifacts.

Templates receive a normalized semantic context.

---

Example Template

dto.create.hbs

export class Create{{entity.pascal_case_key}}Dto {
{{#each entity.fields.mutation.api_create}}
  {{snake_case_key}}!:
  {{typescript_type}};
{{/each}}
}

---

Example Generated Output

export class CreateUserDto {
  email!: string;
  status!: UserStatus;
}

---

Template Context

All exposed template variables are normalized to snake_case.

Example context:

{
  "entity": {
    "key": "user",

    "pascal_case_key": "User",

    "fields": {
      "all": [],

      "strings": [],

      "enums": [],

      "query": {
        "select": [],
        "default_select": [],
        "search": [],
        "sort": []
      },

      "mutation": {
        "api_create": [],
        "api_update": [],
        "immutable": [],
        "generated": []
      }
    },

    "relations": {
      "all": []
    },

    "checks": {
      "all": []
    },

    "indexes": {
      "all": []
    }
  }
}

---

Query Builder

Semantic query capabilities are defined fluently.

query().select().defaultSelect().sort().search().build();

---

Mutation Builder

Mutation semantics describe API/system behavior.

mutation().apiWritable().immutableAfterCreate().build();

Supported semantics include:

• api writable
• system writable
• readonly
• immutable
• immutable after create
• generated
• computed
• persisted

---

Relations

Relations are fully typed semantic metadata.

relationField(this, AppEntityConfig, {
  relation: {
    kind: "one_to_many",

    remote_field: () => new AppEntityConfig().fields.ownerId,

    cascade: true,
  },

  query: {
    select: true,
  },
});

---

Semantic Validation Rules

Checks are represented as semantic ASTs.

No raw SQL.

checks = [
  {
    name: "email_not_empty",

    rule: field(() => this.fields.email).notEmpty(),
  },
];

---

Workflows / Transitions

Transitions are defined independently from enum fields.

transition({
  field: () => this.fields.status,

  initial: this.fields.status.values.active,

  terminal: [this.fields.status.values.deleted],

  transitions: {
    active: ["suspended", "deleted"],
    suspended: ["active", "deleted"],
    deleted: [],
  },
});

---

Compiler Pipeline

TypeScript Configs
        ↓
Runtime Metadata Extraction
        ↓
Semantic Inference Engine
        ↓
Normalized Manifest
        ↓
Handlebars Template Compilation
        ↓
Generated Source Code

---

Design Goals

Codepurify SHOULD:

• infer semantic groups automatically
• remain architecture agnostic
• support multiple output ecosystems
• expose normalized template context
• prioritize metadata semantics over implementation details

Codepurify SHOULD NOT:

• hardcode framework architecture
• assume NestJS patterns
• assume ORM implementations
• force a specific folder structure
• expose raw database concerns directly

---

Long-Term Vision

Codepurify aims to become a universal semantic metadata compiler capable of generating:

• backend architectures
• frontend forms
• APIs
• SDKs
• validation layers
• schemas
• admin systems
• documentation
• infrastructure metadata

from a single semantic domain definition.

---

Example Future Ecosystem

codepurify
@codepurify/compiler
@codepurify/runtime
@codepurify/templates
@codepurify/typeorm
@codepurify/graphql
@codepurify/zod
@codepurify/react-form
@codepurify/openapi
@codepurify/fastapi

---

License

MIT