Generation

Code generation is more than rendering a template string. Codepot treats it as an inspectable pipeline with explicit inputs, plans, file outcomes, and diagnostics.

General pipeline

Prototype generation

CodepotG derives a normalized context from OpenAPI and applies Jinja packs. It supports bundled languages, custom packs, dry runs, lifecycle policies, guarded refresh, commands, and reports.

codepotx generation

codepotx compiles authoring and template artifacts, builds a complete GenerationPlan, renders every virtual file, and then applies managed writes and manifest cleanup.

Generation depends on authoring and templating ports rather than concrete engine classes, allowing memory adapters and alternate source providers.

Determinism

A deterministic generator should produce the same plan and content from the same semantic inputs, template pack, variables, and configuration.

Stable digests and normalized artifacts make caching, comparisons, dry runs, and reviews possible.

Reports

A useful result includes more than success or failure:

• planned paths;
• created, updated, unchanged, skipped, refused, and deleted files;
• diagnostics;
• commands;
• cleanup decisions;
• rollback information;
• durations and counts.

Every frontend should present the same underlying result according to its user experience.