Skip to main content

Documentation Index

Fetch the complete documentation index at: https://archie.com/docs/llms.txt

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

Specifications are what code generation reads. When you click Generate, Archie’s build agents read the entire specification set and produce frontend, backend, or full-stack code based on the stack you picked.

The whole spec set, one generation

Code generation is not run per spec type. There is no “generate just visual design” or “regenerate just functional design” mode. The agents need the full picture to generate consistent code. The split happens on a different axis: the stack you pick at generation time.
StackWhat it generatesReads from specs
Full-StackFrontend, backend, APIs, and database schemaAll spec groups (Functional, Visual, Technical)
Frontend-OnlyUI components, pages, client-side routingFunctional (user types, features) + Visual + Technical (strategy)
Backend-OnlyAPIs, services, database schema, server logicFunctional (features, requirements) + Technical (strategy, data model)

When edits affect code

Spec edits don’t touch existing code until you re-run generation. Editing the data model, adding a feature, swapping a UI framework — all of these queue up for the next generation pass. A few edits are exceptions and apply immediately:
  • Toggling a system service that has runtime configuration (e.g. enabling SMS) takes effect once the corresponding integration credentials are set
  • Changes in the Data Model editor commit to the database via tracked migrations rather than going through code generation

What gets regenerated vs. preserved

Archie tries hard to preserve hand edits in non-generated files. Edits inside generated files may be merged or overwritten depending on the change. The build view shows a diff before applying so you can review what’s about to happen. For protected work — custom logic in custom functions, hand-written tests, your own utility files — keep them outside the generated paths and Archie won’t touch them.

Generation phases

A typical full-stack generation moves through these phases:
  1. Queued — generation request accepted
  2. Initializing — agents load the spec set
  3. Configuring foundations — base config, dependencies, scaffolding
  4. Generating — agents write code module by module
  5. Building — the generated code compiles and tests run
  6. Completed (or Failed with a reason)
You watch progress live in the build view. If a phase fails, your specs and previous code remain intact — credits are only consumed for completed work.

FAQ

Not as a separate generation type, but you can edit one feature’s spec and run a regular generation — the agents focus changes on the affected files. Most generations are partial in practice.
Re-running generation from scratch is supported, but most builders find it faster to edit specs and run incremental generations. The exception is if your blueprint drifts far from your intent — then start from a refreshed blueprint.
No. Editing specs is free. Code generation is what consumes credits — the generation cost depends on the stack (Full-Stack, Frontend-Only, Backend-Only) and the size of your spec set.
Existing code stays intact. The new generation either completes or rolls back — you don’t end up with a half-applied diff. Open the activity feed to see why generation failed.