Workspace

Open a factory to run work. App settings holds API keys and the same factory list for configuration. The Warehouse is every file that entered or left a factory.

Each chip is a configuration plane. Values live on the table named on that plane — not in a host file.

Faces and voices live on this key (and optional user). Any factory that asks for face or audio copies these files onto the job. Keywords let a brief say “Use Bob talking to Karen” without putting people on a catalog shelf.

Craft binds (style, rules) are catalog nodes. Source files live on the project. Only members of this project's active context are sent to chat and jobs. Unlisted files stay in the library and are out of scope.

Active context

Craft binds

Library nodes

One wizard for every factory. Factory, strategy, layers, and leaves come from the API. No product-specific branches in the UI.

1. 1 Factory
2. 2 Strategy
3. 3 Context
4. 4 Brief & files
5. 5 Submit

Launch gauges, strategy quality bands, and context grades are Postgres rows — not a vibe.

Process choices for this factory. Pick one to read stages and deliverables.

Layers come from the selected strategy map. The same layer can sit on many strategies.

Defaults and factory identity live on the factory row. Save writes Postgres — not a local preference file.

Issue manager or user keys. Types and privileges are Postgres rows — this screen only calls the API.

Suites are Postgres rows. Glance is last run and x/y pass/fail. Open a job to see the same page as every other job. Read-only — new cases are authored as DATA.

Modular node graph: canvas is React Flow; system of record is Postgres (df_pipeline_plan + edges + edge maps). Click a node to open that factory job. No JSON bags.

Operator handbooks are generated from live factory, package, and app catalog rows — same structure for every factory. The ? on a factory card opens this same handbook. Update documentation with Factory Builder Document factory (or Document package / Document UX app) before or after deploy.

ChatGPT-style threads (Postgres history). Ask questions, recall prior work via Memory (warehouse/jobs/chat by time + text), or produce multi-factory jobs through the pipeline node graph. Other factories share the same memory timeline.