Blueprint Results Rail: Generate Prompt by Prompt

Version: 2.15.0, Date: September 4, 2026

Highlights

• A results rail on the run page: every prompt is listed up front and appears the moment it lands.
• One request per prompt: a failure no longer costs you the whole set.
• Full Set mode: produce the unified prompt and one per subtopic in a single generation.
• Categories for libraries: file your prompt libraries the same way you file everything else.

The blueprint results rail replaces the old single block of output on the Generate card with the layout the analyzers already use: a list of every prompt on the left, and the selected prompt’s content on the right. Generation now happens one prompt at a time, so results arrive as they are ready and a problem with one prompt leaves the rest untouched.

Blueprints

• New: The run page shows a results rail listing every prompt before any of them exist, each row selectable at any point, with its own state (queued, generating, completed or failed) and the subtopic description underneath.
• New: Generation runs one request per prompt instead of one request covering the whole set. Your cost is unchanged, because the engine always made a separate model call per subtopic. What changes is that each prompt appears as soon as it is finished, and a failure or a Stop keeps everything generated so far. The run count still increments once per generation, not once per prompt.
• New: Prompt, Output and Details tabs are reachable for every prompt. A prompt that has not arrived yet explains why on each tab rather than blocking you from looking.
• New: Details is grouped into General Info, Prompt Info and Run Info, and copy and export follow the same three sections.
• New: Full Set mode sits alongside Unified and Separate on the Topics card, producing the unified prompt and one prompt per subtopic in a single generation.
• New: Set actions in the rail, each showing a count: Regenerate Prompts, Generate Added Topics, Rerun Outputs and Run Added Topics. Generate Added Topics understands what changed: it generates newly ticked topics, drops the ones you unticked, and keeps every prompt and output that survives.
• New: A Stop control ends the run in progress, available in the rail and in the Prompt and Output panels.

• Improved: Run results now carry token counts, whether your own key was used, a timestamp and processing time, so Run Info tells you as much as Prompt Info does.
• Improved: Credits and tokens are never shown side by side. Each row reports whichever meter was actually used for that call, so there is nothing to reconcile.
• Improved: Subtopic descriptions are now sent through to the engine. They were being dropped before reaching it.
• Improved: The Blueprint editor’s action bar is split into a save row and a lifecycle row, so Publish and Delete are no longer mixed in with saving.
• Improved: The run page and both live previews default to two columns and remember the choice you make.

• Fixed: The run page showed a red Required badge on Topics without ever enforcing it. Topics is now genuinely required, so the badge means what it says.

Libraries

• New: Prompt libraries can be given categories. The category picker is on the save to library form, the library create page and library management.
• New: The category filter is on the Libraries page, next to search, and on the Prompts page next to the library selector. Prompts take the categories of the library they live in.
• New: Auto-write buttons on the library name and description, drafted from the context you have already written. A long main idea is summarized into a short name rather than cut off partway through.

Building Blocks

• Improved: The hand built topics window gains the Describe action the Topic Pack builder already had, filling in captions only for subtopics that do not have one.