ChatMaestro — Screen Gallery

This gallery collects the 67 static screen mockups of the ChatMaestro experiment-orchestration platform.

How to Review the Screens

There are two ways to review the screens. You can open each screen as a static page — an HTML rendering or a PDF snapshot — or you can review the screens live in the mockup viewer, which is the preferred way.

Static pages. Click any card below to open that screen's static HTML page in a new browser tab, or click its PDF link for a fixed snapshot. A static page shows a single screen on its own, without the surrounding application.

The mockup viewer (preferred). The mockup viewer is a small application for viewing the mockup screens; it is not the ChatMaestro application itself. Within it you can:

navigate among and between the screens from a left-hand rail;
turn a screen's design annotations on and off;
search for a particular screen by name; and
run a slide show that steps through every screen in order (or only the screens matching a search), with the annotations kept on from one screen to the next.

Clicking a button or a tab performs the appropriate screen navigation, opens the appropriate pop-up modal dialog, or opens the appropriate side drawer, exactly as the real application would. The only difference is that no action is actually carried out, and all of the data on display is fictional.

Requirements. The mockup viewer requires JavaScript to be enabled. It is best viewed on a laptop or desktop display of roughly 1280 pixels wide or wider, although the interface is responsive and also adapts to tablet and phone form factors.

Open the mockup viewer

Auth & entry

Sign-in through an OAuth provider or an emailed one-time code; there are no passwords.

Enrollee landing

An enrollee's post-login list of the studies they have been assigned.

Enrollee experience

The chatbot an enrollee interacts with, kept deliberately minimal and blinded by default: a turned-back message (worded per the run's Socratic switch), Rewind to here on every exchange, Clear chat & start new, and the filling-up notice at 80% of the memory budget.

Enrollee preferences

An enrollee's profile, appearance, and notification settings.

Confirm · leave the study pool

The enrollee's own pool-wide opt-out: what ends, what stays, and how to rejoin.

Enrollee messages

An out-of-band message thread between an enrollee and the experimenter.

Attach text to a message

An enrollee attaches a plain-text file to a chat message when the run permits it; this is separate from the Documents library.

Home & authoring

A role-aware home page: experimenters see their runs, recent results, and spend, while administrators see an overview of experimenter activity.

Lists the experiments, which are reusable packages, together with the runs launched from each; a finished run is cleaned up or purged from its own row, without opening it.

Manages the reusable steering messages that a study varies across its runs; a nudge is locked once a run references it.

Experiment composer

Defines a reusable experiment — its prompts, enrollee-facing interface, scoring, and run defaults — in a modal used for both New and Edit (opens over the Experiments list).

Model configurations

Defines the reusable model side of an experiment: its constituent models, combine method, and self-improvement settings.

Builds cohorts and their membership, enforcing the disjoint-membership rule.

Launches a run, taking its nudge, model configuration, and retention from the experiment's defaults, then freezes the configuration and checks for overlap.

Experiment editor

Edits an existing experiment, pre-filled and marked as editing, with the admin-only ownership transfer on its own strip.

Edits an existing nudge, pre-filled; a nudge referenced by a run is locked.

Nudge detail (read-only)

Shows one nudge read-only, the state a locked nudge opens in; cloning it is the way to change it.

Model-configuration editor

Edits an existing model configuration — its members, weights and combine method — pre-filled and marked as editing.

Edits an existing cohort and its membership, pre-filled.

Monitor & operate

A master-detail view of the active runs and their enrollees.

One run's frozen composition — its condition snapshot, models, participants, and history — organized into tabs.

Observes a single enrollee and offers per-enrollee operations.

Experimenter messages

A broadcast composer alongside the experimenter's support inbox.

Results & analysis

Results and analytics

Presents per-factor scores, cross-run comparison, operational statistics, and natural-language querying.

Ask drawer, over results

The contextual Ask drawer, which returns a narrative answer with an accompanying table or chart, shows the generated SQL, and offers a partitioned rail of saved questions.

Ask engine, global

The same Ask engine without a scope, offering a browsable catalog of saved questions and cross-entity answers.

Saved questions (experimenter)

The experimenter's copy of the Saved-questions screen: their own drafts (edit, share, delete), drafts shared with them, and the team's approved and built-in questions, read-only.

Ask, fill parameters

A pop-up that collects the inputs for a parameterized saved question through a category drop-down, a dimension selector, and a scalar field.

Ask, save a question

The save-question dialog, where the user names the question and chooses, for each detected value, whether to parameterize it or keep it fixed.

Administration & system

Administrators manage experimenter accounts here, inviting them, enabling or disabling them, and transferring their data.

The administrator's roster of the shared enrollee pool, where enrollees are added, invited, and placed into cohorts; removal is blocked while an enrollee is in use.

Curates the installation's model menu, recording each model's aptitudes, benchmarks, and enabled state.

Scoring defaults

The read-only environment default for the trusted judge, copied into each new experiment.

A read-only log of every privileged action.

The administrator's triage board for bug reports and product feedback, a generic subsystem attached to the platform.

A read-only inspector for the embeddings and their registered sources.

The retrieval-augmented-generation library, showing each document's extraction and embedding status.

Saved questions (admin)

Governance for the Ask engine: every saved question at every trust level. Drafts are reviewed and approved with a note; a shared button is taken back by unapproving it or retiring it (undoable), and only a draft or a retired question can be deleted.

Resets or reseeds the demo database in test and demo environments, taking an internal safety snapshot first.

Email delivery settings, whether Supabase or SMTP, together with the invitation templates.

Catalog-model editor

Edits a catalog model: its identity, display name, aptitudes, expertise sentence, prices and enabled flag.

Enrollee editor

Edits one enrollee's account details from the enrollee roster.

Experimenter editor

Edits one staff account from the Experimenters screen, including its role.

Saved-question editor

Edits a saved question's label, prompt, parameters and notes; an owner opens it for their own draft, an admin for anything.

Document detail

Shows one document's metadata, extraction state and where it is used, read-only.

Dialogs & shared

The invitation dialog, with a template picker and a preview.

Creates a cohort by naming it and selecting enrollees from the shared pool.

New model configuration

Creates a model configuration from its constituent models, combine method, and self-improvement settings.

Creates or edits a nudge, capturing its name, steering text, and notes.

Registers a model in the installation's catalog, with its aptitudes, benchmarks, and enabled state.

Bulk-import a CSV, XLSX, or JSON file — validated row by row, then inserted and audited.

Add models (picker)

Picks models from the catalog to add to a configuration, hiding those already selected.

Add context files (picker)

Picks context documents to attach to a run, hiding those already attached.

See all responses

A viewer for the responses from multiple models, ranked and downloadable.

Exports run or results data, letting the user choose the format, scope, and columns.

Download or export

Downloads a chat transcript or exports run and results data, with a choice of format, scope, and filename.

Adds a document to the library by uploading a file — Word, Excel, or PDF, with scanned files sent through OCR — or by linking an https URL, then extracts and embeds it.

User preferences

Profile, appearance, notification, and table preferences.

A single audit entry, showing the event and its raw detail.

A single issue's detail, showing its metadata, telemetry, screenshot, and triage disposition.

A submission dialog for a bug report, capturing a summary, a severity, an automatic screenshot, and an attachment.

A submission dialog for product feedback — a like, a dislike, or an idea — with a note.

Confirm, clean up a run

Confirms a cleanup of a finished run, keeping its scorecard. Archive first is ticked by default and can be declined with a warning; a failed archive deletes nothing.

Confirm, reset the database

Confirms wiping the data tables, whether a reset or a reseed, taking a safety snapshot first.

Confirm, delete a definition

Confirms deleting a reusable definition, which is blocked while a live run uses it.

Confirm, remove an enrollee

Confirms soft-withdrawing an enrollee from a run, keeping their data.

Confirm, purge a run

Confirms the admin-only purge of a finished run, which removes the run and its scorecard. Archive first is ticked by default and can be declined with a warning; a failed archive deletes nothing.

Transfer ownership

Reassigns one item's owner — admin only, and blocked while the owner has a live run.

Transfer everything on offboarding

Reassigns every item owned by one experimenter in a single step, for offboarding.

Copyright © 2026 Agilis Software LLC. All rights reserved.