Ask AI
Ask AI Start conversation ↵

I'm an AI assistant with Mewbo's codebase and documentation in context.

Ask me anything about Mewbo.

EXAMPLE QUESTIONS

Describe an app and build it

You build an app by describing it. The builder agent does the rest. One example runs through this page, a workspace of markdown trackers and a job-applications.csv under the intent Visualize all the tracker files in my workspace.


Creating an app

Open the Apps tab, describe your app, and pick a workspace. That is either the app's own isolated space or an existing project whose files the builder should read. A building card tails progress in the gallery and flips live within minutes.

Over REST, create with POST/api/apps and read state with GET/api/apps/{app_id}.


What the builder does

The diagram below is the sequence. What each step guarantees is the part worth reading.

  • Files match by pattern, never by name. New files of the same shape join on the next refresh.
  • Every document carries a natural key and a provenance field. A refresh updates documents instead of duplicating them.
  • The Streamlit frontend renders whatever the pipelines last wrote, read through the injected SDK.
  • Deterministic work becomes plain code with no model call. The declared refresh is a cron schedule, on_demand, or agentic mode where it needs model inference.
  • The dry run writes nothing durable, and the submit contract rejects an app that could never refresh.

The agent follows app-builder.md.

flowchart LR
    I([Intent]) --> E[Scan files] --> M[Design collections] --> F[Write frontend]
    F --> PP["Write pipelines\n+ schedule"] --> DR[Dry-run] --> S([Submit]) --> L([Live])

Use cases

The running example is the simplest shape, one view over a folder. Four others are worth knowing.

Tracker with a form. The one case where an app writes as well as reads.

A habit tracker where I log today's habits and see my streaks.

Scheduled digest with a model step. One bounded model call per item.

Digest my unread email into a checklist every morning, grouped by urgency.

On demand utility. Refreshes on a button, not a clock.

On demand, summarize the open issues in this repo grouped by area.

Sync plumbed through the CLI. Shelling out to git, tea, or gh stays plain code, and declares which vetted binary and remote host it needs.

Every 6 hours, sync commit history and open issues from the Gitea remote.


Writing a good intent

  • Name the data source. A folder, a CSV, your email, a repo.
  • Say what live means. Every morning, whenever the files change, or only when I ask. That phrase becomes the schedule.
  • Say if you want to enter data. That adds a form pipeline.

All three at once, Visualize all the tracker files in my workspace, refreshed whenever the files change.


→ Living with an App: health, refreshes, versions, repair.

→ Agentic Apps: the overview and the app-versus-widget distinction.