Skip to main content

Generate a clickable HTML approval page for reviewing UI fixes inside terminal-based AI agents.

Project description

ApprovalUI

License: MIT Python 3.10+ Tests PyPI

A tiny open-source loop for reviewing UI fixes inside terminal-based AI agents.

Terminal agents are great for code. They are bad at UI feedback, because UI feedback needs eyes, not paragraphs.

ApprovalUI renders a clickable HTML approval page from a simple JSON spec. You open it in a browser, tick approve or reject per item, generate a review, and paste it back into the agent.

Binary signal. Screenshot attached. No drift.

ApprovalUI demo


Why

When you iterate on frontend with a terminal agent, the feedback loop is verbal:

“The divider looks weird — can you make it full height?”

The agent reads text, not pixels. You end up describing the same fix five times.

ApprovalUI short-circuits that loop. The agent writes a structured spec of the fixes, you review them visually, and the page gives you a structured text block to paste back.

Works with Claude Code, Codex, Kimi, or any terminal agent that can write JSON and open a browser.


How it works

  1. The agent writes a JSON spec describing each UI fix: title, issue number, screenshot, root cause.
  2. approvalui turns the spec into a self-contained HTML page.
  3. You open the page, review each item, tick Approve or Reject, add comments.
  4. Click Generate review to get a structured text block you paste back into the agent.

Install

pip install approvalui

Or clone and install in editable mode:

git clone https://github.com/floomhq/approvalui.git
cd approvalui
pip install -e ".[dev]"

Quick start

approvalui example/fixes.json example/approval.html
open example/approval.html

Tick approve or reject, add a comment, and click Generate review.


JSON spec format

{
  "title": "Workeros — Fix Approval",
  "instructions": "Tick the ones you approve, click Generate review, paste it back to me.",
  "items": [
    {
      "id": 1218,
      "title": "Brain split-view divider now runs FULL HEIGHT to the bottom",
      "status": "pending",
      "root_cause": "wrapper used min-h-full so Collection height:100% could not resolve",
      "screenshot": "screenshot.svg"
    }
  ]
}

Only id and title are required. root_cause, screenshot, and instructions are optional.


Generated review format

[APPROVE] #1218 — Brain split-view divider now runs FULL HEIGHT to the bottom
[REJECT] #1219 — Sidebar scrolls horizontally on mobile — comment: still broken on iPhone SE

Documentation


Files

  • src/approvalui/ — the Python package.
  • example/ — sample spec, screenshot placeholder, and generated page.
  • SKILL.md — Floom skill instructions for agent integration.

Contributing

See CONTRIBUTING.md.


License

MIT. See LICENSE.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

approvalui-0.1.1.tar.gz (181.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

approvalui-0.1.1-py3-none-any.whl (8.5 kB view details)

Uploaded Python 3

File details

Details for the file approvalui-0.1.1.tar.gz.

File metadata

  • Download URL: approvalui-0.1.1.tar.gz
  • Upload date:
  • Size: 181.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for approvalui-0.1.1.tar.gz
Algorithm Hash digest
SHA256 71a1e5a1bf144eae592ec6477e6a9ec87664ad5dc4cfd8e677f461883dbf6b89
MD5 69b37cddadb3365caaa2157630a2ec2d
BLAKE2b-256 38fcab9a99b32162142a846bb573ba55258a7e195cc68db615659d408bc8191f

See more details on using hashes here.

File details

Details for the file approvalui-0.1.1-py3-none-any.whl.

File metadata

  • Download URL: approvalui-0.1.1-py3-none-any.whl
  • Upload date:
  • Size: 8.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for approvalui-0.1.1-py3-none-any.whl
Algorithm Hash digest
SHA256 0d91c462c931305f199464c7866adb49ade88141560d61acd7e72f1049c2bd2a
MD5 89a87a1949b032448a569c220c86423d
BLAKE2b-256 431efcb66872005138b81647441294f77c2c6e11a0ddc5e4a22f5410ef32601e

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page