ICICLE Insights measures how the ICICLE institute's open-source work is used. It collects stars, forks, traffic, downloads, pulls and deployments from the platforms the institute publishes on. It keeps the history and shows it on a public dashboard.
- Dashboard: https://insights.pods.icicleai.tapis.io
- Inside TapisUI: https://icicleai.tapis.io/#/insights, under ICICLE Services → Insights
- API reference:
/docson any deployment, generated from/openapi.json
| Platform | What is tracked | Collected |
|---|---|---|
| GitHub | Repositories: stars, forks, watchers, 14-day views and clones | Yes |
| Hugging Face | Models and datasets: 30-day downloads, likes, lifetime downloads | Yes |
| Patra | Model cards and datasheets: deployments, card details | Yes |
| GHCR | Container images: 30-day pulls, lifetime pulls | Yes |
| npm | Packages | Listed only |
| PyPI | Packages | Listed only |
npm and PyPI are not collected on purpose. Their download counts cannot tell a person from a CI runner or a mirror refreshing its cache.
- Read the dashboard. Pick a platform and a time range, then drill into any resource. Start with Tour the dashboard.
- Run the catalog. Administrators sign in with a Tapis account to add accounts, resources, credentials and releases. Start with Administer Insights.
- Change the code. A Swift server and a SvelteKit dashboard. Start with Run Insights locally.
| Part | Technology |
|---|---|
| API and collectors | Swift 6.4, Vapor 4, Fluent |
| Storage | PostgreSQL 18 |
| Job queue and rate limits | Valkey 9 (any Redis-protocol server) |
| Dashboard | SvelteKit (Svelte 5), Tailwind 4, shadcn-svelte, built with Bun |
| Sign-in and secrets | Tapis tokens and Tapis Vault |
| Delivery | One container image, ghcr.io/icicle-ai/insights |
The image runs as three processes: the API, a queue worker and a scheduler. Architecture explains how they fit together.
cp .env.example .env # then fill in the Tapis values; see the tutorial
just web-install
just run # API on http://127.0.0.1:8080
just web # dashboard on http://localhost:5174This needs PostgreSQL and Valkey running locally. The full walkthrough is Run Insights locally.
Everything is under docs/, in four kinds:
- tutorials to learn by doing
- how-to guides for one task
- reference tables to look things up
- explanations of why it works the way it does
National Science Foundation (NSF) funded AI institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) (OAC 2112606).
See LICENSE.