Repository navigation
Releases
Squipi Mac Preview v3
Download Squipi for Mac
Download Squipi-macOS-AppleSilicon.zip below. Unzip it, drag Squipi.app to Applications, and open it. Requires Apple Silicon and macOS 26 or later. This pre-release is signed with Developer ID and notarized by Apple.
The research worker downloads Python and dependencies on first use. Ollama and a Qwen model, or a Nebius key, need separate setup. The optional research pack downloads an isolated R runtime. No personal data, keys, or model weights are included.
Kumo Forecast uses a managed Python; Git, Apple Command Line Tools, internet, model weights and additional storage are required. Its setup has stubbed tests but has not been validated end to end on a fresh Mac. Kumo Relational and Tabular require manual setup.
The file named Squipi-macOS-yet-to-be-notarized-preview-v3.zip is the older preview. Use the notarized download above.
Try the synthetic household data
Download synthetic-households.csv below. It contains 48 fictional records. Attach it in Chat and ask: “Remove the name, phone and email columns, keeping the others. Save a cleaned file and create a Word report with descriptive statistics, labelled charts, key findings and limitations.”
Validation and source
Apple accepted the notarization submission. The app and the app extracted from the downloadable archive passed signature, staple and Gatekeeper checks on the development Mac.
A previous extracted v3 app was run on the development Mac with a fresh isolated worker environment and a local scripted model substitute: it profiled 48 synthetic household rows, removed the requested personal columns, and generated a Word report with five embedded charts. Forecast installer and adapter tests passed with a stub model. The notarized download has not been tested on a separate clean Mac. Review research outputs before relying on them.
Source code is in Squipi-source-v3.zip and the repository. Squipi is MIT licensed; third-party software and models have their own terms.
Squipi Mac Preview v2
Download Squipi for Mac
Download Squipi-macOS-UNNOTARIZED-preview-v2.zip, unzip, and drag Squipi.app to Applications. Requires Apple Silicon and macOS 26 or later. This pre-release is ad-hoc signed, not notarized. If blocked, try opening it once and use System Settings → Privacy & Security → Open Anyway. Do not disable Gatekeeper.
The core research worker downloads Python and dependencies on first use. Ollama and the Qwen model or a Nebius key need separate setup. The optional research pack downloads an isolated R runtime. No personal data, keys, or model weights are included.
Kumo Forecast setup now uses a managed Python instead of requiring a preinstalled python.org copy; Git and Apple Command Line Tools, internet, model weights and additional storage are required. The setup path has stubbed tests but has not been validated end to end on a fresh Mac. Kumo Relational and Tabular currently require manual setup; see the repository documentation.
Validation and source
The extracted app was run on the development Mac with a fresh isolated worker environment and a local scripted model substitute: it profiled 48 synthetic household rows, removed the requested personal columns, and generated a Word report with five embedded charts. The original file was unchanged. Forecast installer and adapter tests passed with a stub model. We have not tested the GUI, Gatekeeper flow, local Qwen, or Kumo on a separate clean Mac. Review research outputs before relying on them.
Source code is in Squipi-source-v2.zip and the repository. Squipi is MIT licensed; third-party software and models have their own terms.
Squipi — Hackathon Preview and Source
Download Squipi for Mac
Download Squipi-macOS-UNNOTARIZED-preview.zip below, unzip it, drag Squipi.app into Applications, and open it. This is an ad-hoc signed preview and is not notarized by Apple. If macOS blocks it, try opening it once, then go to System Settings → Privacy & Security → Open Anyway and follow the prompts. Do not disable Gatekeeper.
Requires Apple Silicon and macOS 26 or later. The app itself does not require a build, but its research tools download Python and packages during setup. Local AI requires Ollama and a separately downloaded Qwen model; Nebius requires your own API key. Optional Kumo specialists require separate setup and model weights. No private research data, personal memory, API keys or model weights are included.
Source code
Download Squipi-MIT-source.zip or browse the repository for the source, documentation and tests. The app source is MIT licensed; third-party components and model weights have their own licences. The source release is not a notarized installer.
The downloadable preview was checked for a valid ad-hoc signature and archive integrity. It has not been tested on a separate clean Mac. Research results should be reviewed before use.