IRIS Fieldwork

IRIS Fieldwork

IRIS Fieldwork is an original local operations workspace for InterSystems IRIS 2026.2. A responsive, dependency-free browser interface connects to a Python gateway; IRIS credentials remain in the gateway process. Live management APIs and an original Embedded Python/ObjectScript adapter supply the data.
Six views cover web applications, permissions, security and secrets, tasks, system telemetry, and logs. Operators can inspect configuration and nested records, enable or disable eligible web applications, review existing-resource grants on eligible roles, and run, suspend or resume verified user tasks. Wallet tools create collections and create, replace or remove write-only HTTP KeyValue secrets with explicit hosts and TLS. Audit and journal searches expose bounded record metadata with explicit pending-result refresh.
The interface begins in read-only mode. Enabled commands show a target-specific review, fresh-state checks and readback. Historical PID correlation is labelled as uncertain, and unavailable sources remain visible. Secret values are never fetched for display or verification.

Installation and walkthrough

Use the README installation instructions to prepare local credentials, start the IRIS container with Docker Compose, and run the Python gateway. Follow DEMO.md for the reproducible walkthrough and demonstration fixtures. The gateway and IRIS host port are bound to loopback by default.
Fresh Linux Docker Compose reproduction passed seven authenticated and four unauthenticated access checks; validation evidence is published. Docker Desktop on Windows and macOS has not been tested.

Published package-manager installation

Version 0.2.0 packages the complete gateway, UI and runtime adapter for Linux IRIS 2026.2 in USER. See the IPM installation guide. Installation creates no credentials or demo fixtures; the foreground launcher uses an existing account and binds only to loopback. Ownership checks refuse conflicting installations and changed files.
Version 0.2.0 was published to the public community IPM registry on 24 September 2026. A fresh consumer installed it solely from https://pm.community.intersystems.com using zpm "install iris-fieldwork -v", without a copied application source or test registry. Four authenticated requests, four anonymous denials and eleven gateway routes across all six sections passed; see community-registry verification. The UI still requires its separate foreground Python gateway and an existing authorized IRIS account. IPM publication is verified; the contest has not credited the IPM bonus. The older release note above records the earlier test-registry stage.

Scope and limitations

This is an AI-assisted development prototype for a disposable local instance. It uses an administrative account and is not a production security boundary. Resource creation, TLS/X.509/OAuth writes, process/device mutations and complete subsystem log coverage are not implemented. Stored secret values are not verified; replacement/removal cannot be undone here. Fresh-state checks do not lock out concurrent administrators.
The source is available under the MIT license, with a NOTICE. The v0.2.0 prerelease adds package-manager installation to the local prototype. InterSystems confirmed contest approval on 21 September 2026, and the entry is visible on the public contest roster. The official bonus results record 11 points. No prize or revenue is claimed.

Demonstration videos

• Overview — 2:58: a web-app enable/disable cycle with verified restoration and a completed bounded audit search; permission, wallet and task examples stop at review.
• Task lifecycle and triage — 2:22: verified suspension and resumption, one demonstration-task run with later completion evidence, and the limits of PID correlation. The initial unsuspended state and read-only workflow guard are restored.
• REST exploration and bounded journal reads — 2:14: a live journal-file inventory read, a reference-only write entry, and one bounded journal search that completes after refresh with two metadata records. Local filtering does not issue another search; global nodes and stored values remain withheld.
  All three videos provide English captions and show edited sequences of actual captured UI states using a disposable administrative test account. AI assistance and prototype limitations are disclosed. They document real local behavior. Video publication does not establish an awarded bonus or prize.