Midair

Midair keeps the quality record of one piece of software in one place. It runs the test suites, it analyses the sources, and it watches the machines that the software runs on. It keeps each result together with the configuration that produced it.

The record that holds all of that is a workspace. A workspace names the source repository, the test suite, and the machine that does the work. From a workspace you start three kinds of work:

• Testing — Midair checks out the test suite on a node and runs it. It brings the log back, and it turns the log into a result tree with a verdict for each test.

• Analysis — Midair deploys the analyser on a node and examines the sources. It reports defects, code metrics, the dependency manifest with the known advisories, the authors of the flagged code, and the deviations from a coding standard.

• Monitoring — an agent on a device reports the state of that device. Midair compares each report with the accepted baseline, and it lists the differences.

A fourth kind of work joins the first three into a loop. An attack vector names one avenue into the system, takes what the analyser found suspicious along it, asks an AI provider what to test, has the test suite written, runs it, reads the result, and starts again from what the round could not settle. See Attack vectors.

Each of the three keeps a history. Every start makes a new record, and the old records stay as they were.

What the screens show

The left side of the window holds one entry for each area:

• Dashboard — the state of the workspaces that are in scope: the map, and the counts for each area.

• Vision — one column for each workspace: the checks that you can start, and the items that ask for attention; the attack vectors under probe; and the vulnerability database. See Vision and Attack vectors.

• Monitoring — the devices and their differences from the baseline. See Monitoring.

• Analysis — the analysis runs, the analysis templates, and the analyser builds. See Analysis.

• Testing — the test runs, the test templates, the test suites, and the test environment builds. See Testing.

• Workspaces — the workspaces themselves. See Workspaces.

• Nodes — the machines that run the work. See Nodes.

• Users — the accounts and the teams. See Users, teams and access.

• Settings — the instance configuration and the integrations. See Settings.

The picker at the top of the window sets the scope. Each screen then shows one workspace, or all the workspaces that the account may see.

Ways in, other than the browser

• API tokens and the HTTP API — a personal API token with named scopes, and the HTTP API that it opens.

• Command-line client — midair-cli, which runs the work on the machine that holds the checkout and sends the result to the server.

• Continuous integration — the Jenkins plugin, the incoming webhooks, the status badges, and the instance state endpoint.

• MCP endpoint — the Model Context Protocol endpoint, which gives an AI assistant the same operations under one account.

Where to start

• First use of a new instance: Getting started.

• The words that this manual uses: Concepts and Glossary.

• Workspace configuration: Workspaces.

License

Copyright Interpretica, Unipessoal Lda. Write to sales@interpretica.io, or see https://interpretica.io.