i6eal.de
DEEN
Start a project
Source-backed public-sector code intelligence

The AI Code Radar for German public administration

A permanent, machine-readable ledger of AI building blocks published in public openCode repositories. Every finding retains its GitLab project ID, commit SHA and exact file path.

official openCode sourceexact file evidenceno fuzzy joinsno generative AI
Data state 18 July 2026Observed since 18 July 2026Schema ki-code-radar/v1
The evidence boundary

Published code is observable. Deployment is not.

The radar records exact package dependencies and container images in public openCode repositories. It does not prove that a component is deployed, procured or used in production.

project ID + commit SHA + exact file path
Evidence snapshot

Which AI building blocks are visible in public-sector code?

Every mark resolves to an exact repository and source file. Descriptions help discover candidates, but never create evidence.

30repositories
187evidence rows
45technologies
157with an exact version

Observed over time

The first checkpoint starts the series. Later runs can reveal additions, removals and version changes.

The longitudinal series currently contains one verified checkpoint.

Source and citation

Built to be checked, cited and revisited

The public projection is deliberately bounded. Permanent dossiers retain the stable keys and direct source links required to reproduce every displayed claim.

Open official openCode source ↗

i6eal (2026): German public-sector AI Code Radar — 30 public repositories, 45 exact technology identities and 187 published code evidence rows, data state 18 July 2026. https://i6eal.de/en/tools/ki-code-radar/

Clearly explained

Does an evidence row prove that a technology is in production?
No. It proves only that an exact technology reference was published in a specific file at a specific commit. Deployment, procurement and productive use require separate sources.
How is a technology identified?
Through exact package names or container identities. Project descriptions can help discover candidates, but never become evidence.
Can repository names be joined to organizations in other datasets?
Not by name alone. The radar permits a join only when an explicit repository URL exists in the other source. Similar names are not merged.
What becomes valuable over time?
Repeated checkpoints can show first and last observation, additions, removals and exact version changes. The history starts only with observations the collector actually made.
Does the collector use generative AI?
No. Retrieval, candidate bounding, file parsing, identity resolution and publishing are deterministic and rule-based.

Need an auditable code evidence product for your field?

We build durable public-interest data products with exact identities, explicit limits and sources that remain inspectable.

Discuss a data projectExplore all tools
Related tools

Useful next steps

These tools complement the current result.

View all tools
MonitorGerman public-sector AI registerWhere does German public administration already use AI? Search published system profiles from federal, state and municipal bodies, with lifecycle, governance fields, stable identifiers and original evidence.IndexAI Procurement & Supplier IndexWhich AI contracts does the public sector award, who wins, and which technology dependencies are explicitly disclosed? Procurement updates daily from TED and oeffentlichevergabe.de; suppliers use exact source IDs only.MonitorPublic AI signal chronicleSee what actually changed across public AI data: new projects, updated evidence and corrections in an append-only chronicle with observation time, source identifier and direct link.