Sovereignty Commitments
Relevant source files
The following files were used as context for generating this wiki page:
Sovereignty Commitments are the core design principles of Opificia AI. They ensure that organizations can run the software factory without depending on external infrastructure, specific AI models, or third-party permissions. These commitments serve as technical rules that the code must follow and against which the system is tested.
The project defines sovereignty through three primary questions: where the software runs, who holds the encryption keys/access to data, and which legal jurisdiction governs the operation. The commitments aim to provide the operator with total control over these answers.
Sources: SOVEREIGNTY.md:3-16, README.md:19-24
Core Architectural Commitments
The project adheres to eight specific commitments that dictate the architecture of the software factory.
Deployment and Connectivity
Opificia must be fully self-hostable and capable of operating in air-gapped environments. This means no feature requires internet access or services hosted by the project steward. Telemetry is disabled by default and, if enabled, only communicates with operator-defined endpoints.
The diagram shows that the core system functions entirely within a local or air-gapped network, with external connections being strictly optional and operator-controlled.
Sources: SOVEREIGNTY.md:21-36, website/index.html:105-115
Model and Data Control
The system remains model-agnostic by using open, documented interfaces. Open-weight models are treated as first-class citizens. All data, including code, prompts, and logs, remains in storage locations configured by the operator. The project does not collect data or use it for model training.
Sources: SOVEREIGNTY.md:38-54, publiccode.yml:29-37
Human-in-the-Loop Governance
By default, the system prevents any AI agent from merging pull requests or triggering deployments without recorded human approval. While operators can change this, the system logs such changes in an audit trail.
The sequence diagram illustrates the default requirement for human intervention before code is committed to the version control system.
Sources: SOVEREIGNTY.md:60-65, website/index.html:130-134
Technical Implementation and Testing
Each commitment includes a specific technical test to verify compliance during development.
| Commitment | Technical Test | Source |
|---|---|---|
| Runs Anywhere | Full feature set works with no outbound internet access. | SOVEREIGNTY.md:28-29 |
| No Phone-Home | Default installation makes zero network connections to unconfigured endpoints. | SOVEREIGNTY.md:35-36 |
| Model-Agnostic | Reference test suite passes using a local open-weight model. | SOVEREIGNTY.md:46-47 |
| Open Standards | All external core interfaces follow a published specification. | SOVEREIGNTY.md:53-54 |
| Data Sovereignty | Verification that no data is sent outside configured storage/destinations. | SOVEREIGNTY.md:58-59 |
| Human Decision | Default agent cannot merge/deploy without recorded human approval. | SOVEREIGNTY.md:64-65 |
| Supply Chain | Independent rebuilds produce identical artifacts (reproducible builds). | SOVEREIGNTY.md:70-71 |
| Reference Infrastructure | Public listing of providers and their headquarters' countries. | SOVEREIGNTY.md:76-77 |
Governance of Commitments
The Sovereignty Commitments are managed as policy files. Changes to these commitments require a formal Request for Comments (RFC) and approval from the steering team.
Change Process
- Proposal: Significant changes to
SOVEREIGNTY.mdrequire an RFC. - Review: The RFC remains open for comments for at least 14 days.
- Approval: Changes must be approved by the steering team (a subset of maintainers).
- Recording: Once accepted, commitments are recorded as Architecture Decision Records (ADRs).
Sources: GOVERNANCE.md:26-31, 39-41, SOVEREIGNTY.md:7-9
Development Constraints
While the project uses third-party AI tools (like Claude) for its own development, the resulting software must remain independent of these tools.
- No Tool Dependency: Opificia cannot require a specific AI provider to build or run.
- Assisted-by Trailers: Commits with AI help must include an
Assisted-bytrailer in the git history. - Confidentiality: No secrets or confidential project data are sent to third-party AI services.
Sources: SOVEREIGNTY.md:88-102, AGENTS.md:36-37
Summary
Sovereignty Commitments define Opificia AI as a self-contained, model-agnostic software factory. They prioritize air-gap capability, data privacy, and human control. These commitments are not just goals but are enforced through architectural requirements and automated tests, ensuring that users retain full control over their automated software development environment.