SuperAPI AI Directory

d-inference

Private Inference Network on Idle Macs

Open website

About this tool

d-inference is an open-source repository maintained by Layr-Labs on GitHub, with the title and positioning of "Private Inference Network on Idle Macs".

Purpose and Positioning The repository hosts Darkbloom's Apple Silicon provider, native inference runtime integration, and marketing site, while retaining a snapshot of the console source code and its original build, test, and hosting configuration. The provider uses MLX to run models locally and can connect to the Darkbloom network or expose a local OpenAI-compatible API. This means it can be used both as a local inference endpoint and connected to a broader network environment.

Repository Ownership and Boundaries The materials clearly delineate code ownership: the centralized backend, console, and admin code belong exclusively to the darkbloom-platform repository, and new backend, console, or admin code should not be implemented, committed, pushed, or uploaded in this repository. The existing console-ui/ snapshot remains local, but retaining the snapshot does not mean that new development ownership has transferred back from platform. The device-local Swift API still belongs to provider code. Before pushing, check git remote -v and the expected diff. Source code ownership does not equal deployment authority: infrastructure, hosting, and traffic changes require specific human approval; releases and model registration remain external API operations.

Usage It is recommended to start the usage path with installation, then proceed to quickstart. Local inference uses direct mode; model adaptation requires reference to hardware requirements; cache storage is used for encrypted SSD storage and daily write limits. Tool versions are recorded in mise.toml, and supported commands are recorded in the Makefile. Building the provider requires Apple Silicon, the documented Xcode toolchain, and recursive MLX dependencies; for related rules, see the build, test, navigation, and contribution instructions.

Security and Limitations The provider is a plaintext inference endpoint. Network encryption does not apply to the process itself or the coordinator's temporary request handling.To hide prompts. Security assurances should refer to the encryption and storage trust model and provider attestation, and should not be inferred solely from repository separation. Public golden vectors are used to check fixed expected behavior, not as private cross-implementation qualification or real cluster evidence. The documentation index and retained console architecture are available for reference; platform APIs and consumer guides are maintained in the platform documentation. The existing changelog history is a record of work before and after the ownership split and does not constitute evidence of new releases or deployments.

License and Other For license information, see LICENSE, and for acknowledgements, see ACKNOWLEDGEMENTS.md. The published privacy and terms are not changed by source code cleanup. Specific pricing, rates, platform support scope, or performance data are not provided in the repository; this content is not yet available.