Model delivery

Build a Checksum Manifest for Model Downloads

Publish artifact identity, length, digest, format, revision, and source together.

Editorial review pending · Not indexed · Fact IDs: COVE-MODEL-01, COVE-MODEL-02

Direct answer

Publish a single manifest that bundles the artifact identity, its length, its digest, its format, its revision, and its source, so that every download carries its own recorded description and reviewers do not need to guess what the file is supposed to be. Stay strictly within these six descriptors and avoid reading anything else into them.

Evidence to collect

The claim becomes reviewable only when the following evidence is attached to the same artifact and test run:

  • Pinned artifact URL
  • Exact length and digest
  • Manifest verification test

A completed byte stream is not enough. Final placement, integrity, runtime load, and recovery must also pass.

Implementation workflow

Record the pinned artifact URL alongside the exact length and digest, and run a manifest verification test so that the two recorded values are compared against the file actually received, rather than against assumptions about the source or the publisher.

  1. Pin the official artifact revision.
  2. Record expected length and trusted digest.
  3. Check storage for temporary and update copies.
  4. Download to a temporary path and resume safely.
  5. Verify before final placement.
  6. Keep rollback and repair paths.

Failure patterns to prevent

  • Publishing a digest without revision
  • Treating length as integrity

Publishing a digest without the matching revision, or treating the stated length as proof of integrity, leaves the manifest incomplete, so any release statement should be limited to what these recorded evidence items actually show.

Minimum reproducibility record

LayerRecord
DeviceManufacturer, model, chipset, RAM class, operating-system build
SoftwareApplication version and git commit
ModelFamily, variant, revision, format, file length, hash, quantization
RuntimeName, revision, requested backend, observed backend evidence
WorkloadFixture revision, input hash, prompt hash, output policy
OutcomeCompleted, failed, cancelled, fallback, and privacy-safe diagnostics

Release checklist

  • ☐ The primary query is answered without an unsupported number.
  • ☐ Every artifact and runtime is pinned.
  • ☐ The representative task and failure policy are explicit.
  • ☐ Lifecycle, cancellation, cleanup, and fallback are tested.
  • ☐ User-content and network boundaries are documented.
  • ☐ Result wording applies only to the recorded configuration.
  • ☐ The page links to raw method or evidence when results are added.

Sources and related evidence

Chinese deployment and troubleshooting content is organized in the .奇连 AI 端侧专题.