andreagrandi / andreagrandi/draftomen

Publish a pinned and licensed narration model artifact

Aperta
#456 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub
draftomen enhancement size: S
Lingua principale
Python
Stelle
0
Fork
0
Merge medio
1h 23m
PR unite (30g)
181

Descrizione

## Problem

Users need an exact tested artifact, not an unversioned upstream download whose behavior or licensing may change silently.

## Proposed change

Implement the artifact-distribution decision from #454. Publish or reference the immutable tested files with model identifier/revision, format, quantization/build, expected sizes and checksums, compatibility identifier, license and attribution. Record provenance and reproducible preparation where applicable. Keep weights outside normal application downloads. No in-app downloader or automatic model updates in this slice.

## Acceptance Criteria

- [ ] AC1: An artifact retrieval check obtains the exact declared revision and verifies every required file's size and checksum.
- [ ] AC2: A manifest review finds identifier, revision, format, quantization/build, compatibility, sizes, hashes, and stable locations; no `latest` dependency exists.
- [ ] AC3: A licensing/provenance check confirms redistribution or upstream-fetch terms and all required notices accompany the chosen distribution.
- [ ] AC4: A baseline package-content check shows model weights are absent from the default application artifact.

## Implementation classification

- **Estimated size:** S
- **Orchestration risk:** High
- **Reason:** Artifact integrity, provenance, and immutable distribution are security and compatibility boundaries.

## Dependencies

Direct prerequisites: #454. Consumers: Download and verify narration artifacts on explicit request, Package the optional local inference component.

Guida per i contributori

Nessuna guida per i contributori indicizzata per questo repository

Valutazione

Questa issue non è ancora stata valutata.

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.