Skip to content

[EPIC]: Cross-Suite Model Downloader Service #2465

@sakshijha11

Description

@sakshijha11

Feature Description

Implement a shared downloader capability that:

  • Reads per-application model registry configuration (model name, source URL, target path, version metadata).
  • Downloads required model artifacts into runtime storage used by downstream services.
  • Skips already valid artifacts (idempotent behavior).
  • Reports per-model success, skip, and failure outcomes.
  • Fails fast with actionable logs for blocking issues.
  • Integrates with setup-stage orchestration flows (compose profiles, helm jobs, or pre-deploy hooks).
  • Supports release readiness checks and rollback procedures tied to model availability.

Business Value

  • Eliminates manual model copy across teams and environments.
  • Reduces deployment failures caused by missing or incorrect model files.
  • Standardizes model onboarding for new workloads across the suite.
  • Improves release reliability with one common gate for model readiness.

Scope

  • Standalone containerized downloader service reusable by all suite applications.
  • Config-driven model registry format per app and workload including URLs, paths, and versions.
  • Idempotent download behavior that skips existing valid artifacts.
  • Retry and error handling with clear model-level status output.
  • Setup-stage orchestration integration via compose or helm profile and job.
  • Release gate checks and rollback runbook tied to model availability.

Dependencies

  • ML and artifact teams publishing model URLs and versioned artifacts.
  • Environment access, proxy, and network policy support for artifact repositories.

Solution Description

Build a centralized, standalone model provisioning service that runs before application startup and prepares required AI model artifacts in a consistent, validated location.
The service will be configuration-driven, containerized, and reusable across all suite applications without app-specific code changes. It will support model-level status reporting, resilient download behavior, and deployment gate integration so releases are blocked when required models are unavailable or invalid.

Acceptance Criteria

Any suite application can provide a registry and use the same downloader image without code changes.
Service downloads configured artifacts to target storage and reports success, skip, and failure per model.
Service exits with non-zero code on blocking failures and provides actionable logs.
Setup orchestration can run the downloader before application startup.
Go and no-go release checklist includes verified model readiness evidence.
Rollback path is documented and tested for artifact and download failures.

Alternative Description

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    FEATUREFeature RequestTRIAGEIssues to be triaged

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions