[v0.1] Environment:Model management automation

[rootfs]

Acceptance

• Automated model download/verification from HuggingFace
• version pinning; graceful fallback on missing models or revisions
• HuggingFace model uploading CI to ensure models are fully evaluated before overwriting existing models.

[lengrongfu]

@rootfs Hi, about this issue can provide more info, we want do something what? thanks

[rootfs]

@lengrongfu

The semantic-router project currently has manual and fragmented model management:

• Manual Model Downloads. Models are downloaded via in Makefile:

make download-models 

• Init containers in Kubernetes deployments. Manual hf download commands during development
• Hardcoded Model References.
• No Version Pinning: Current downloads use default branches (main) without specific version control

For this task, the following are expected:

Automated Model Download/Verification from HuggingFace:

• Automatic model discovery and download during application startup
• Model integrity verification (checksums, file completeness)
• Caching to avoid re-downloading existing models with the same version

Version Pinning with Graceful Fallback

• Configuration-driven version specification (commit hashes, tags)
• Fallback hierarchy: specific version → latest stable → main branch
• Version compatibility checking between models and application code
• Rollback capability when new versions cause issues

HuggingFace Model Uploading CI

• Automated model evaluation pipeline before publishing
• Model quality gates (accuracy thresholds, performance benchmarks)
• Safe model publishing that doesn't overwrite production models until validation passes
• Model versioning and tagging in HuggingFace Hub
• Rollback mechanism for problematic model releases