Summary

Add a CLI binary (aprender) for model inspection, conversion, and quality checking. This mirrors the alimentar CLI for data tooling.

Motivation

Currently aprender is library-only. Users need to write Rust code to inspect models. A CLI would enable:

• Quick model inspection without code
• CI/CD integration for model validation
• Parity with alimentar data tooling CLI

Proposed Commands

aprender <COMMAND>

Commands:
  info      Display model information (architecture, params, size)
  convert   Convert between formats (apr, safetensors, gguf)
  inspect   Inspect layer weights and activations
  validate  Validate model integrity (no NaN, architecture valid)
  score     Quality score with letter grade (see #104)
  export    Export to deployment formats
  help      Print help

Examples

# Model info
aprender info model.apr
# Architecture: MLP [784, 128, 10]
# Parameters: 101,770
# Size: 407 KB
# Format: APR v1.2

# Convert formats
aprender convert model.apr --to safetensors -o model.safetensors

# Validate model
aprender validate model.apr
# ✓ Model loads successfully
# ✓ No NaN/Inf values
# ✓ Architecture valid

# Quality score (from #104)
aprender score model.apr --suggest

Implementation Notes

• Add [[bin]] target in Cargo.toml
• Use clap for argument parsing (like alimentar)
• Reuse existing library functions for inspection/conversion

Acceptance Criteria

• aprender info shows model metadata
• aprender convert supports apr/safetensors/gguf
• aprender validate checks model integrity
• aprender score implements Add model quality scoring system with improvement suggestions #104
• --json flag for CI/CD integration

Related

• Add model quality scoring system with improvement suggestions #104: Model quality scoring system
• alimentar CLI: Data tooling reference implementation

🤖 Generated with Claude Code