Runs evaluation and prints metrics.
Usage
mix arcana.eval.run
mix arcana.eval.run --mode hybrid
mix arcana.eval.run --generate --sample-size 50
mix arcana.eval.run --format json > results.jsonOptions
--mode- Search mode: semantic, fulltext, hybrid (default: semantic)--source-id- Limit to specific source--generate- Generate test cases first if none exist--sample-size- Sample size for generation (default: 50)--format- Output format: table, json (default: table)--fail-under- Exit 1 if recall@5 below threshold (for CI)
Configuration
Requires :repo to be configured in your application:
config :arcana, repo: MyApp.RepoIf using --generate, also requires :llm to be configured.