Evaluate - ScaleDown

/slm-agent:evaluate scans your project, writes a migration plan, and applies the changes to your code with your approval.

The migration report

Evaluation saves scaledown-report.md in your project root, following a fixed schema (version "1"):

Section	Content
Summary table	Files scanned, files with AI calls, total opportunities, providers detected
Complexity breakdown	Calls at each level, trivial to highly complex
Per-opportunity sections	Files, line numbers, complexity score, confidence and action for each of `sd_classify`, `sd_extract`, `sd_summarize`, `sd_compress`
Complex call decompositions	For calls scoring 3+/5, which sub-tasks can move to a ScaleDown model and which still need a frontier LLM
HTTP API reference	Request and response shapes for `/classify`, `/extract`, `/summarization/abstractive` and `/compress/raw/`

## Summary
| Metric | Value |
|---|---|
| Files scanned | 42 |
| Files with AI API calls | 7 |
| Total integration opportunities | 11 |
| Providers detected | openai, langchain |

MCP tools

Tool	Description
`get_ai_detection_patterns`	Regex patterns for all supported AI providers
`get_integration_template`	Before/after code showing how to integrate ScaleDown
`generate_migration_plan`	Structured markdown plan from a list of findings
`save_migration_report`	Writes the plan to `scaledown-report.md`

Analysis runs locally. The only network calls are your application’s runtime calls to ScaleDown, using the SCALEDOWN_API_KEY you provide.