Skip to content

prism index

Build a compact, token-efficient index of all classes in an IRIS namespace.

Uses the IRIS %Dictionary SQL metadata tables to extract class structure without fetching source files. This lets AI agents understand a large IRIS codebase using a fraction of the tokens needed to read every document.

Usage

prism index [OPTIONS]

Options

Option Description
--namespace, -n IRIS namespace to index. Defaults to IRIS_NAMESPACE.
--system Include system classes (%Library, %SYS, %Api, etc.).
--prefix Only index classes starting with this prefix (e.g. MyApp).
--summary Only show counts (classes, methods, properties). No class details.

Examples

Quick overview

prism index --summary

{
  "namespace": "USER",
  "classes": 6480,
  "methods": 50036,
  "properties": 24456
}

Index all custom classes

prism index

Returns a JSON object with:

  • statistics: counts of classes, persistent classes, methods, properties, SQL procedures
  • classes: array of compact class summaries (name, super, properties, methods, parameters, SQL procedures)
  • dependencies: class → superclass mapping

Filter by prefix

prism index --prefix MyApp

Only includes classes whose name starts with MyApp.

Include system classes

prism index --system

Includes %Library.*, %SYS.*, %Api.* and other system classes.

Output shape

{
  "namespace": "USER",
  "statistics": {
    "classes": 179,
    "persistent": 12,
    "methods": 1061,
    "properties": 394,
    "sql_procedures": 9
  },
  "classes": [
    {
      "name": "MyApp.Model",
      "super": "%Persistent",
      "properties": {"Name": "%String", "Age": "%Integer"},
      "methods": {"Save": "%Status", "Load": "MyApp.Model"}
    }
  ],
  "dependencies": {
    "MyApp.Model": "%Persistent"
  }
}

Token efficiency

The index uses %Dictionary SQL metadata — the IRIS compiler's own metadata tables — so no source files are fetched. Benchmark against IPM (InterSystems Package Manager, 179 classes):

Metric Reading all source prism index Savings
Size ~345K tokens ~24K tokens 93%
API calls ~170 5 97%
Time ~30s 0.64s 47×