Skip to main content
Glama

decken_compare_models

Compare VLM models on a slide image by running bounded diagnostic reconstructions and reporting whole-page MAE, native object counts, latency, and USD cost.

Instructions

Run bounded diagnostic reconstruction across explicit VLM modelIds on a slide image and compare wholePageMae, native object counts, latency, and USD cost.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "imagePath": "<imagePath>",
      "modelIds": "<modelIds>"
    },
    "name": "decken_compare_models"
  }
}

Input Schema

NameRequiredDescriptionDefault
modelIdsYesExplicit list of VLM model identifiers to compare.
imagePathYes
maxOutputTokensNo

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv1.1.0

TDQS

B3.2/5.0
Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full burden, and it does disclose real traits: the run is 'bounded', it targets specific modelIds, it consumes VLM inference (latency and USD cost are reported), and it performs actual reconstruction on the slide. It omits whether artifacts are persisted or written to disk, whether the operation is safe to repeat, and any cost/rate warning beyond implying cost is measured.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

A single dense sentence with the action front-loaded and the compared metrics trailing; nothing is wasted. Slightly dense jargon (wholePageMae, 'native object counts') compresses but does not obscure.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Listing the compared metrics partly substitutes for a missing output schema, which helps. However, for a multi-model inference tool with no annotations and only 33% parameter documentation, it should say what imagePath accepts, that runs can be costly/slow per model, and whether results are persisted.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is only 33%: only modelIds carries a description. The description reinforces that modelIds must be explicit but says nothing about imagePath format/resolution expectations or about maxOutputTokens (1024-16384), which is the parameter that actually enforces the 'bounded' claim. It fails to compensate for the coverage gap.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

Specific verb+resource: 'bounded diagnostic reconstruction across explicit VLM modelIds on a slide image' plus the explicit metric list it compares. An agent can tell this is the model-comparison/eval path rather than the plain reconstruct_slide path, but the description never names reconstruct_slide as the alternative, so differentiation is inferred from 'compare' and the metric list rather than stated.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The word 'diagnostic' frames usage as an evaluation/probing exercise rather than production reconstruction, and 'bounded' hints at capped runs. There is no explicit when-to-use, when-not, or named alternative among the sibling reconstruct/generate tools, so the guidance is only implied.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.