Skip to main content
Glama
Sign In
deepset-ai

deepset-mcp

Official
by deepset-ai
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

get_experiment_run

Reads one scoring run: how its cells ended, a summary per Metric, then the grid rows.

Instructions

Reads one scoring run: how its cells ended, a summary per Metric, then the grid rows.

A scoring run is a Metric Group judging chosen Sessions of one pipeline. Find one with list_metric_groups (last_run) or list_experiments (a side's scoring_runs). It judges each chosen Session with each Evaluator; one cell per pair. counts says how the cells ended: errored is the Evaluator's own code breaking (see the cell's error_detail), not a low Metric; failed is a cell that never ran and can be retried. metrics summarises each Evaluator's Metric over every row, rows past max_rows included: mean, min and max for a Score, a count per label for a Label, and how often it did not apply. Each row is a Session with its cells; a cell's metrics carry the value per address (session, turn/2, ...) and the rationale the Evaluator wrote, which is what to group failures by. run.selection says whether the Sessions were SAMPLED or HAND_PICKED; only sampled runs say anything about traffic as a whole. :param pipeline_name: Name of the pipeline. :param metric_group_id: The id of the Metric Group the run scored with. :param run_id: The run's id. :param max_rows: Most rows to return; the summaries still cover every row. :returns: The run report, or an error message.

The output is automatically stored and can be referenced in other functions. Returns a formatted preview with an object ID (e.g., @obj_123). Use the object store tools in combination with the object ID to view nested properties of the object. Use the returned object ID to pass this result to other functions.

Input Schema

NameRequiredDescriptionDefault
run_idYes
max_rowsNo
pipeline_nameYes
metric_group_idYes

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed3 schema fields changedv0.1.31
    • removedInput schema / properties / experiment_id
      Removed value: -{
-  "title": "Experiment Id",
-  "type": "string"
-}
    • addedInput schema / properties / metric_group_id
      Added value: +{
+  "title": "Metric Group Id",
+  "type": "string"
+}
    • changedInput schema / required
      Previous value: -[
-  "pipeline_name",
-  "experiment_id",
-  "run_id"
-]New value: +[
+  "pipeline_name",
+  "metric_group_id",
+  "run_id"
+]
  2. Addedv0.1.29

TDQS

A4.1/5.0
Behavior4/5

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

With no annotations, the description carries the full burden and does so well: it distinguishes 'errored' (Evaluator code breaking, see error_detail) from 'failed' (never ran, retryable), explains that summaries cover rows past max_rows, and warns that only SAMPLED runs generalize to traffic. It omits any auth/permission or rate-limit context.

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?

Long but dense and front-loaded: the first sentence states the return shape, and each subsequent sentence earns its place by explaining semantics rather than restating the name. The trailing object-store/object-ID boilerplate is templated and slightly padded, but still operationally relevant.

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

Completeness4/5

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

For a complex, no-annotation, no-output-schema read tool this covers the important ground: return structure, error-vs-failure semantics, summarization scope, and the selection caveat. Missing only ancillary operational details such as permissions and payload size limits.

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

Parameters3/5

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

Schema description coverage is 0% across 4 params, so the description must compensate. It meaningfully clarifies max_rows ('Most rows to return; the summaries still cover every row') and gives terse docstrings for the other three, but adds no format or sourcing detail beyond the identifiers' identity.

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

Purpose5/5

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

States a precise verb+resource ('Reads one scoring run') and then enumerates the exact return shape — cell outcomes, per-Metric summary, grid rows. It also defines what a scoring run is, which lets an agent distinguish it from run-adjacent siblings like get_evaluation_try or get_pipeline_trace.

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

Usage Guidelines4/5

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

It explicitly routes the agent to the tools that produce the required identifiers ('Find one with list_metric_groups (last_run) or list_experiments (a side's scoring_runs)'), which is genuine when-to-use guidance. It stops short of stating when-not-to-use this tool or any exclusion conditions.

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

Deploy Server

Other Tools