Skip to main content
Glama
Cloto-dev

CPersona

Official
by Cloto-dev
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

calibrate_threshold

Calibrate vector search thresholds from random-pair cosine distributions, filtering unrelated memories and adapting similarity cutoffs to embedding models and corpus data.

Instructions

Auto-calibrate the vector search threshold from the null (random-pair) cosine distribution. Samples random memory pairs and places the threshold ABOVE the null mean so unrelated pairs are rejected. method='separation' (default) learns the operating point from two populations — null pairs vs temporally-adjacent same-session positives (nearest-neighbour fallback when too few exist); method='percentile' uses a quantile of the null distribution (robust to anisotropic models such as bge-m3); method='zscore' uses mean + z*std. No labels used, purely statistical. Adapts to both embedding model and corpus characteristics.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
methodNo'separation' (default; two-population — learns the operating point from null pairs vs temporally-adjacent same-session positives, falling back to nearest-neighbour when too few exist), 'percentile', or 'zscore'
agent_idYesAgent ID whose memories to sample
z_factorNoZ-score multiplier for method='zscore' (default: 1.0, higher = stricter)
percentileNoNull-distribution quantile for method='percentile' (default: 0.95, higher = stricter)
sample_sizeNoNumber of embeddings to sample (default: 200)
session_keyNoOpaque session identity you declare: a partition hint, not authentication and not a data filter. Selects which no-persist pause applies to this call. Omit to share one bucket with every caller that omits it. Full text on recall.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changedv2.6.1
    • addedInput schema / properties / session_key / maxLength
      Added value: +256
  2. Changed1 schema field changedv2.5.10
    • addedInput schema / properties / session_key
      Added value: +{
+  "default": "",
+  "description": "Opaque session identity you declare: a partition hint, not authentication and not a data filter. Selects which no-persist pause applies to this call. Omit to share one bucket with every caller that omits it. Full text on recall.",
+  "type": "string"
+}
  3. Changed1 schema field changedv2.5.4
    • addedInput schema / properties / method / enum
      Added value: +[
+  "separation",
+  "percentile",
+  "zscore"
+]
  4. Addedv2.5.2
  5. Removedv2.5.1
  6. Changed3 schema fields changedv2.4.34
    • addedInput schema / properties / method
      Added value: +{
+  "description": "'percentile' (default), 'zscore', or 'separation' (two-population, learns the operating point from null vs nearest-neighbour positives)",
+  "type": "string"
+}
    • addedInput schema / properties / percentile
      Added value: +{
+  "description": "Null-distribution quantile for method='percentile' (default: 0.95, higher = stricter)",
+  "type": "number"
+}
    • changedInput schema / properties / z_factor / description
      Previous value: -"Z-score multiplier (default: 1.0, higher = stricter)"New value: +"Z-score multiplier for method='zscore' (default: 1.0, higher = stricter)"
  7. First observedv0.1.0

TDQS

A3.6/5.0
Behavior3/5

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

Annotations declare readOnlyHint=false and non-idempotent, and the description says the calibration is 'purely statistical' with no labels, adding algorithmic transparency. However, it does not disclose operational behavior: whether the computed threshold is persisted to the agent, whether it overwrites a prior threshold, the cost of sampling 200 embeddings, or any auth requirement.

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?

The purpose is front-loaded in the first sentence, followed by method semantics. It is a bit dense with parenthetical detail, but nearly every clause (null mean placement, no-labels, model/corpus adaptation) carries information an agent needs.

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 6-parameter, no-output-schema tool, the description thoroughly explains the calibration algorithm and method choices, which is the core complexity. It leaves minor gaps (persistence of the resulting threshold, response shape), which keeps it from a 5.

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

Parameters4/5

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

Schema coverage is 100%, so the baseline is 3, but the description adds genuine value beyond the schema by explaining the tradeoffs between the method enum values (separation's two-population learning, percentile's robustness to anisotropic models like bge-m3, zscore's mean+z*std), which helps an agent pick the right value.

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?

The description states a specific verb and resource: 'Auto-calibrate the vector search threshold from the null (random-pair) cosine distribution.' An agent immediately understands the operation. It does not, however, name or contrast with any sibling tool (e.g. set_recall_precision), so it stops short of a 5.

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?

It gives guidance on which of the three methods to choose (percentile for anisotropic models like bge-m3), which is useful, but it never states when to call this tool versus alternatives such as set_recall_precision, nor any precondition like requiring existing memories. Usage is implied rather than explicit.

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