skills.rest
acking-you
acking-youCOMMUNITY
Explore all skills by this author →
GitHub

kiro-kmodel-calibrator

Recalibrate per-model Kmodel cache coefficients from recent successful usage samples.

Updated Aug 23, 2026
Search Tags:#calibration#normalization#data-analytics#sampling#kiro#kmodel#usage-events
One-click install
npx skills add https://github.com/acking-you/static_flow --skill kiro-kmodel-calibrator
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kiro-kmodel-calibrator
Source: https://github.com/acking-you/static_flow/tree/main/skills/kiro-kmodel-calibrator
Command: npx skills add https://github.com/acking-you/static_flow --skill kiro-kmodel-calibrator

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Recalibrate conservative Kiro cache-estimation coefficients using recent successful usage samples before applying admin updates to Kmodel settings.

Core Features & Use Cases

  • Reproduce offline calibration by aggregating llm_gateway_usage_events data with provider_type="kiro" and status_code=200 within the last 30 days.
  • Normalize model aliases and compute Tin, Tout, Cobs, and ratio to derive per-model Kmodel guidance (p50, p80, p90) for safe production planning.
  • Use case: when pricing, multipliers, or new models require a conservative starting Kmodel without auto-writing to live config.

Quick Start

Run the calibrator to generate per-model Kmodel recommendations from recent usage data, then manually apply the suggested Kmodel values in the admin UI after review.

Frequently Asked Questions about kiro-kmodel-calibrator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I recalibrate Kmodel coefficients from recent LLM usage events?▼

To recalibrate Kmodel coefficients, aggregate llm_gateway_usage_events data filtered by provider_type="kiro" and status_code=200 within the last 30 days to compute per-model p50, p80, p90 Kmodel recommendations.

What is Kmodel calibration for Kiro cache estimation?▼

Kmodel calibration is the process of recalibrating conservative Kiro cache-estimation coefficients. It normalizes model aliases and computes Tin, Tout, Cobs, and ratio to derive safe per-model Kmodel guidance for production planning.

How to calculate per-model p50, p80, and p90 Kmodel values from usage data?▼

Calculate per-model Kmodel values by aggregating 30 days of llm_gateway_usage_events with valid credits and status_code=200. This normalizes aliases and computes Tin, Tout, Cobs, and ratios to output p50, p80, p90 percentiles.

Does the Kmodel calibrator automatically update live configuration settings?▼

No, the Kmodel calibrator does not automatically write to live configuration. It outputs per-model Kmodel recommendations for manual review and application in the admin UI to ensure safe production planning.

When do I need to recalculate Kiro cache estimation coefficients?▼

You need to recalculate Kiro cache estimation coefficients when pricing, multipliers, or new models require a conservative starting Kmodel. This ensures cache estimates remain accurate before applying admin updates.

Related Skills

Explore More Data & Analytics Skills→
yangqiongyangqiong

Data Cleaning Skill

Cleans raw datasets by handling missing values, duplicates, and data types.

Community
Intermediate
frumu-aifrumu-ai
114

development-estimation

Generate software development effort estimates with P50/P80/P90 confidence ranges.

Official
Advanced
FrancisVargaFrancisVarga

w88 Data Model

Define w88 gaming platform event types, entity relationships, and join keys.

Community
Intermediate
alvarobarttalvarobartt
938

hf-mem

Estimate VRAM and parameter footprint for Safetensors-based Hugging Face models.

Community
Advanced