Settings

Configuration browser for your Caliper deployment. All values are read-only and reflect the live system state.

API & SDK

Endpoints and credentials for the Caliper ingestion SDK

API endpoint

https://caliper-xi.vercel.app/api

Demo API key

caliper_demo_key_public

Authentication

X-API-Key header

Rate limits

Unconfigured (demo)

// Example: track an event

await caliper.track({
user_id: 'user_abc',
event_name: 'buy_section_view',
experiment_id: 'hero_cta_test'
});

GitHub repository

Statistical Configuration

Thresholds and parameters used by the aggregator

Confidence level

95% (α = 0.05)

Test direction

Two-sided

SRM detection alphaStatsig/Eppo Std

0.0001

mSPRT prior tau (τ)

0.1

CUPED variance reduction

Enabled (auto-fit per experiment)

Bonferroni correction

Not applied (single-comparison)

Active methods:

Two-proportion z-testWelch's t-testχ² SRM detectionCUPED (Deng et al. 2013)mSPRT (Johari et al. 2015)

AI Readouts

Bedrock-powered narrative summaries of experiment results

Provider

Amazon Bedrock

Primary model

claude-haiku-4-5

Fallback model

amazon.nova-lite-v1:0

Region

us-east-1

Max tokens / readout

600

Temperature

0.3

Output format

Structured JSON (verdict, summary, recommendation, confidence)

Triggers

On-demand + auto on experiment stop

Readouts generatedLive

Data & Storage

Multi-database architecture for hot ingestion + warm analytics

Hot Store

DynamoDB

caliper-main

On-demand billing

NEW_AND_OLD_IMAGES streams

Single-table design (PK/SK + GSI1)

Warm Store

Aurora PostgreSQL

PostgreSQL 17.7

Serverless v2 (0.5–2 ACU)

SSL required

dbt-modeled mart layer

Region

us-east-1

Events ingested (lifetime)Live

31,054

User assignments (lifetime)Live

10,020

Analytics Pipeline

Scheduled dbt-core on AWS Lambda for segment-level analytics

Engine

dbt-core 1.8.0 + dbt-postgres

Deployment

AWS Lambda container image

Function

caliper-dbt-runner

Image registry

ECR (caliper-dbt)

ScheduleEventBridge

Every 15 minutes

Models

4 (staging × 2, intermediate, mart)

Tests

26 dbt tests + 1 custom assertion

Last successful runHealthy

1m ago

Model layers:

stg_eventsstg_assignmentsint_user_outcomesmart_segment_results

Real-Time Aggregator

Stream-driven Lambda computing statistics on every event batch

Function

caliper-aggregator

Runtime

Python 3.12 on arm64

Trigger

DynamoDB Streams (caliper-main)

Cold startPure Python Stats

~3ms

Layer

AWSSDKPandas-Python312-Arm64:27

Unit tests

33 (validated vs scipy reference)

Pure Python implementations:

• Normal CDF via math.erf (C library precision)

• Normal PPF via AS241 (Wichura 1988, accuracy 1e‑9)

• Regularized incomplete beta via Lentz's continued fraction

• Regularized lower incomplete gamma for χ² p-values

Caliper · Built for the H0 Hackathon · Track 2 (Monetizable B2B)