Glossary

A carbon-negative, high-performance blockchain platform used by JustKalm for Digital Product Passports. Features instant finality, low fees, and Proof-of-Stake consensus.

A standardized measure of user satisfaction with application response time. Scores range from 0 to 1, where 1 is perfect satisfaction. Calculated based on satisfied, tolerating, and frustrated user thresholds.

A unique secret identifier used to authenticate your requests to the JustKalm API. Must be kept confidential and never exposed in client-side code or version control.

curl -H 'Authorization: Bearer jk_live_xxxxx' https://api.justkalm.com/v1/valuate

A lightweight version of an iOS app (under 10MB) that provides specific functionality without full installation. Triggered by NFC, QR codes, or App Clip Codes.

A set of HTML attributes that add accessibility information to web elements. Helps screen readers understand dynamic content and custom widgets.

<button aria-label='Close dialog' aria-expanded='false'>

Greenhouse gas emissions prevented by choosing a lower-impact alternative. Buying secondhand instead of new typically avoids significant manufacturing emissions.

A statistical method that updates probability estimates as new evidence becomes available. Used in our confidence scoring to combine prior market knowledge with item-specific features.

An access token included in the Authorization header to authenticate API requests. Format: 'Authorization: Bearer <token>'. Our tokens use JWT format with embedded claims.

The gradual accumulation of substances (like heavy metals or PFAS) in an organism over time, faster than they can be broken down or excreted. Higher trophic levels show higher concentrations.

Security verification using unique physical characteristics like fingerprints (Touch ID), facial features (Face ID), or iris patterns. More secure and convenient than passwords.

A deployment strategy with two identical production environments. New versions deploy to the idle environment, then traffic switches instantaneously. Enables instant rollback.

A comprehensive certification that addresses environmental, health, and safety issues throughout the textile manufacturing process. More focused on production practices than end-product testing.

A classification of brands into value segments: Ultra-Luxury, Luxury, Premium, Contemporary, Mid-Market, Value, and Unclassified. Based on typical resale performance, price points, and brand equity.

Storing frequently accessed data in fast-access memory to reduce load on primary systems and improve response times. We use multi-layer caching for optimal performance.

The total greenhouse gas emissions caused by a product, expressed in CO2 equivalent. Includes manufacturing, transportation, use, and disposal phases.

California law providing consumers with rights over their personal information held by businesses. Includes right to know, delete, opt-out of sale, and non-discrimination.

A distributed network of servers that deliver content from locations geographically close to users. Reduces latency and improves load times for static assets.

The practice of intentionally injecting failures into systems to test resilience. 'Break things on purpose to learn how they break by accident.'

Automated pipelines that test, build, and deploy code changes. CI validates every commit; CD pushes validated changes to production automatically.

An economic model that eliminates waste by designing products for durability, reuse, and recycling. JustKalm enables circular economy by facilitating informed resale decisions.

A composite score (0-100) measuring the environmental sustainability of extending a product's lifecycle through resale, repair, or recycling. Incorporates material durability, repairability, and end-of-life recyclability.

Recent sales of similar items used as reference points for valuation. Our algorithm identifies the most relevant comps based on brand, condition, category, and recency.

A standardized rating of product condition: New With Tags (NWT), New Without Tags (NWOT), Like New (LN), Excellent (EX), Very Good (VG), Good (G), Fair (F), Poor (P). Condition significantly affects valuation.

A 0-100 score indicating how confident our model is in its prediction. Higher scores indicate more comparable data points, better feature coverage, and more reliable estimates.

A training approach where AI models are guided by a set of principles (a 'constitution') to produce helpful, harmless, and honest outputs. Developed by Anthropic.

The maximum amount of text (in tokens) that a language model can process in a single request. Larger context windows enable processing longer documents but increase latency and cost.

US law establishing safety standards for children's products, including limits on lead and phthalates. Relevant for resale of children's items.

A resampling technique where data is split into k folds, and the model is trained k times, each time using a different fold for validation. Provides more robust accuracy estimates than single holdout.

Requirements that data be stored and processed within specific geographic boundaries. JustKalm offers regional data residency options for enterprise customers.

The average number of days between listing and sale. Lower values indicate high demand; higher values may suggest overpricing or niche appeal.

URLs that navigate directly to specific content within a mobile app, bypassing the home screen. Essential for seamless web-to-app transitions and marketing campaigns.

// iOS: AppDelegate handles URL schemes
// Android: Intent filters in manifest

A model of how a product's value decreases over time. Different product categories, brands, and even specific items have unique depreciation patterns shaped by durability, trends, and scarcity.

A blockchain-anchored record containing verified product information including origin, materials, authenticity, ownership history, and environmental impact. Enables transparent provenance tracking.

Tracking requests across multiple services in a distributed system. Each request gets a unique trace ID, enabling end-to-end visibility of latency and errors.

Processing data geographically close to where it's needed, reducing latency and bandwidth costs. JustKalm uses edge locations in 15+ regions for optimal global performance.

A dense numerical vector representation of data (text, images) that captures semantic meaning in a high-dimensional space. Similar items have embeddings that are close together mathematically.

Chemicals that can interfere with the body's hormone (endocrine) system. May affect development, reproduction, neurological function, and metabolism. Common examples include BPA, phthalates, and certain pesticides.

Chemicals that interfere with the body's hormonal system. Found in some plastics, coatings, and textiles. Can affect development, reproduction, and metabolism at very low doses.

A machine learning approach that combines multiple models to produce better predictions than any single model. Our valuation engine uses an ensemble of gradient boosting, neural networks, and specialized category models.

The ability to understand and explain why a model made a particular prediction. We use SHAP values and feature importance to provide transparent explanations for valuations.

A policy approach where producers take responsibility for their products' entire lifecycle, including end-of-life management. Drives design for recyclability and take-back programs.

The automatic switching to a redundant system when the primary system fails. Enables high availability by minimizing service disruption during failures.

The estimated price at which a product would sell in an open market transaction between a willing buyer and seller, both with reasonable knowledge and no pressure to complete the transaction.

The process of converting raw product data (title, description, images) into structured numerical features that machine learning models can process. Combines rule-based extraction with neural network approaches.

A toggle that enables or disables features without code deployment. Allows gradual rollouts, A/B testing, and instant kill-switches for problematic features.

A measure of how much each input variable contributes to model predictions. Helps explain model behavior and identify which product attributes most affect valuation.

The process of taking a pre-trained model and further training it on domain-specific data. Our valuation models are fine-tuned on millions of authenticated resale transactions.

The transaction cost for executing operations on a blockchain network. Varies based on network congestion and computational complexity.

European Union regulation governing personal data protection and privacy. Grants individuals rights over their data including access, rectification, erasure, and data portability.

The world's leading textile processing standard for organic fibers. Includes ecological and social criteria across the entire textile supply chain. Requires minimum 70% organic fibers.

An ensemble machine learning technique that builds models sequentially, with each new model correcting errors of the previous ones. Powers XGBoost, LightGBM, and CatBoost.

A query language for APIs that allows clients to request exactly the data they need. JustKalm offers GraphQL for complex queries requiring multiple related resources.

When an AI model generates plausible-sounding but factually incorrect information. Critical to monitor in production systems. We use retrieval-augmented generation to minimize hallucinations.

A fixed-size fingerprint generated from data using a cryptographic function. Any change to the original data produces a completely different hash, enabling tamper detection.

A risk assessment score indicating potential exposure to harmful chemicals or materials. Scale: Low (0-25), Moderate (26-50), High (51-75), Critical (76-100). Based on material composition analysis and regulatory databases.

A validation technique where a portion of data is withheld from model training and used only for testing accuracy. Our holdout set is 20% of data, stratified by category and price range.

Adding more machines to handle increased load, rather than upgrading existing machines (vertical scaling). Enables near-linear capacity growth without service interruption.

A unique identifier you provide to ensure an API request is only processed once, even if retried. Prevents duplicate operations during network issues or timeouts.

fetch('/v1/valuate', { headers: { 'Idempotency-Key': crypto.randomUUID() } })

The process of using a trained model to make predictions on new data. In production, inference latency directly affects API response time.

Managing infrastructure through version-controlled configuration files rather than manual processes. Enables reproducible, auditable, and automated deployments.

A compact, URL-safe token format for securely transmitting information between parties. Contains encoded claims that can be verified and trusted.

A methodology for assessing environmental impacts throughout a product's life—from raw material extraction through manufacturing, use, and end-of-life disposal or recycling.

AI models trained on massive text datasets that can understand and generate human-like text. Used in JustKalm for natural language product descriptions and intelligent feature extraction.

The average absolute difference between predicted and actual values, expressed in dollars. Unlike MAPE, MAE treats all errors equally regardless of item value, making it more robust for items with varying price ranges.

A measure of prediction accuracy calculated as the average of absolute percentage errors between predicted and actual values. Lower is better. Industry standard for evaluating forecasting models.

The breakdown of materials in a product (e.g., 80% cotton, 20% polyester). Used for valuation accuracy, health risk assessment, and circularity scoring. Extracted from product descriptions and labels.

The degradation of model accuracy over time as real-world data patterns change. Can be caused by market shifts, seasonality, or changes in consumer behavior. We monitor for drift continuously and retrain models monthly.

The communication layer between JavaScript and native mobile code in hybrid apps. Enables calling device features (camera, GPS) from web-based interfaces.

A unique digital asset on a blockchain representing ownership of items like digital art, collectibles, or in our case, product authenticity certificates.

The ability to understand internal system state from external outputs (logs, metrics, traces). Enables proactive issue detection and rapid debugging in complex distributed systems.

An independent certification system for textiles that tests for harmful substances. OEKO-TEX Standard 100 is the most common certification, indicating products are free from harmful levels of 100+ substances.

An international testing and certification system for textile products. OEKO-TEX Standard 100 certifies products are free from harmful substances at levels that pose health risks.

The release of volatile organic compounds (VOCs) from materials over time. Common in plastics, foams, synthetic fabrics, and treated materials, especially when new or exposed to heat.

Data or transactions recorded directly on the blockchain, providing immutable, publicly verifiable records. On-chain data has the highest trust level but higher storage costs.

A specification for describing REST APIs in a machine-readable format. Our OpenAPI spec enables automatic SDK generation, documentation, and testing tools.

Percentile latency metrics. p50 is median response time (50% of requests are faster). p95 means 95% of requests complete within this time. p99 is the 99th percentile, critical for tail latency SLAs.

A class of synthetic chemicals also called 'forever chemicals' due to their environmental persistence. Associated with health concerns including cancer, thyroid disease, and immune dysfunction. Found in water-resistant textiles, non-stick coatings, and some cosmetics.

A group of synthetic chemicals called 'forever chemicals' due to extreme environmental persistence. Found in some textiles and coatings. Associated with health risks including immune suppression and cancer.

A group of chemicals used to soften plastics (plasticizers). Found in some clothing, toys, vinyl, and packaging. Associated with endocrine disruption, particularly concerning for children's products.

A measure of how demand changes in response to price changes. High elasticity means small price changes significantly affect sales volume.

A web application that uses modern web capabilities to deliver app-like experiences. Supports offline mode, push notifications, and home screen installation.

A security vulnerability where malicious instructions embedded in user input manipulate AI model behavior. Mitigated through input sanitization and system prompt isolation.

A blockchain consensus mechanism where validators are chosen based on staked tokens rather than computational work. More energy-efficient than Proof of Work by 99%+.

The complete documented history of an item's origin, ownership, and custody chain. Cryptographically verified provenance establishes authenticity and prevents counterfeit goods.

A statistical measure (0-1) indicating how well the model's predictions match the actual data. An R² of 0.95 means 95% of variance in prices is explained by our model.

An AI technique that grounds language model outputs in retrieved factual documents. Reduces hallucinations by providing the model with verified reference material.

The maximum number of API requests allowed per time period. Exceeding rate limits results in HTTP 429 (Too Many Requests) errors. Limits vary by plan tier and endpoint.

EU regulation addressing the production and use of chemical substances. Restricts or bans harmful chemicals in products sold in Europe.

EU regulation requiring companies to identify and manage chemical risks in products sold in Europe. Sets restrictions on substances of very high concern (SVHCs).

Deliberately testing AI systems by trying to make them fail or produce harmful outputs. Essential for identifying vulnerabilities before deployment.

The predicted price range at which a product can be resold in the secondary market, accounting for condition, demand, seasonality, and market trends.

An architectural style for web APIs using standard HTTP methods (GET, POST, PUT, DELETE). JustKalm's primary API follows REST principles with JSON request/response bodies.

The square root of the average of squared differences between predicted and actual values. Penalizes larger errors more heavily than MAE, making it sensitive to outliers.

Assistive technology that converts on-screen content to speech or braille. Used by blind and visually impaired users. Proper semantic HTML and ARIA labels are essential for compatibility.

Pre-built libraries for Python, JavaScript/TypeScript, Ruby, Go, PHP, Java, and Rust that simplify integration with the JustKalm API. Includes type definitions, automatic retries, and error handling.

import { JustKalm } from '@justkalm/sdk';
const client = new JustKalm({ apiKey: process.env.JUSTKALM_API_KEY });

The percentage of listed items that actually sell within a given time period. Higher rates indicate stronger demand and appropriate pricing.

A game-theoretic approach to explain model predictions. SHAP values show how each feature contributes to pushing the prediction from the baseline to the final output.

A formal commitment to service quality metrics including uptime percentage, response time, and support responsiveness. Breach of SLA may trigger credits or penalties.

Self-executing code stored on a blockchain that automatically enforces agreement terms when conditions are met. Used in JustKalm for material verification and authenticity attestations.

Self-executing code stored on a blockchain that automatically enforces agreement terms. Enables trustless transactions without intermediaries.

// Algorand PyTeal smart contract
@app.create
def create():
    return pt.Approve()

An auditing standard measuring how well a company safeguards customer data. SOC 2 Type II reports cover security, availability, processing integrity, confidentiality, and privacy over time.

The basic unit of text processing in language models. Roughly corresponds to word fragments. GPT-4 uses ~1.3 tokens per word on average. API pricing is per-token.

Our evaluation of potential chemical hazards based on product materials, manufacturing origin, brand practices, and regulatory databases. Provides a risk score, not a definitive safety determination.

Using a model pre-trained on a large dataset as a starting point for a new task. Dramatically reduces training time and data requirements.

The predicted range within which the true value is likely to fall, expressed as a min-max range. Based on Bayesian uncertainty quantification and historical prediction variance.

The percentage of time a service is operational and accessible. Measured as (Total Time - Downtime) / Total Time × 100. Does not include scheduled maintenance windows.

International standards for making web content accessible to people with disabilities. WCAG 2.1 AA is the common compliance target. Covers perceivability, operability, understandability, and robustness.

An HTTP callback that delivers real-time notifications to your server when events occur in JustKalm (e.g., batch processing complete, async valuation ready).

A model's ability to perform tasks it wasn't explicitly trained on. Modern LLMs can classify, translate, and analyze without task-specific training data.