---
title: "Clairic in Practice: Five Steps from Unstructured Input to a Typed Record"
description: "Clairic is a System One extraction model that turns unstructured input into typed values in a single forward pass. This five-step flow shows how to declare fields, call it, and use the results."
---

# Clairic in Practice: Five Steps from Unstructured Input to a Typed Record

**Type:** [How to](https://yomu.fyi/visual-explainers/category/how-to.md)

![A left-to-right five-step flow: declare typed fields, send unstructured input, read a value or null with offsets, check confidence, and persist typed values. A checkpoint band below lists under-one-second latency, no output token charges, and valid value or null.](https://yomufyi.s3.us-east-1.amazonaws.com/visual-explainers/ece8a374-e1ed-4a5c-bff3-6f68cfc49a30/image-1a94bd0f-b0b0-4145-b0c8-c571ed0db3ea.webp)

> Clairic is a System One extraction model that turns unstructured input into typed values in a single forward pass. This five-step flow shows how to declare fields, call it, and use the results.

## Explanation

Clairic is a preview extraction model on Workers AI that reads unstructured input in one forward pass and returns typed values such as numbers, dates, durations, and spans, each with character offsets and a confidence score. Because it generates no output tokens, responses come back in under a second and every answer is either a valid typed value or null. The five steps below cover declaring the fields you want, sending the input, reading values and offsets, validating confidence, and writing the typed result into a store that downstream systems can query instead of re-parsing raw text.

## FAQ

### Why does Clairic produce no output tokens, and what does that change for cost?

Clairic processes only the input tokens in a single forward pass and never generates tokens, so responses come back in under one second. There are no output token charges, which means cost tracks the input you send rather than the length of the answer you get back.

### When should I pick an extraction model over a decision model?

Decision models such as Jev or Clef output probabilities, which suit scoring and classification. Clairic returns a typed value, estimate, or forecast with a confidence score, so choose it when you need a number, date, duration, or span rather than a likelihood.

### What should my pipeline do when Clairic returns null or low confidence?

Null is a valid answer, not a failure, so treat it as 'not present in this input' and route that record accordingly. Confidence scores and character offsets let you gate on a threshold and keep the source span for review instead of re-parsing the raw text.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents.md), [Clairic](https://yomu.fyi/topic/clairic.md), [Extraction Models](https://yomu.fyi/topic/extraction-models.md), [Inference](https://yomu.fyi/topic/inference.md), [Workers AI](https://yomu.fyi/topic/workers-ai.md)

## Related reading

### [Real-time fraud detection with Feature Views](https://yomu.fyi/post/real-time-fraud-detection-with-feature-views.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Shyam Sankararaman, Hedvig Vahlberg
- Published: Nov 6, 2026
- Type: Tutorial

Real-time fraud detection needs three feature types: recent activity (e.g., five-minute transaction count), spending baselines (30-day average and standard deviation), and current payment details. Databricks Feature Views address this by defining each feature once for both training and serving, producing point-in-time correct training data and online inference pipelines. The walkthrough follows a $900 payment from a Kafka stream through RollingWindow and SawtoothWindow aggregations to a fraud score from Model Serving. Feature Views prevent training-serving skew and data leakage; on-demand features compare the current amount against stored baselines at request time. Unity Catalog governs definitions, permissions, and lineage, while MLflow records model dependencies and Model Serving retrieves features from the Online Feature Store.


### [Make Kubernetes logs easier to correlate with traces using the log normalizer operator](https://yomu.fyi/post/make-kubernetes-logs-easier-to-correlate-with-traces-using-the-log-nor.md)
- Company: [Red Hat](https://yomu.fyi/company/red-hat.md)
- Author: Denis Moskalenko
- Published: Nov 6, 2026
- Type: Explainer

The log normalizer operator for Red Hat OpenShift injects a sidecar into selected pods to normalize heterogeneous log formats without application code changes. It detects formats such as JSON, klog, Zap console, logfmt, and plain text, reassembles multiline entries like stack traces, and emits a consistent structured JSON record. A Drain-style template miner clusters structurally similar lines and assigns template IDs as additional metadata. The sidecar attaches Kubernetes resource attributes and promotes trace and span IDs from common field names to top-level fields for correlation. Normalized records remain available through kubectl logs and can optionally be exported over OTLP to an OpenTelemetry collector.


### [How to build an AI agent that remembers returning customers](https://yomu.fyi/post/how-to-build-an-ai-agent-that-remembers-returning-customers.md)
- Company: [Twilio](https://yomu.fyi/company/twilio.md)
- Author: Jesse Sumrak
- Published: Nov 6, 2026
- Type: Tutorial

AI agents that fail to recognize returning customers force them to repeat information, with 74% of consumers reporting repetition and 71% willing to abandon conversations early. Twilio’s approach uses three capabilities: identity resolution to match profiles across channels, recall to pull observations and summaries from Conversation Memory, and write-back to save new facts via the Memory API. Conversation Memory stores structured traits, extracted observations from conversations, and short conversation summaries, with the Recall API returning up to 20 observations and 5 summaries per profile. Agent Connect loads relevant context before the agent replies, running in Orchestrator mode with memory modes of always or once, and it uses ConversationRelay for voice calls. New facts are saved mid-conversation by agent tools calling the Memory API directly because automatic extraction only runs after a conversation closes, and memory carries across channels and human handoffs through Twilio Flex.


### [How TUI scales threat modeling for 100+ teams with Amazon Bedrock](https://yomu.fyi/post/how-tui-scales-threat-modeling-for-100-teams-with-amazon-bedrock.md)
- Company: [AWS](https://yomu.fyi/company/aws.md)
- Author: Vushesh Babu Adhikari
- Published: Nov 5, 2026
- Type: Explainer

TUI’s four-person security team faced a bottleneck: manual threat modeling for 100+ digital product teams, each assessment taking 4-6 hours. They built a system using Amazon Bedrock with Claude Sonnet 4.5 and deployed on Amazon Bedrock AgentCore runtime after a 10-week proof of concept with AWS. Instead of loading all 192 security standards into prompts, they implemented metadata-driven selection, reducing context from 338 KB to 21 KB and improving consistency. The LangGraph-based agentic workflow uses self-correcting threat generation and SME comparison for accuracy and completeness. In production, assessment time dropped by up to 70%, sessions per month increased from 3-4 to 10-12, and 255 users registered in eight weeks, making threat modeling mandatory for all new services.


### [Introducing Clairic, a new System One class of extraction models](https://yomu.fyi/post/introducing-clairic-a-new-system-one-class-of-extraction-models.md)
- Company: [Cloudflare](https://yomu.fyi/company/cloudflare.md)
- Author: Kevin Flansburg
- Published: Oct 19, 2026
- Type: Announcement

Cloudflare announces Clairic, a preview extraction model hosted on Workers AI, designed as a System One model for fast, one-pass value extraction from unstructured inputs. Unlike decision models such as Jev or Clef that output probabilities, Clairic returns extracted, computed, estimated, or forecasted typed values like numbers, dates, durations, and spans with character offsets and confidence scores. It processes only input tokens in a single forward pass, resulting in under-one-second responses and no output token charges, ensuring every answer is a valid typed value or null. The model handles extraction tasks like invoice numbers, estimation with distributions providing a median and 80% interval, and forecasting per future step with intervals, all without token generation. Cloudflare plans to release Clairic as open-weights soon and offers it via Workers AI with a typed field request structure accessible through env.AI.run() or the REST API.
