# Deepchecks Documentation

> Documentation for Deepchecks

## Guides
- [What is Deepchecks?](https://deepchecks.readme.io/docs/what-is-deepchecks.md): Deepchecks is an evaluation platform for LLM-based and agentic AI applications - helping you understand how your AI system performs, find where it fails, and continuously improve it from development to production.
- [Quickstart: SDK](https://deepchecks.readme.io/docs/quickstart-sdk.md): Send your first interactions to Deepchecks and see evaluation and observability results in under 5 minutes, using the Python SDK.
- [Quickstart: UI](https://deepchecks.readme.io/docs/quickstart-ui.md): Upload a CSV file and explore Deepchecks evaluation results without writing any code - in just a few minutes.
- [Supported Use Cases](https://deepchecks.readme.io/docs/supported-use-cases.md): Deepchecks supports a wide range of LLM applications - from RAG and Q&A to multi-agent workflows. Each use case maps to one or more interaction types that configure properties, annotations, and evaluation automatically.
- [Key Concepts](https://deepchecks.readme.io/docs/key-concepts.md): A reference guide to every core concept in Deepchecks - the data model, evaluation building blocks, and how they fit together.
- [Navigating the UI](https://deepchecks.readme.io/docs/navigating-the-ui.md): A guide to every screen in the Deepchecks UI - what it shows, when to use it, and how it fits into your evaluation workflow.
- [Integration Overview](https://deepchecks.readme.io/docs/integration-overview.md): Connect your real LLM pipeline data to Deepchecks. Choose the integration path that fits your stack - framework auto-instrumentation, the Python SDK, or CSV upload.
- [Auto-Instrumentation (Frameworks)](https://deepchecks.readme.io/docs/auto-instrumentation.md): If you use LangGraph, CrewAI, Google ADK, or LangChain, add a few lines of setup code and Deepchecks captures every trace, span, and metric automatically.
- [Offline Trace Export](https://deepchecks.readme.io/docs/offline-trace-export.md): Export traces to a JSON file when you cannot send data to Deepchecks in real time, then upload them in a separate step.
- [Python SDK Integration](https://deepchecks.readme.io/docs/sdk-integration.md): Send data from your custom LLM pipeline to Deepchecks using the Python SDK - batch uploads for evaluation, streaming for production, and everything in between.
- [Upload Agentic Data](https://deepchecks.readme.io/docs/upload-agentic-data.md): Understand how Deepchecks models agentic hierarchical data - spans, traces, parent/child relationships - and how to upload it from a custom agent framework.
- [Upload via CSV](https://deepchecks.readme.io/docs/upload-via-csv.md): Upload interaction data to Deepchecks from the browser using a CSV file - no code required.
- [Data Fields Reference](https://deepchecks.readme.io/docs/data-fields-reference.md): Complete reference for every data field Deepchecks supports - content fields, metadata fields, and span fields - with descriptions and usage guidance.
- [How Evaluation Works](https://deepchecks.readme.io/docs/evaluation-overview.md): Once your data is in Deepchecks, the platform evaluates every interaction automatically. This section explains how evaluation works and what the results mean.
- [Properties](https://deepchecks.readme.io/docs/properties-guide.md): Properties are the foundation of evaluation in Deepchecks - individual quality scores calculated on every interaction, used for annotations, comparison, filtering, and root cause analysis.
- [Built-in Properties](https://deepchecks.readme.io/docs/builtin-properties.md): Deepchecks' proprietary quality models - a complete catalog of what is calculated on your interactions automatically, organized by category.
- [Quality & Accuracy Properties](https://deepchecks.readme.io/docs/quality-and-accuracy-properties.md): Built-in properties that measure whether your LLM output correctly and fully addresses the input - relevance, completeness, instruction fulfillment, and more.
- [Safety & Risk Properties](https://deepchecks.readme.io/docs/safety-and-risk-properties.md): Built-in properties that detect harmful content, PII leakage, refusals, and system errors in your LLM application's outputs.
- [Text Quality & Style Properties](https://deepchecks.readme.io/docs/text-quality-properties.md): Built-in properties that measure how well-written, readable, and appropriately structured your LLM application's outputs are.
- [RAG Use-Case Properties](https://deepchecks.readme.io/docs/retrieval-use-case-properties.md): Overview of the classification of documents and built-in retrieval properties on the Deepchecks app
- [Agent Use-Case Properties](https://deepchecks.readme.io/docs/agent-use-case-properties.md): Deepchecks provides built-in properties tailored for agentic workflows, giving you structured, research-backed ways to evaluate how agents plan, act, and use tools effectively
- [Prompt Properties](https://deepchecks.readme.io/docs/prompt-properties.md): Create custom LLM-as-a-judge evaluators that score interactions on criteria you define - in natural language, without code.
- [Data Fields for Evaluation](https://deepchecks.readme.io/docs/data-component-for-prompt-properties.md)
- [Improve Guidelines with AI](https://deepchecks.readme.io/docs/improve-guidelines-with-ai.md): AI-powered tool to refine and optimize your property guidelines.
- [Number of Judges](https://deepchecks.readme.io/docs/number-of-judges.md): Run multiple independent LLM evaluators per property and aggregate their results to reduce variance on high-stakes scores.
- [Refining Properties with Few-Shot Examples](https://deepchecks.readme.io/docs/refining-properties-with-feedback.md): Improve property accuracy by adding real interactions as examples that guide the evaluator toward your quality standards
- [User-Value (Custom) Properties](https://deepchecks.readme.io/docs/custom-properties.md): Learn what types of user-value properties exist in Deepchecks LLM Evaluation and how they are defined
- [Session-Level Properties](https://deepchecks.readme.io/docs/session-level-properties.md): Learn about Deepchecks' session-level properties - what they're used for and how to configure them
- [Activate or Pause Property Calculations](https://deepchecks.readme.io/docs/activate-or-pause-property-calculations.md): Learn how to put properties on "Pause" mode, allowing you to calculate them ad-hoc on any interaction while saving costs during automatic calculation upon data upload
- [Automatic Annotations](https://deepchecks.readme.io/docs/auto-annotation.md): Deepchecks automatically labels every interaction as Good, Bad, or Unknown using a configurable pipeline - defined per interaction type.
- [Manual Annotations](https://deepchecks.readme.io/docs/manual-annotations.md)
- [Analyze & Improve Overview](https://deepchecks.readme.io/docs/analyze-improve-overview.md): Find problems, test solutions, and pick the best version - the complete improvement loop from investigation to production monitoring.
- [The Interactions and Sessions Screens](https://deepchecks.readme.io/docs/interactions-and-sessions-screen.md): How to navigate, filter, and search the Interactions and Sessions screens - the starting point for any quality investigation in Deepchecks.
- [Root Cause Analysis](https://deepchecks.readme.io/docs/root-cause-analysis.md): Investigate why your LLM application fails - property explainability, annotation breakdowns, automated insights, and failure mode analysis.
- [Navigating the Single Interaction/Session Screen](https://deepchecks.readme.io/docs/navigating-the-session-view.md): Walk through individual traces span by span - inspect data, evaluation results, and the execution hierarchy in one screen.
- [Dataset Management](https://deepchecks.readme.io/docs/dataset-management.md): Create and manage curated evaluation sets for systematic testing across versions - build them manually, from production data, or with AI generation.
- [Generate Test Data](https://deepchecks.readme.io/docs/ai-data-generation.md)
- [Version Comparison](https://deepchecks.readme.io/docs/version-comparison.md): Compare versions of your LLM application side by side - from high-level score overviews to granular interaction-by-interaction differences.
- [Cost Tracking](https://deepchecks.readme.io/docs/cost-tracking.md): Automatic LLM cost tracking and analysis across interactions, sessions, and versions based on configured model pricing
- [Production Monitoring](https://deepchecks.readme.io/docs/production-monitoring-1.md): Track your LLM application's performance over time in production - annotation trends, property scores, time-range comparisons, and external integrations.
- [Configure Your Application](https://deepchecks.readme.io/docs/configure-your-application.md): How to set up your Deepchecks application - interaction types, properties, auto-annotation, sampling, cost, document classification, and more.
- [Evaluate a RAG Pipeline](https://deepchecks.readme.io/docs/evaluate-a-rag-pipeline.md): End-to-end guide to evaluating a retrieval-augmented generation pipeline in Deepchecks - from uploading data to tuning annotations and comparing versions.
- [Evaluate an Agentic pipeline](https://deepchecks.readme.io/docs/agent-evaluation-use-case.md): End-to-end guide to evaluating multi-step agentic workflows in Deepchecks - from configuring deployments and generating test data, through automated simulation, to component-level evaluation and root-cause analysis.
- [Run Agent Simulations (KYA)](https://deepchecks.readme.io/docs/kya-run-agent-simulations.md): How to proactively test your agent using Know Your Agent (KYA) - configure a deployment, generate test scenarios with AI, and run them against your agent at scale.
- [Create and Refine a Prompt Property](https://deepchecks.readme.io/docs/defining-and-refining-prompt-properties.md): Step-by-step walkthrough: create a numerical prompt property, test it on real interactions, and iteratively improve it until it produces reliable scores.
- [Configure Auto Annotation Rules](https://deepchecks.readme.io/docs/configure-auto-annotation.md): How to select the right evaluation properties, thresholds and aggregate into a high quality evaluation flow.
- [Configuring Auto Annotation in the UI](https://deepchecks.readme.io/docs/configuring-auto-annotation-in-the-ui.md)
- [Selecting the Right Properties](https://deepchecks.readme.io/docs/selecting-the-right-properties.md)
- [Threshold Detection](https://deepchecks.readme.io/docs/adjusting-the-thresholds.md)
- [Auto Annotation Design](https://deepchecks.readme.io/docs/building-the-auto-annotation.md)
- [The Configuration YAML](https://deepchecks.readme.io/docs/the-auto-annotation-configuration-yaml.md)
- [Run a Version Comparison](https://deepchecks.readme.io/docs/version-comparison-flow.md): Step-by-step guide to comparing two or more versions of your LLM application in Deepchecks - from uploading against a shared dataset to picking a winner.
- [Set Up Production Monitoring](https://deepchecks.readme.io/docs/set-up-production-monitoring.md): How to start monitoring your LLM application in production - configure sampling, connect external tools, start logging, and know what to watch for.
- [Integrate into CI/CD](https://deepchecks.readme.io/docs/ci-cd.md): Add automatic quality gates to your CI/CD pipeline - run your new version against an evaluation set and fail the build if quality drops below threshold.
- [Pentest Your LLM App](https://deepchecks.readme.io/docs/pentesting-your-llm-pipeline.md): Test your LLM application against a broad set of adversarial prompts - prompt injections, jailbreaks, PII extraction, bias triggers - and analyze resilience.
- [Export Data for Offline Analysis](https://deepchecks.readme.io/docs/export-data-for-offline-analysis.md): Download enriched interaction data from Deepchecks - with property scores, annotations, and topics - for fine-tuning, external analysis, or CI/CD integration.
- [Multi-Turn Agent Demo: Google ADK Academic Researcher](https://deepchecks.readme.io/docs/multi-turn-agent-demo-google-adk-academic-researcher.md): Evaluating a multi-turn, multi-agent application built on Google ADK - from auto-instrumentation to component-level scoring, multi-turn session quality, system metrics, and root cause analysis.
- [Logging the Data](https://deepchecks.readme.io/docs/uploading-the-data-adk-academic-researcher.md)
- [Analyze Multi-Agent Performance](https://deepchecks.readme.io/docs/analyze-multi-agent-adk-academic-researcher.md)
- [Evaluate Multi-Turn Sessions](https://deepchecks.readme.io/docs/analyze-multi-turn-adk-academic-researcher.md)
- [Observability and System Metrics](https://deepchecks.readme.io/docs/observability-adk-academic-researcher.md)
- [Root Cause Analysis](https://deepchecks.readme.io/docs/root-cause-analysis-adk-academic-researcher.md)
- [Multi-Agent Demo: Content Creator Crew](https://deepchecks.readme.io/docs/multi-agent-demo-content-creator-crew.md): Evaluating and debugging a Multi-Agent application, step by step.
- [Logging the Data](https://deepchecks.readme.io/docs/uploading-the-data-content-creator.md)
- [Analyze Performance](https://deepchecks.readme.io/docs/analyze-performance-content-creator.md)
- [Root Cause Analysis (RCA)](https://deepchecks.readme.io/docs/root-cause-analysis-rca-content-creator.md)
- [Compare Versions](https://deepchecks.readme.io/docs/compare-versions-content-creator.md)
- [Q&A Demo: GVHD Data](https://deepchecks.readme.io/docs/qa-demo-gvhd-use-case.md): Evaluating and debugging a Q&A application, step by step
- [Uploading the Data](https://deepchecks.readme.io/docs/uploading-the-data.md)
- [Identify Problems Using Properties, Estimated Annotations and Insights](https://deepchecks.readme.io/docs/identify-problems-using-properties-estimated-annotations-and-insights.md)
- [User-Value Properties and Prompt Properties](https://deepchecks.readme.io/docs/user-value-properties-and-prompt-properties.md)
- [Compare Between Versions](https://deepchecks.readme.io/docs/compare-between-versions.md)
- [Monitor Production Data and Research Degradation](https://deepchecks.readme.io/docs/monitor-production-data-and-research-degradation.md)
- [Summarization Demo: E-Commerce Data](https://deepchecks.readme.io/docs/summarization-demo.md): Evaluating and debugging a Summarization application, step by step
- [Uploading the Data](https://deepchecks.readme.io/docs/uploading-the-data-summarization.md): Follow step by step to upload the data in two ways - python SDK or Deepchecks UI
- [Configuring the Automatic Annotation](https://deepchecks.readme.io/docs/use-llm-properties-summarization.md)
- [Compare Between Versions](https://deepchecks.readme.io/docs/compare-between-versions-summarization.md)
- [Production Monitoring](https://deepchecks.readme.io/docs/production-monitoring-3.md)
- [Classification Demo: Movie Genre](https://deepchecks.readme.io/docs/classification-demo-movie.md): Evaluating and debugging a classification application, step by step.
- [Uploading the Data](https://deepchecks.readme.io/docs/classification-demo-uploading-the-data.md)
- [Evaluation Set Analysis](https://deepchecks.readme.io/docs/evaluation-set-analysis.md): Analyzing performance on the labeled evaluation set.
- [Production Monitoring](https://deepchecks.readme.io/docs/production-monitoring-2.md): Tracking model performance in the wild
- [Deepchecks API Key](https://deepchecks.readme.io/docs/deepchecks-api-key.md)
- [Access Control](https://deepchecks.readme.io/docs/access-controls.md): Learn how Deepchecks manages user permissions using role-based access control and how to assign and manage roles in your organization.
- [Compliance](https://deepchecks.readme.io/docs/compliance.md): This page covers compliance features and capabilities available in the Deepchecks platform
- [Usage Management and Optimization](https://deepchecks.readme.io/docs/get-the-most-out-of-your-dpus.md): Understand how DPUs work on Deepchecks SaaS, monitor your usage, and optimize spend across sampling, model selection, properties, and data volume.
- [Framework Integrations](https://deepchecks.readme.io/docs/frameworks.md): Deepchecks seamlessly integrates with leading LLM application frameworks, enabling you to monitor, evaluate, and trace your agentic workflows with minimal setup.
- [CrewAI](https://deepchecks.readme.io/docs/crewai.md): Deepchecks integrates seamlessly with CrewAI, providing automated tracing, evaluation, and observability for multi-agent workflows and tool-assisted pipelines
- [LangGraph](https://deepchecks.readme.io/docs/langgraph.md): Deepchecks integrates with LangGraph to provide tracing, evaluation, and observability for agents and their interactions across the LangGraph workflow.
- [Google ADK](https://deepchecks.readme.io/docs/google-adk.md): Deepchecks integrates with Google ADK to provide tracing, evaluation, and observability for agents and their interactions across the Google ADK workflow.
- [Production Monitoring Integrations](https://deepchecks.readme.io/docs/production-monitoring.md): After customizing Deepchecks for a given LLM-based app, utilize the artifacts created to continuously monitor both quality and risk metrics in production
- [Datadog](https://deepchecks.readme.io/docs/datadog-integration.md)
- [New Relic](https://deepchecks.readme.io/docs/new-relic-integration.md): Applies also to any APM with accessible SDK to upload data
- [AWS Cloudwatch](https://deepchecks.readme.io/docs/aws-cloudwatch-integration.md)
- [Self-Hosted Deployments](https://deepchecks.readme.io/docs/deepchecks-self-hosted-deployments.md): Infrastructure prerequisites and deployment guide for Deepchecks Self-Hosted Enterprise
- [Helm Chart Reference](https://deepchecks.readme.io/docs/helm-chart-reference.md): Versioning, compatibility, and upgrade reference for the Deepchecks llm-stack Helm chart
- [Model Configuration for Self-Hosted Deployments](https://deepchecks.readme.io/docs/model-configuration-for-self-hosted-deployments.md)
- [Deepchecks in AWS SageMaker](https://deepchecks.readme.io/docs/aws-sagemaker.md): How to set up and use Deepchecks as a native AWS SageMaker Partner AI App - running entirely in your AWS account with your own Bedrock models.
- [Working with LLM Features - Deepchecks on SageMaker](https://deepchecks.readme.io/docs/working-with-llm-features-deepchecks-on-sagemaker.md): Learn how to optimize Deepchecks’ LLM features on SageMaker, including model selection and processing speed, to balance cost, capacity, and reliability.
- [Using the Python SDK on SageMaker](https://deepchecks.readme.io/docs/sdk-on-sagemaker.md): How to install and initialize the Deepchecks Python SDK when running on AWS SageMaker - authentication, environment variables, and the SigV4 signing flow.
- [Optimizing LLM Costs in Deepchecks on SageMaker](https://deepchecks.readme.io/docs/optimizing-llm-costs-in-deepchecks-on-sagemaker.md): Learn to reduce LLM costs in your Deepchecks SageMaker deployment by configuring cost-effective models, controlling data volume, optimizing property usage, and monitoring spend.
- [Using the Owner Panel](https://deepchecks.readme.io/docs/using-the-owner-panel.md)
- [AWS CloudWatch Integration](https://deepchecks.readme.io/docs/aws-cloudwatch.md): Send Deepchecks evaluation metrics and application logs to your AWS CloudWatch account - for unified monitoring and log aggregation alongside your other AWS services.

## API Reference
- [DeepchecksLLMClient](https://deepchecks.readme.io/reference/deepchecks-llm-client.md)
- [DeepchecksCallbackHandler](https://deepchecks.readme.io/reference/deepchecks-callback-handler.md)
- [Data Types](https://deepchecks.readme.io/reference/data-types.md)
- [Create Application](https://deepchecks.readme.io/reference/createapp-1.md): Create Application
- [List Applications](https://deepchecks.readme.io/reference/listapps-1.md): List applications
- [Getting Started with our REST API](https://deepchecks.readme.io/reference/getting-started-with-your-api.md): This page will help you get started with our REST API
- [Retrieve Backend Version](https://deepchecks.readme.io/reference/retrieve_backend_version_api_v1_backend_version_get-1.md)
- [Get Interactions By Filter](https://deepchecks.readme.io/reference/listinteractionsbyfilter-1.md): List interactions by filter
- [Download All Interactions By Filter](https://deepchecks.readme.io/reference/listinteractionsdownloadbyfilter-1.md): List interactions download by filter
- [Get Interaction Completion Status](https://deepchecks.readme.io/reference/get_interaction_completion_status_api_v1_application_versions__application_version_id__interactions_complete_status_post-1.md)
- [Delete Interactions](https://deepchecks.readme.io/reference/delete_interactions_api_v1_application_versions__application_version_id__interactions_delete-1.md)
- [Create Interactions](https://deepchecks.readme.io/reference/publiccreateinteractions-1.md): Create interactions
- [Get Raw Spans By Filter](https://deepchecks.readme.io/reference/listrawspansbyfilter.md): List raw spans by filter (same filtering logic as get-interactions-by-filter)
- [List Application Versions](https://deepchecks.readme.io/reference/listappversions-1.md): List application versions
- [Create Application Version Router](https://deepchecks.readme.io/reference/createappversion-1.md): Create application version

## Changelog
- [0.45.0 Release Notes](https://deepchecks.readme.io/changelog/0450-release-notes.md)
- [0.44.0 Release Notes](https://deepchecks.readme.io/changelog/0440-release-notes.md)
- [0.43.0 Release Notes](https://deepchecks.readme.io/changelog/0430-release-notes.md)
- [0.42.0 Release Notes](https://deepchecks.readme.io/changelog/0420-release-notes.md)
- [0.41.0 Release Notes](https://deepchecks.readme.io/changelog/0410-release-notes.md)
- [0.40.0 Release Notes](https://deepchecks.readme.io/changelog/0400-release-notes.md)
- [0.39.0 Release Notes](https://deepchecks.readme.io/changelog/0390-release-notes.md)
- [0.38.0 Release Notes](https://deepchecks.readme.io/changelog/0380.md)
- [0.37.0 Release Notes](https://deepchecks.readme.io/changelog/0370.md)
- [0.36.0 Release Notes](https://deepchecks.readme.io/changelog/0360-release-notes.md)