--- title: Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI description: A hands-on tutorial walks through building an eval-driven development pipeline for AI agents on Red Hat OpenShift AI 3.4+, combining MLflow for trace capture,... canonical: https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI | daily.dev og:description: A hands-on tutorial walks through building an eval-driven development pipeline for AI agents on Red Hat OpenShift AI 3.4+, combining MLflow for trace capture,... og:url: https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi og:image: https://api.daily.dev/og/posts/k7WDxXyFi.png og:image:alt: Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI **[Red Hat Developer](https://daily.dev/sources/rhdev)** · 22 min read · 0 upvotes · 0 comments ## Summary A hands-on tutorial walks through building an eval-driven development pipeline for AI agents on Red Hat OpenShift AI 3.4+, combining MLflow for trace capture, IBM CLEAR for LLM-as-a-judge error analysis, and EvalHub for orchestration. It covers deploying MLflow and EvalHub, uploading agent traces, configuring standard versus SPARC (tool-level) evaluation modes, submitting evaluation jobs via REST API, and interpreting per-agent scores, issue catalogs, and HTML dashboards in MLflow. Includes an appendix with full trace-generation scripts for a simulated multistep research agent. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is the difference between CLEAR standard mode and SPARC mode when evaluating AI agent tool calls? Standard mode (separate_tools: false) scores an agent step's reasoning and tool calls together as one holistic interaction, giving a single quality score per step. SPARC mode (separate_tools: true) splits tool calls into their own rows and evaluates each one individually with a dedicated reflection engine for argument correctness and tool selection, revealing whether an agent reasons well but picks wrong tools or calls the right tool with malformed arguments. SPARC issues roughly twice as many judge LLM calls as standard mode. _Teams debugging tool-selection failures in agent pipelines can track evaluation techniques like these on daily.dev._ ### What span structure does IBM CLEAR expect from MLflow traces of an AI agent? CLEAR expects an AGENT root span containing a CHAIN span, with CHAT_MODEL (or MODEL/GENERATION) spans for LLM calls and TOOL spans for tool executions. A span is treated as an LLM call if its span_type is CHAT_MODEL, MODEL, or GENERATION, if it has a gen_ai.operation.name attribute, or if its outputs contain a choices key. AGENT and CHAIN spans act as unscored wrappers, while TOOL spans are only scored by CLEAR in SPARC mode. _Developers instrumenting agent observability pipelines follow structural details like this on daily.dev._ ### Which agent framework and observability framework combinations does IBM CLEAR support for evaluation? CLEAR supports langgraph with mlflow, langgraph with langfuse, and crewai with langfuse, but crewai with mlflow is not supported. When mlflow is used as the observability framework, only langgraph is currently supported as the agent framework for trace parsing. _Anyone picking an agent framework and observability stack can compare compatibility notes like these on daily.dev._ ## Similar posts on daily.dev - [Eval-driven development: Build and evaluate reliable AI agents](https://daily.dev/posts/eval-driven-development-build-and-evaluate-reliable-ai-agents-xenqh9mz3) · Red Hat Developer · 1 upvotes · 0 comments - [EvalHub: Because "looks good to me" isn't a benchmark](https://daily.dev/posts/evalhub-because-looks-good-to-me-isn-t-a-benchmark-osnz9ucrd) · Red Hat Developer · 0 upvotes · 0 comments - [How EvalHub manages two-layer Kubernetes control planes](https://daily.dev/posts/how-evalhub-manages-two-layer-kubernetes-control-planes-mprwmrrvs) · Red Hat Developer · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#red-hat](https://daily.dev/tags/red-hat) [View this post on daily.dev](https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI","url":"https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi"},"datePublished":"2026-09-03T07:29:54.419Z","dateModified":"2026-09-03T08:25:11.894Z","description":"A hands-on tutorial walks through building an eval-driven development pipeline for AI agents on Red Hat OpenShift AI 3.4+, combining MLflow for trace capture,...","image":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","thumbnailUrl":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","isAccessibleForFree":true,"articleSection":"Red Hat Developer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Red Hat Developer","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/2a8895f3216c4900bdc455585a4ad3e4","url":"https://daily.dev/sources/rhdev"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,llm,ai-agents,red-hat","timeRequired":"PT22M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Red Hat Developer","item":"https://daily.dev/sources/rhdev"},{"@type":"ListItem","position":3,"name":"Evaluate AI agents with IBM CLEAR & EvalHub on OpenShift AI"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/evaluate-ai-agents-with-ibm-clear-evalhub-on-openshift-ai-k7wdxxyfi#faq","mainEntity":[{"@type":"Question","name":"What is the difference between CLEAR standard mode and SPARC mode when evaluating AI agent tool calls?","acceptedAnswer":{"@type":"Answer","text":"Standard mode (separate_tools: false) scores an agent step's reasoning and tool calls together as one holistic interaction, giving a single quality score per step. SPARC mode (separate_tools: true) splits tool calls into their own rows and evaluates each one individually with a dedicated reflection engine for argument correctness and tool selection, revealing whether an agent reasons well but picks wrong tools or calls the right tool with malformed arguments. SPARC issues roughly twice as many judge LLM calls as standard mode. Teams debugging tool-selection failures in agent pipelines can track evaluation techniques like these on daily.dev."}},{"@type":"Question","name":"What span structure does IBM CLEAR expect from MLflow traces of an AI agent?","acceptedAnswer":{"@type":"Answer","text":"CLEAR expects an AGENT root span containing a CHAIN span, with CHAT_MODEL (or MODEL/GENERATION) spans for LLM calls and TOOL spans for tool executions. A span is treated as an LLM call if its span_type is CHAT_MODEL, MODEL, or GENERATION, if it has a gen_ai.operation.name attribute, or if its outputs contain a choices key. AGENT and CHAIN spans act as unscored wrappers, while TOOL spans are only scored by CLEAR in SPARC mode. Developers instrumenting agent observability pipelines follow structural details like this on daily.dev."}},{"@type":"Question","name":"Which agent framework and observability framework combinations does IBM CLEAR support for evaluation?","acceptedAnswer":{"@type":"Answer","text":"CLEAR supports langgraph with mlflow, langgraph with langfuse, and crewai with langfuse, but crewai with mlflow is not supported. When mlflow is used as the observability framework, only langgraph is currently supported as the agent framework for trace parsing. Anyone picking an agent framework and observability stack can compare compatibility notes like these on daily.dev."}}]} ```