Agentic SRE App Release Notes
Note
Version 0.2.0 marks the Beta release of the ITRS Agentic SRE app. For more information, see the latest documentation for Use Agentic SRE for operations and support in ITRS Analytics.
This app release requires a minimum ITRS Analytics Platform version of 2.17.0 and Web Console version 3.8.0 to operate.
Agentic SRE Beta 0.7.1 Copied
Released: 30 October 2026 Beta
This app is packaged with ITRS Analytics 2.18.16+p2.18.13.
Issues fixed Copied
These are the issues we have fixed in this release:
| Issue key | Release description |
|---|---|
| IAA-1261 | Fixed an issue where non-admin users repeatedly received a “Forbidden” toast notification every few seconds while the Agentic SRE panel was open. The system-status indicator in the chat header now displays only for admin users, preventing unauthorized status requests and recurring error toasts. |
Agentic SRE Beta 0.7.0 Copied
Released: 19 October 2026 Beta
This app is packaged with ITRS Analytics 2.18.15+p2.18.10.
Highlights Copied
These are the highlights of this release:
- Telemetry in the answer — Ask
Insightsto show logs, a metric over a period, a severity change, or traces. The answer shows those records, and a follow-up stays on the entity from the previous turn. - Dashboards from chat — The chat now generates a dashboard preview for a selected entity or named scope, which you can open and review before saving. The reply names the metrics that were chosen, and you can export the definition before or after saving. For known limitations of dashboards created from chat in this beta, see Create a dashboard from chat.
- Evaluation — New graders cover documentation search, safety, tool arguments, and conversations of more than one turn. The test score measures answer quality. Speed and token use are reported separately.
- Administration — Route one LLM through an HTTP proxy, export a conversation, and read times in the date format, time format, and time zone from your user settings.
New features and enhancements Copied
These are the new features and enhancements of this release:
| Issue key | Release description |
|---|---|
| IAA-534 | Added Retrieval and Faithfulness graders. Retrieval scores the documents that were found against the question and the expected answer. Faithfulness scores whether the answer’s claims follow from those documents. A run with no retrieval record is N/A, not zero. See Understand Eval results. |
| IAA-536 | Added evaluation tests with more than one turn in the same conversation, sent through the chat path. Context carry-over checks that later turns still use the entity from an earlier turn, including when a test changes subject on purpose. |
| IAA-537 | Added grading of an investigation’s tool trace, not only the final wording. The report can score expected tool arguments, the number of tool calls, the time window, whether the named entity was resolved, whether the investigation was sound, whether the requested task was completed, and what the turn did after an injected platform failure. |
| IAA-539 | Added a Safety grader and tests for instructions that try to override the assistant’s rules, and for requests that are out of scope. Only tests written for this check are scored. See Understand Eval results. |
| IAA-643 | Updated the agent runtime so saved model settings are sent with each request. Azure OpenAI remains available, and an Ollama model uses the model name configured for that LLM. |
| IAA-679 | The test score now includes only graders that measure answer quality. Duration, token use, and consistency stay on the report and do not change that score. A report stored before this split is marked ungated. The accuracy grader uses the weighted criteria in its prompt, and a string-content check with no expected strings is reported as a configuration error instead of a half score. See Understand Eval results. |
| IAA-680 | Added graders for whether an answer is direct, whether it says where the information came from, whether it matches an expected format, how close it is to the expected answer, and whether a clarifying question addressed the real ambiguity. |
| IAA-681 | Added a Tool Args grader that checks whether each tool was called with the expected arguments, such as the entity, filters, and time window. A tool that was not called is N/A. See Understand Eval results. |
| IAA-745 | Numeric Invariant, Entity Resolution, Context Carry-over, Temporal, and Error Recovery stay listed in the evaluation wizard and cannot be selected. Each says that it needs freshly seeded platform data. See Run an Eval campaign and review the results. |
| IAA-838 | Charts on a dashboard created from chat were placed on a Trends tab, and table height follows the tile’s row limit instead of a large fixed height. A collapsed tile no longer overlaps the tiles below it. |
| IAA-871 | The chat reply explains each tile on a dashboard created from chat and why it was included. A tile that lists only what is in trouble names its signal. See Create a dashboard from chat. |
| IAA-923 | You can take the JSON of a dashboard created in chat, including a preview you have not saved, with Menu > Export as JSON in the Dashboards app. The chat does not add a separate download. See Create a dashboard from chat. |
| IAA-930 | Insights can investigate a question by checking platform data and showing the records that came back, rather than only a written summary. See Explore telemetry with Insights. |
| IAA-931 | SRE routes a request to see logs, errors, traces, when something started, what changed, or a comparison to Insights investigate. A status question stays an entity summary. A question about why stays root cause analysis. A definition question stays Support. |
| IAA-933 | Raised the conversation limit and made it configurable. The defaults are a cost of 10, 200 actions, and 5,000,000 tokens. When the limit or the tool-call loop is exhausted, the answer keeps the text already streamed and adds one line that the investigation stopped and what it had established. It does not substitute a summary that was not checked. See Read an empty or partial answer. |
| IAA-934 | A question such as “When did this start?” returns the instant the severity changed and quotes the signal message. See Explore telemetry with Insights. |
| IAA-935 | A metric question returns the series together with the severity that was active over it. See What the answer can show. |
| IAA-936 | A telemetry check takes either one entity, matched exactly, or a scope that can include the entities beneath a parent. A call that supplies both or neither is rejected. An exact match that finds nothing, while the same identity has children, returns those child identities instead of an empty result. |
| IAA-937 | Log searches can be filtered by the severity the platform assigned to each line. A message search is no longer described to the assistant as that severity filter. See Explore telemetry with Insights. |
| IAA-938 | When a metric series hits its 200-point cap, the answer says the series was shortened and states the window that was returned. |
| IAA-939 | Trace results can be ordered by duration and limited to spans at least as long as a minimum duration. Each span includes its own time, separate from the time spent in the spans it called. |
| IAA-940 | The period you name in a question, such as “over the last three days” or “this morning”, is the period that is searched. |
| IAA-941 | A follow-up such as “When did this issue start?” stays on the entity from the previous turn. See Explore telemetry with Insights. |
| IAA-942 | Tool calls made while Insights answers are labelled in the chat. |
| IAA-945 | A status question is handled by Insights in status mode. The separate entity-summary recipe is retired. |
| IAA-947 | Added chat views for a metric series, a severity timeline, and a log grid. Each view is controlled by its own option and is off until that option is enabled. See What the answer can show. |
| IAA-948 | Added a trace waterfall view in chat, controlled by its own option. |
| IAA-958 | Applied each model’s token price to the conversation, so the conversation cost limit can stop a long investigation. A model with no listed price does not count toward that limit. |
| IAA-999 | When an attached grader scored no tests, the evaluation report names it and says that the empty column is not a quality signal and is not in the test score. A report where every grader scored, or a report from before this field existed, does not show that note. See Understand Eval results. |
| IAA-1001 | The entity a turn resolves is remembered, so a follow-up such as “When did this start?” keeps that entity even when no earlier tool result named it. |
| IAA-1030 | Replaced the Trends tab. A dashboard created from chat uses one tab: a full-width chart of several metrics, single-metric charts three to a row, and one table limited to one row with severity first. A chart that points at a dropped metric is removed and reported. Some generated chart tiles include a current-value setting that the Dashboards app does not display yet, so those tiles show the chart only. See Known limitations. |
| IAA-1056 | Times in chat answers use the date format, time format, and time zone from your ITRS Analytics user settings. |
| IAA-1065 | Metric charts, severity timelines, log grids, and trace waterfalls are enabled by default. A client that cannot draw one of these views shows the same information as a table. An installation can turn the four view options off. |
| IAA-1068 | Added a Proxy section on an LLM so that model can be reached through an HTTP proxy. The proxy applies to that LLM only, and saving applies it without an app restart. See Route an LLM through an HTTP proxy. |
| IAA-1083 | Added export of a conversation’s full record as JSON, from the chat for your own conversation and from Audit for any conversation in the tenant. The two permissions are separate. See Export a conversation. |
| IAA-1098 | When an installation enables interface options that are not in the default chart, chat and Admin show a warning banner that cannot be dismissed and names each of those options. A default installation shows no banner. See Review service readiness in Status. |
| IAA-1140 | Audit is available in Admin > AI without a Preview label. Usage and Connectors are available as preview features in that section only. Usage counts AI usage against the license entitlement and requires one. No connectors are recommended yet. |
Issues fixed Copied
These are the issues we have fixed in this release:
| Issue key | Release description |
|---|---|
| IAA-401 | After you rotate an API key in the Secrets app, a call that the provider rejects reads the new secret. A replacement that the provider does not reject is picked up on the next lookup after the cache interval, which is 15 minutes by default. See Store provider credentials in the Secrets app. |
| IAA-548 | A failed Test LLM check is shown as failed on the audit turn, matching the conversation status. |
| IAA-567 | When the signed-in user’s token has expired, platform calls continue with the app’s own credentials so a long-running request can still finish. |
| IAA-584 | When entity disambiguation asks you to choose, typing the option number selects that entity. The number is the position in the list, starting at 1. See Answer a question from the chat. |
| IAA-615 | A value on a dashboard created from chat is no longer marked out of date because its column had no time window. Metrics that no entity in the scope exposes are left out before the dashboard is built, instead of being chosen and shown as empty. |
| IAA-623 | Removed deprecated OpenAI chat model names from the model list. |
| IAA-626 | Fixed streaming calls that could fail with no reported error, and shutdown handling that could lose the shutdown signal. A method that checked for a missing value and then used it anyway was corrected. List sort order was left unchanged. |
| IAA-682 | Evaluation averages no longer include grader results that the report shows as N/A. |
| IAA-683 | Running an evaluation from the wizard keeps the test details that the format, safety, clarification, and tool-argument graders need. |
| IAA-684 | The evaluation grader list includes the graders an SRE run can select, including clarification. |
| IAA-707 | The provider list in the LLM editor now requires permission to manage AI settings. The lists used by chat, including models, agents, feature options, and your own chat settings, stay available without that permission. |
| IAA-708 | Deleting a test did not change which later tests were selected; that selection was already remapped. Editing the initial blank test no longer changes every test you add afterwards. |
| IAA-709 | Validation failures are now reported when you save an LLM, an MCP server, or a tool. Save was already disabled while the form was invalid. The new message covers a rejection that happens after the form was opened, such as a duplicate name. |
| IAA-712 | Removed checks for missing values that could not occur. Those checks did not change chat or investigation behavior. |
| IAA-713 | Limited text matching on a chat message so a long message cannot hold the request for a long time. This includes matching that decides whether a message confirms or rejects a dashboard, and matching used to detect an application-definition question. |
| IAA-741 | An unreadable usage counter is kept so it can be repaired. The next usage write no longer replaces it and understates the open day’s usage. |
| IAA-742 | A failure to store the price for a billing period is reported as a failure. The period is no longer treated as priced when nothing was stored. |
| IAA-743 | A telemetry check no longer fails when a time value is missing or not a positive instant. |
| IAA-746 | Audit records a failed tool call as failed, and a completed action as completed. |
| IAA-747 | Errors from administration calls include a description, instead of an unknown failure with no reason. |
| IAA-748 | A refused deletion of a tool or an MCP server is reported as refused, and the item stays in the configuration. |
| IAA-750 | A documentation search that is limited to one product no longer answers a retry from documents the first search excluded. |
| IAA-751 | Stopped offering a vector search that the active index cannot answer, so the assistant does not call a search that always returns nothing. |
| IAA-756 | A missing or invalid identifier in an administration request returns the reason, instead of an unknown error. |
| IAA-757 | A backend entry with no name is skipped, instead of emptying the backend list. |
| IAA-758 | A failure to load Audit is shown as an error, instead of an empty conversation list. |
| IAA-759 | Opening a second LLM, MCP server, or tool no longer shows the details of the first one. |
| IAA-761 | A refused MCP connection test names the endpoint, instead of reporting a blank failure. |
| IAA-762 | An MCP connection test no longer holds the request for 45 seconds when the transport is left on automatic. |
| IAA-769 | Saving chat settings when chat storage is not configured reports that the store is missing, instead of a server fault. |
| IAA-772 | A grader that was selected but scored no tests stays on the evaluation report and says that it could not measure. |
| IAA-785 | Removed metric forecasts from answers. The reported trend could be flat or reversed. |
| IAA-800 | A clarifying question is no longer padded with documentation citations that are unrelated to the question. |
| IAA-807 | Fixed log searches that used a time window a year earlier than the period you asked for. |
| IAA-814 | Usage is still recorded when chat history storage fails, and turning off chat history no longer turns off usage metering. Reading usage is a separate permission from reading every conversation or changing LLM settings. |
| IAA-837 | Corrected a text pattern on the page address that could run for a long time on an unusual path. |
| IAA-844 | If the conversation history cannot be read, the assistant is told that earlier history is missing rather than that there was none. Messages already in the conversation are not followed by older messages placed after them. |
| IAA-845 | Stopped a later usage calculation from replacing an earlier one for the same period with a less complete result. |
| IAA-869 | Fixed dashboard creation so a later request does not reuse the previous scope, a clarifying question reaches you, and only Save or Discard stores the preview. Dropped filters are named in the reply. |
| IAA-914 | Fixed dashboard confirmation so an unclear reply does not drop the preview, and Save applies to the preview you were offered. A reply that did not build a dashboard is not treated as one. |
| IAA-921 | The install job no longer turns evaluation-data seeding on by default. An installed cluster is not seeded automatically. |
| IAA-928 | Code samples in an answer are shown as code. Only graph descriptions are drawn as diagrams. |
| IAA-959 | When Save cannot find the preview, the chat now says the preview is no longer on the server and that navigating back inside the Dashboards app is the usual cause, instead of saying the preview expired. Opening the preview and going back still deletes it. See Known limitations. |
| IAA-971 | Aligned the JSON libraries so the app uses one supported Jackson release line. |
| IAA-981 | The conversation cost limit applies to the whole conversation. It is no longer reset on each turn. |
| IAA-984 | A dashboard reply shows the preview link once, and paragraph breaks in the reply are kept. |
| IAA-990 | When a severity-change lookup on one entity finds nothing and that entity is a parent, the same lookup is run again as a scope and returns the changes of the entities beneath it. The answer notes that it did so. A leaf entity with no changes stays empty. |
| IAA-991 | The earlier report that Insights stopped after resolving an entity did not reproduce on a later measurement. When an Insights answer calls no platform tool, it now ends with a line that no telemetry was read. |
| IAA-997 | Re-running an evaluation keeps the request-caching setting, and the report records whether caching was on. |
| IAA-998 | Re-running an evaluation no longer copies the previous report’s notes onto the new report. |
| IAA-1000 | Fixed the follow-up “When did this issue start?” answering about a different entity, or about none. |
| IAA-1027 | Metric charts plot each point at its own time, so the series is visible. |
| IAA-1028 | The log grid and severity timeline fit the chat panel, so the log text and each signal name stay visible. |
| IAA-1043 | Creating a dashboard retries when the draft is not valid, instead of stopping after one attempt. |
| IAA-1081 | Kubernetes evidence is gathered only when the entity is a Kubernetes object. A service with no Kubernetes identity, including a Geneos service, is not investigated with Kubernetes steps. See Perform root cause analysis. |
| IAA-1082 | Root cause analysis uses the time the entity’s severity changed as the start of the problem. A time range in the question is when you started looking. If that start time cannot be determined, the answer says so. |
| IAA-1085 | A follow-up such as “Find the errors in its logs” stays on the entity from the previous turn. The pronoun is not searched as an entity name. |
| IAA-1086 | When a question names a different entity, that name is resolved. The entity remembered from an earlier turn does not replace it. |
| IAA-1087 | The same row name on different hosts is kept as separate entities. It is no longer collapsed to a scope with no host. |
| IAA-1099 | A chat turn that runs longer than the sign-in token still saves its transcript, so the conversation can be reopened with its answer. |
| IAA-1101 | Renamed the Audit feature flag to ai.admin.audit.ui. An override that still uses the previous name must be updated. |
| IAA-1105 | A question such as “Why is this service red?” resolves the named entity and can run root cause analysis. It is no longer treated as a documentation question. |
| IAA-1106 | The entity tile and the status tile use the same name for the scope that was checked. |
| IAA-1107 | An answer about when a problem started includes log lines from the minutes around that instant. |
| IAA-1108 | A request to show named metrics for an entity returns the history and severity bands, instead of only the latest values. |
| IAA-1109 | Root cause analysis reads log lines from each relevant log source around the start of the problem, including a short deploy or configuration line. |
| IAA-1110 | Root cause analysis uses evidence already shown earlier in the conversation. |
| IAA-1111 | A change shortly before the problem starts, which explains the symptoms that follow, is treated as the likely cause. The resource symptom is how that change showed up, and the first suggested step is to undo the change. |
| IAA-1114 | A follow-up such as “Why is it red?” stays on the entity from the previous turn. |
| IAA-1128 | Dashboard tiles show data when the metric is stored in a different unit from the unit on the draft. |
| IAA-1137 | Support answers treat text that claims to change the assistant’s rules as content, not as an instruction. Ordinary requests about length, a table, or a language are unchanged. |
Security updates Copied
ITRS monitors publicly disclosed CVEs that may affect ITRS Analytics and its third-party and open source components. To learn more about how ITRS manages CVEs, see the ITRS Analytics Security Updates page, which also lists the recorded CVEs for each installation package.
The following security vulnerabilities have been addressed in this app release:
| Issue key | Release description |
|---|---|
| IAA-145 | Updated mcp-core to address high-severity vulnerability CVE-2026-35568. |
| IAA-606 | Updated Netty to address high-severity vulnerability CVE-2026-56816. |
| IAA-616 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-54876. |
| IAA-617 | Updated browserslist to address high-severity vulnerability BDSA-2026-26977. |
| IAA-618 | Updated httpcore5 to address high-severity vulnerability CVE-2026-54399. |
| IAA-619 | Updated httpcore5-h2 to address high-severity vulnerability CVE-2026-54428. |
| IAA-648 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-14456. |
| IAA-649 | Updated busybox to address high-severity vulnerability CVE-2026-38753. |
| IAA-650 | Updated busybox to address high-severity vulnerability CVE-2026-38754. |
| IAA-677 | Updated busybox to address high-severity vulnerability CVE-2026-38752. |
| IAA-678 | Updated busybox to address high-severity vulnerability CVE-2026-38755. |
| IAA-693 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-17153. |
| IAA-694 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-17158. |
| IAA-695 | Updated Netty to address high-severity vulnerability BDSA-2026-23495. |
| IAA-696 | Updated Netty to address high-severity vulnerability BDSA-2026-23581. |
| IAA-697 | Updated Netty to address high-severity vulnerability CVE-2026-56816. |
| IAA-698 | Updated Netty to address critical vulnerability CVE-2026-56817. |
| IAA-699 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-55831. |
| IAA-700 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-55833. |
| IAA-701 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-56745. |
| IAA-702 | Updated netty-codec-http3 to address high-severity vulnerability CVE-2026-56816. |
| IAA-703 | Updated netty-codec-http2 to address high-severity vulnerability CVE-2026-56819. |
| IAA-704 | Updated netty-codec-compression to address high-severity vulnerability CVE-2026-59901. |
| IAA-779 | Updated jline3 to address high-severity vulnerability CVE-2026-56740. |
| IAA-780 | Updated jline3 to address high-severity vulnerability CVE-2026-56741. |
| IAA-813 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-30244. |
| IAA-818 | Updated Netty to address high-severity vulnerability BDSA-2026-29111. |
| IAA-825 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-30271. |
| IAA-826 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-30278. |
| IAA-827 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-30279. |
| IAA-830 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-14457. |
| IAA-831 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-18798. |
| IAA-832 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-54874. |
| IAA-833 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-63072. |
| IAA-834 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-63075. |
| IAA-835 | Updated OpenSSL libraries to address high-severity vulnerability CVE-2026-63076. |
| IAA-858 | Updated OpenSSL libraries to address critical vulnerability CVE-2026-63073. |
| IAA-862 | Updated Apache Tomcat to address critical vulnerability CVE-2026-65182. |
| IAA-863 | Updated Apache Tomcat to address critical vulnerability CVE-2026-65905. |
| IAA-864 | Updated Apache Tomcat to address critical vulnerability CVE-2026-68525. |
| IAA-912 | Updated Apache HTTPComponents Core to address high-severity vulnerability BDSA-2026-29026. |
| IAA-913 | Updated jackson-core and jackson-databind to address high-severity vulnerability BDSA-2026-30990. |
| IAA-979 | Updated netty-handler to address critical vulnerability CVE-2026-75595. |
| IAA-1007 | Updated curl to address high-severity vulnerability CVE-2026-13608. |
| IAA-1008 | Updated curl to address critical vulnerability CVE-2026-18924. |
| IAA-1009 | Updated curl to address critical vulnerability CVE-2026-19931. |
| IAA-1010 | Updated curl to address high-severity vulnerability CVE-2026-80229. |
| IAA-1011 | Updated curl to address high-severity vulnerability CVE-2026-80230. |
| IAA-1012 | Updated curl to address high-severity vulnerability CVE-2026-80231. |
| IAA-1013 | Updated curl to address high-severity vulnerability CVE-2026-82208. |
| IAA-1014 | Updated curl to address high-severity vulnerability CVE-2026-82209. |
| IAA-1040 | Updated zlib to address high-severity vulnerability CVE-2026-85091. |
| IAA-1050 | Updated glibc to address high-severity vulnerability CVE-2026-19499. |
| IAA-1051 | Updated python-3.13 to address high-severity vulnerability CVE-2026-82049. |
| IAA-1126 | Updated jackson-databind to address high-severity vulnerability CVE-2026-91776. |
| IAA-1127 | Updated jackson-databind to address high-severity vulnerability CVE-2026-91777. |
| IAA-1141 | Updated source-map-js to address high-severity vulnerability CVE-2026-93749. |
Agentic SRE Beta 0.6.0 Copied
Released: 4 August 2026 Beta
This app is packaged with ITRS Analytics 2.18.8+p2.18.7.
Important
This release changes how the Agentic SRE Beta app is enabled. The app now requires an Agentic SRE license entitlement, and its toggle is hidden in the KOTS Admin Console when the entitlement is unavailable. If you previously enabled Agentic SRE, the selection is not carried over during the upgrade.
Before upgrading, obtain the entitlement and sync or upload the updated license. During or after the upgrade, enable Agentic SRE again under Beta Apps. For requirements and instructions, see Upgrading to ITRS Analytics 2.18.8+p2.18.7.
Highlights Copied
These are the highlights of this release:
- Expanded LLM provider support — Configure, test, and use Google Gemini models through the new Google Gen AI provider. Model lists for supported cloud providers are also updated automatically.
- More accurate entity-aware responses — Entity names in natural-language questions can now be resolved through embedded semantic search. Chat requests can also use multiple entity references and display inline references as context chips.
- Improved service readiness visibility — Administrators can review documentation index, entity index, and LLM readiness on the new auto-refreshing Status page, which is accessible from the chat status indicator.
- Enhanced evaluation and feedback workflows — Create evaluations through a guided wizard, run them with additional timeout and report safeguards, and provide thumbs-up or thumbs-down feedback on chat responses.
New features and enhancements Copied
These are the new features and enhancements of this release:
| Issue key | Release description |
|---|---|
| IAA-310 | Added embedded semantic search for resolving platform entities without requiring a separate search service. See Add entity context to a prompt. |
| IAA-345 | Improved Insights responses by resolving entity names mentioned in natural-language questions. See Investigate what is happening in your environment. |
| IAA-396 | Made it clearer how to rerun Test LLM after a successful or failed test. See Add an LLM. |
| IAA-397 | Improved LLM error messages for authentication, quota, and availability failures while protecting sensitive values. |
| IAA-455 | Automated updates to the available model lists for supported cloud LLM providers. |
| IAA-462 | Reduced unnecessary processing by skipping entity resolution when messages contain no entity reference. |
| IAA-468 | Added Google Gen AI provider support for configuring, testing, and using Gemini models. See Configure a Google GenAI model. |
| IAA-470 | Limited system-wide summaries to explicit overview requests and improved responses when an entity cannot be found. |
| IAA-471 | Improved entity index startup performance by avoiding duplicate processing. |
| IAA-497 | Updated Support guidance to reduce repeated documentation searches before responding. See Get product and setup guidance. |
| IAA-504 | Improved entity disambiguation by filtering weakly related search results. |
| IAA-505 | Improved routing of Insights requests and exposed the reasoning behind the selected response mode. See How to use the AI chat in the ITRS Analytics Web Console. |
| IAA-506 | Added support for natural-language and inline dimension references displayed as context chips in chat. |
| IAA-511 | Disabled the unused hosted models service while retaining an option to enable it. |
| IAA-519 | Added evaluation coverage for SRE routing accuracy, tool selection, and routed agent behavior. See Understand Eval results. |
| IAA-521 | Added an auto-refreshing Status page showing documentation index, entity index, and LLM readiness. |
| IAA-522 | Made the chat status indicator link administrators directly to the Status page. See Review service readiness in Status. |
| IAA-524 | Added timeouts, limited concurrency, and report finalization safeguards for evaluation runs. |
| IAA-528 | Added thumbs-up and thumbs-down feedback for chat responses, with feedback recorded in the audit history. See Provide feedback on an AI response. |
| IAA-544 | Added support for using multiple entity references as separate scopes in one chat request. |
| IAA-545 | Added context-name lookup when clicked entities are saved for later use in chat. |
| IAA-558 | Clarified the Add to chat control and added it directly to entity cards in chat responses. |
| IAA-559 | Added plain-language descriptions of each grader to the Details column. See Understand Eval results. |
| IAA-561 | Reworked evaluation creation into a guided, multi-step wizard. See Run an Eval campaign and review the results. |
| IAA-595 | Hid unsupported OpenAI models and added clear validation when such models are supplied manually. |
Issues fixed Copied
These are the issues we have fixed in this release:
| Issue key | Release description |
|---|---|
| IAA-369 | Fixed LLM save failures closing the editor without showing an error. |
| IAA-394 | Fixed entity index startup processing blocking entity updates and leaving the index incomplete or out of date. |
| IAA-402 | Fixed failed Test LLM checks appearing as complete in the Audit list. |
| IAA-421 | Fixed audit records for unsaved Test LLM configurations omitting the LLM name. |
| IAA-443 | Fixed the Model list showing models from the previously selected provider. |
| IAA-465 | Fixed failed entity resolution producing an unrelated system-wide summary. |
| IAA-489 | Fixed entity index startup delays when available processing resources were overestimated. |
| IAA-490 | Added configurable reasoning effort so supported OpenAI reasoning models can use function tools. |
| IAA-491 | Replaced raw OpenAI error details with concise, readable messages. |
| IAA-492 | Fixed documentation evaluations losing citation URLs and incorrectly receiving zero string-content scores. |
| IAA-494 | Fixed Test LLM failures for Anthropic models that require max_tokens or reject an explicitly supplied temperature. |
| IAA-496 | Fixed unsuitable default settings for Gemini models and added controls for model thinking levels and budgets. |
| IAA-525 | Prevented RCA chat responses from hanging indefinitely when an LLM stream does not end. |
| IAA-543 | Fixed evaluation runs hanging or linking to missing reports when grading, authentication, or report saving fails. |
| IAA-566 | Fixed System Status incorrectly reporting that no LLM is configured. |
| IAA-568 | Fixed uploads of evaluation reports larger than the gateway request limit. |
| IAA-569 | Fixed inapplicable grader results appearing as failed zero scores instead of N/A. |
| IAA-572 | Fixed Test LLM returning an internal server error when reasoning_effort is unsupported by the selected GPT-5 model. |
| IAA-573 | Fixed changes to Reasoning Effort overwriting the name of a copied LLM. |
| IAA-591 | Fixed unsupported OpenAI -pro and -codex models causing Add LLM and Test LLM failures. |
Security updates Copied
ITRS monitors publicly disclosed CVEs that may affect ITRS Analytics and its third-party and open source components. To learn more about how ITRS manages CVEs, see the ITRS Analytics Security Updates page, which also lists the recorded CVEs for each installation package.
The following security vulnerabilities have been addressed in this app release:
| Issue key | Release description |
|---|---|
| IAA-362 | Updated python-3.13 to address high-severity vulnerability CVE-2026-11940. |
| IAA-363 | Updated python-3.13 to address high-severity vulnerability CVE-2026-11972. |
| IAA-383 | Updated antlr to address high-severity vulnerability BDSA-2026-17031. |
| IAA-399 | Updated transformers to address high-severity vulnerability CVE-2026-4372. |
| IAA-423 | Updated curl to address critical vulnerability CVE-2026-10536. |
| IAA-424 | Updated curl to address high-severity vulnerability CVE-2026-11352. |
| IAA-425 | Updated curl to address critical vulnerability CVE-2026-11564. |
| IAA-426 | Updated curl to address high-severity vulnerability CVE-2026-11586. |
| IAA-427 | Updated curl to address critical vulnerability CVE-2026-11856. |
| IAA-428 | Updated curl to address high-severity vulnerability CVE-2026-12064. |
| IAA-429 | Updated libexpat1 to address high-severity vulnerability CVE-2026-56131. |
| IAA-430 | Updated libexpat1 to address high-severity vulnerability CVE-2026-56407. |
| IAA-431 | Updated libexpat1 to address high-severity vulnerability CVE-2026-56408. |
| IAA-432 | Updated curl to address high-severity vulnerability CVE-2026-8286. |
| IAA-433 | Updated curl to address critical vulnerability CVE-2026-8924. |
| IAA-434 | Updated curl to address critical vulnerability CVE-2026-8925. |
| IAA-435 | Updated curl to address critical vulnerability CVE-2026-8926. |
| IAA-436 | Updated curl to address critical vulnerability CVE-2026-8927. |
| IAA-437 | Updated curl to address high-severity vulnerability CVE-2026-8932. |
| IAA-438 | Updated curl to address critical vulnerability CVE-2026-9079. |
| IAA-439 | Updated curl to address high-severity vulnerability CVE-2026-9080. |
| IAA-440 | Updated curl to address high-severity vulnerability CVE-2026-9545. |
| IAA-441 | Updated curl to address high-severity vulnerability CVE-2026-9546. |
| IAA-442 | Updated curl to address high-severity vulnerability CVE-2026-9547. |
| IAA-447 | Updated the Go standard library to address high-severity vulnerability CVE-2026-39822. |
| IAA-456 | Updated python-3.13 to address high-severity vulnerability CVE-2026-15308. |
| IAA-467 | Updated transformers to address high-severity vulnerability CVE-2026-5241. |
| IAA-486 | Updated Apache Tomcat to address critical vulnerability CVE-2026-59083. |
| IAA-487 | Updated Apache Tomcat to address critical vulnerability CVE-2026-59084. |
| IAA-509 | Updated Eclipse Vert.x to address high-severity vulnerability CVE-2026-15075. |
| IAA-510 | Updated Eclipse Vert.x to address high-severity vulnerability CVE-2026-15076. |
| IAA-562 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-55831. |
| IAA-563 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-55833. |
| IAA-564 | Updated netty-codec-http to address high-severity vulnerability CVE-2026-56745. |
| IAA-565 | Updated netty-codec to address high-severity vulnerability CVE-2026-59901. |
| IAA-587 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-17153. |
| IAA-588 | Updated Apache Tomcat to address high-severity vulnerability BDSA-2026-17158. |
| IAA-589 | Updated fast-uri to address high-severity vulnerability BDSA-2026-23246. |
Agentic SRE Beta 0.5.0 Copied
Released: 23 July 2026 Beta
This app is packaged with ITRS Analytics 2.18.7+p2.18.6.
New features and enhancements Copied
These are the new features and enhancements of this release:
| Issue key | Release description |
|---|---|
| IAA-92 | Updated documentation indexing so newly added product repositories are included automatically, with support for filtering repositories as needed. |
| IAA-238 | Updated dependencies to support applying securityContext.applySecurityContextDefaults from instance-config for OpenShift deployments. |
| IAA-262 | Improved the Run Eval test-loading workflow with a multi-group picker, question preview, live filtering, and append-with-deduplication behavior. |
| IAA-263 | Replaced evaluation test groups with tags and reorganized the evaluation protobuf definitions to support more flexible filtering and a cleaner API structure. |
| IAA-293 | Masked API key and secret fields in the LLM editor by default and added a reveal toggle for secure on-demand viewing. |
Security updates Copied
ITRS monitors publicly disclosed CVEs that may affect ITRS Analytics and its third-party and open source components. To learn more about how ITRS manages CVEs, see the ITRS Analytics Security Updates page, which also lists the recorded CVEs for each installation package.
The following security vulnerabilities have been addressed in this app release:
| Issue key | Release description |
|---|---|
| IAA-204 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2026-7210. |
| IAA-232 | Updated axios to address vulnerability BDSA-2026-11999. |
| IAA-251 | Updated golang.org/x/net in the hosted models image to address vulnerability CVE-2026-33814. |
| IAA-252 | Updated golang.org/x/net in the hosted models image to address vulnerability CVE-2026-39821. |
| IAA-253 | Updated Go standard library packages in the hosted models image to address vulnerability CVE-2026-42504. |
| IAA-257 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2026-3644. |
| IAA-258 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2026-4224. |
| IAA-259 | Updated binutils in the hosted models image to address vulnerability CVE-2026-6846. |
| IAA-260 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2025-13462. |
| IAA-265 | Updated netty-handler across Agentic SRE service images to address vulnerability CVE-2026-44249. |
| IAA-266 | Updated netty-handler across Agentic SRE service images to address vulnerability CVE-2026-45416. |
| IAA-267 | Updated netty-resolver-dns in the UI service to address vulnerability CVE-2026-45674. |
| IAA-268 | Updated netty-resolver-dns in the UI service to address vulnerability CVE-2026-47691. |
| IAA-271 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2026-9669. |
| IAA-275 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-34180. |
| IAA-276 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-34181. |
| IAA-277 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-34182. |
| IAA-278 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-34183. |
| IAA-279 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-42764. |
| IAA-280 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-42765. |
| IAA-281 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-45445. |
| IAA-282 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-45447. |
| IAA-283 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-7383. |
| IAA-284 | Updated OpenSSL libraries in the hosted models image to address vulnerability CVE-2026-9076. |
| IAA-291 | Updated starlette in the documentation indexing image to address vulnerability CVE-2026-48818. |
| IAA-292 | Updated starlette in the documentation indexing image to address vulnerability CVE-2026-54283. |
| IAA-296 | Updated form-data to address vulnerability CVE-2026-12143. |
| IAA-360 | Updated Keycloak to address vulnerability BDSA-2026-16693. |
| IAA-361 | Updated Keycloak to address vulnerability BDSA-2026-16698. |
| IAA-372 | Updated Kotlin to address vulnerability CVE-2026-53914. |
| IAA-384 | Updated Apache Tomcat to address vulnerability BDSA-2026-17153. |
| IAA-385 | Updated Apache Tomcat to address vulnerability BDSA-2026-17158. |
Agentic SRE Beta 0.4.1 Copied
Released: 29 June 2026 Beta
This app is packaged with ITRS Analytics 2.18.4+p2.18.5.
New features and enhancements Copied
These are the new features and enhancements of this release:
| Issue key | Release description |
|---|---|
| IAA-238 | Updated dependencies to support applying securityContext.applySecurityContextDefaults from instance-config for OpenShift deployments. |
Agentic SRE Beta 0.4.0 Copied
Released: 11 June 2026 Beta
This app is packaged with ITRS Analytics 2.18.2+p2.18.5.
New features and enhancements Copied
These are the new features and enhancements of this release:
| Issue key | Release description |
|---|---|
| IAA-144 | Updated Python 3.11 in the documentation indexing image to address vulnerability CVE-2026-3298. |
| IAA-146 | Updated lxml in the documentation indexing image to address vulnerability CVE-2026-41066. |
| IAA-147 | Updated glibc across Agentic SRE service images to address vulnerability CVE-2026-5928. |
| IAA-151 | Updated the hosted models image to the latest patched Git package build to address vulnerability CVE-2026-32631. |
| IAA-152 | Updated Python 3.11 base packages in the documentation indexing image to address vulnerability CVE-2026-4786. |
| IAA-153 | Updated glibc across Agentic SRE service images to version 2.43-r7 to address vulnerability CVE-2026-5435. |
| IAA-154 | Updated Python 3.11 base packages in the documentation indexing image to address vulnerability CVE-2026-6100. |
| IAA-157 | Updated axios to address vulnerability BDSA-2026-6792. |
| IAA-158 | Updated DOMPurify to address vulnerability BDSA-2026-8016. |
| IAA-159 | Updated DOMPurify to address vulnerability BDSA-2026-8017. |
| IAA-160 | Updated DOMPurify to address vulnerability BDSA-2026-8018. |
| IAA-167 | Updated core libc packages across Agentic SRE service images to address vulnerability CVE-2026-5450. |
| IAA-168 | Updated OpenTelemetry Java instrumentation to address vulnerability BDSA-2026-5716. |
| IAA-171 | Updated Netty dependencies to address vulnerability BDSA-2026-9430. |
| IAA-172 | Updated Netty dependencies to address vulnerability BDSA-2026-9433. |
| IAA-173 | Updated netty-codec-dns in the UI service to address vulnerability CVE-2026-42579. |
| IAA-174 | Updated netty-codec across Agentic SRE service images to address vulnerability CVE-2026-42583. |
| IAA-175 | Updated netty-codec-http across Agentic SRE service images to address vulnerability CVE-2026-42584. |
| IAA-176 | Updated netty-codec-http2 and related HTTP codec dependencies across Agentic SRE service images to address vulnerability CVE-2026-42587. |
| IAA-177 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-33811. |
| IAA-178 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-33814. |
| IAA-179 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-39820. |
| IAA-180 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-39836. |
| IAA-181 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-42499. |
| IAA-182 | Updated the Go standard library in the hosted models image to address vulnerability CVE-2026-42501. |
| IAA-206 | Updated Spring AI chat client and model dependencies across Agentic SRE service images to address vulnerability CVE-2026-41712. |
| IAA-207 | Updated the Spring AI chat client across Agentic SRE service images to address vulnerability CVE-2026-41713. |
Agentic SRE Beta 0.3.0 Copied
Released: 6 May 2026 Beta
This app is packaged with ITRS Analytics 2.18.0+p2.18.3.
Issues fixed Copied
These are the issues we have fixed in this release:
| Issue key | Release description |
|---|---|
| N/A | Agentic SRE Beta app 0.3.0 is a maintenance release focused on stability and dependency updates for ITRS Analytics Platform 2.18.0. It also improved overall reliability and stability, including a fix for a chat UI issue where settings could repeatedly be saved when no LLM was configured, which could cause the browser to become unresponsive. |
Agentic SRE Beta 0.2.0 Copied
Released: 1 April 2026 Beta
This app is packaged with ITRS Analytics 2.17.11.
Version 0.2.0 marks the Beta release of the ITRS Agentic SRE app. For more information, see the Use Agentic SRE for operations and support in ITRS Analytics documentation. This app release requires a minimum ITRS Analytics Platform version of 2.17.0 and Web Console version 3.8.0 to operate.
Release highlights Copied
The following highlights summarize what is included in this release:
- ITRS Analytics Agentic SRE Beta app 0.2.0 introduces usability and control improvements, including:
- stronger role-based access control (RBAC) for protected admin operations;
- the ability to stop a request while it is still processing; and
- a Test button in the LLM editor to help admins validate model behavior more easily.
- Reliability and UX consistency were improved by work on request cancellation behavior, clearer backend error propagation to the UI, and persistence and functional test fixes.
- Platform and architecture improvements were added, including architecture standardization and migration to the newer Web Platform Micro-Frontend (MFE) core.
- Operational and security hardening included dependency CVE fixes and adding standard Kubernetes labels to workloads.
Disclaimer
The information contained in this document is for general information and guidance on our products, services, and other matters. It is only for information purposes and is not intended as advice which should be relied upon. We try to ensure that the content of this document is accurate and up-to-date, but this cannot be guaranteed. Changes may be made to our products, services, and other matters which are not noted or recorded herein. All liability for loss and damage arising from reliance on this document is excluded (except where death or personal injury arises from our negligence or loss or damage arises from any fraud on our part).