You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have verified that my UTMStack version is up-to-date
Describe the bug
The Anthropic provider selector in the SOC-AI integration guide offers three predefined model IDs that are no longer available in the Anthropic public API. Any alert parsed with the default selection fails in the Anthropic messaging API with a 4xx error, and the user interface displays the message "The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
Current dropdown value
Real status at Anthropic
claude-sonnet-4-20250514
Retired on 2026-06-15 (deprecated 2026-04-14). Requests fail.
claude-opus-4-20250514
Retired on 2026-06-15 (deprecated 2026-04-14). Requests fail.
claude-haiku-4-20250414
Not in current models, legacy list, or deprecations.
Sonnet and Opus worked until Anthropic retired those snapshots on 2026-06-15
Regression Issue
Select this option if this issue appears to be a regression.
Expected Behavior
The Anthropic provider dropdown should list only model IDs that are currently active in the Anthropic Messages API. Selecting any option from the dropdown, saving the configuration, and running an analysis on an alert should successfully return a classification from Claude, with no "processing error" shown in the UI.
Ideally the dropdown should use Anthropic's stable, dateless API aliases (e.g. claude-opus-4-8, claude-sonnet-5, claude-haiku-4-5) so that this class of bug does not recur silently when Anthropic retires the underlying snapshot. Aliases keep resolving to a supported model when the snapshot changes; dated IDs do not.
The exact list can be adjusted — the requirement is that every hardcoded value must be a currently active API alias or model ID listed at https://docs.claude.com/en/docs/about-claude/models/overview at the time of the fix. The Custom model... option must remain available for operators who need to pin a specific dated snapshot.
Optional UI improvement: update the Anthropic provider description in the same file to point operators to Anthropic's model overview page, so they know they can pick Custom model... if they need something newer than the hardcoded list.
Current Behavior
When an operator selects any of the three preloaded Anthropic models and triggers an alert analysis, the SOC-AI plugin sends a POST to https://api.anthropic.com/v1/messages with an invalid model ID.
The SOC-AI plugin retries three times, then marks the alert as failed in the v11-soc-ai OpenSearch index (status: Error).
The SOC-AI panel in the alert-management UI then renders the generic error block:
"Error! The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
Reproduction Steps
Deploy or run any UTMStack build that includes the current SOC-AI multi-provider guide (introduced in PR Release/v11.2.6 #2025).
Log in as an admin and open Integrations → SOC-AI configuration guide.
In the Model dropdown, pick any of the three preloaded options:
claude-sonnet-4-20250514
claude-haiku-4-20250414
claude-opus-4-20250514
Set Max Tokens to 4096 (or any positive value), save the configuration, and enable auto-analyze (or leave manual).
Go to Alerts, open any alert, and either wait for auto-analysis or click "Analyze with SOC AI".
Observe the SOC-AI panel for that alert.
Observed result: The panel shows the red error block:
"Error! The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
Expected result: The panel should show a Claude-produced classification, reasoning, and next steps.
Workaround (until the fix lands): in the same dropdown, choose Custom model... and manually enter a currently active Anthropic model ID such as claude-opus-4-8, claude-sonnet-5, or claude-haiku-4-5.
Acknowledgements
Describe the bug
The Anthropic provider selector in the SOC-AI integration guide offers three predefined model IDs that are no longer available in the Anthropic public API. Any alert parsed with the default selection fails in the Anthropic messaging API with a 4xx error, and the user interface displays the message "The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
claude-sonnet-4-20250514claude-opus-4-20250514claude-haiku-4-20250414Sonnet and Opus worked until Anthropic retired those snapshots on 2026-06-15
Regression Issue
Expected Behavior
The Anthropic provider dropdown should list only model IDs that are currently active in the Anthropic Messages API. Selecting any option from the dropdown, saving the configuration, and running an analysis on an alert should successfully return a classification from Claude, with no "processing error" shown in the UI.
Ideally the dropdown should use Anthropic's stable, dateless API aliases (e.g.
claude-opus-4-8,claude-sonnet-5,claude-haiku-4-5) so that this class of bug does not recur silently when Anthropic retires the underlying snapshot. Aliases keep resolving to a supported model when the snapshot changes; dated IDs do not.Suggested new options for the Anthropic block:
The exact list can be adjusted — the requirement is that every hardcoded value must be a currently active API alias or model ID listed at https://docs.claude.com/en/docs/about-claude/models/overview at the time of the fix. The
Custom model...option must remain available for operators who need to pin a specific dated snapshot.Optional UI improvement: update the Anthropic provider
descriptionin the same file to point operators to Anthropic's model overview page, so they know they can pickCustom model...if they need something newer than the hardcoded list.Current Behavior
When an operator selects any of the three preloaded Anthropic models and triggers an alert analysis, the SOC-AI plugin sends a POST to https://api.anthropic.com/v1/messages with an invalid model ID.
The SOC-AI plugin retries three times, then marks the alert as failed in the
v11-soc-aiOpenSearch index (status: Error).The SOC-AI panel in the alert-management UI then renders the generic error block:
"Error! The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
Reproduction Steps
4096(or any positive value), save the configuration, and enable auto-analyze (or leave manual).Observed result: The panel shows the red error block:
"Error! The SOC-AI integration could not analyze this alert due to a processing error. Please try again."
Expected result: The panel should show a Claude-produced classification, reasoning, and next steps.
Workaround (until the fix lands): in the same dropdown, choose
Custom model...and manually enter a currently active Anthropic model ID such asclaude-opus-4-8,claude-sonnet-5, orclaude-haiku-4-5.Possible Solution
No response
Additional Information/Context
No response
UTMStack Version
v11.2.12
Operating System and version
Ubuntu 24
Hypervisor and Version | Server Vendor and Model
dev-env
Browser and version
Chrome