What's Changed
- ROB-1624 Notify Slack when a release workflow fails by @Avi-Robusta in #2606
- Add allowed_tools to MCP servers to load only some of a server's tools by @elidrissi-amine in #2598
- feat(alert): add webhook destination by @mrchocha in #2612
- ROB-616: Migrate AWS MCP to the hosted AWS MCP Server by @Avi-Robusta in #2529
- Add notify input to eval-benchmarks workflow for quiet manual runs by @moshemorad in #2627
- ROB-1490 Cleanup repeated Holmes server logs by @moshemorad in #2608
- ROB-1544 Stop storing a JSON-string copy of messages on ai_answer_end by @naomi-robusta in #2630
- ROB-1290 Send the conversation ID to relay on Robusta-hosted completions by @RoiGlinik in #2567
- Weekly Benchmark Results 2026-10-21_07-21 by @github-actions[bot] in #2635
- FRO-714 Build the docs with mkdocs-material 9.7.7 and add a manual job that rebuilds every published version by @ezra-robusta in #2615
- ROB-1606 Keep the sanitized cause on failed runs by @naomi-robusta in #2596
- ROB-1707 Remove follow_up_actions from the chat API by @naomi-robusta in #2651
- ROB-1623 Switch the eval judge to gpt-6.1-sol-medium and add a judge benchmark skill by @Avi-Robusta in #2658
- ROB-1722: Remove deprecated qwen-next-80B-instruct from the default benchmark models by @Sheeproid in #2662
- ROB-1544 Publish a contract hash per remote tool and the Holmes version per toolset row by @naomi-robusta in #2624
- ROB-1552 Retry LLM rate-limit, overload and 503 errors with backoff instead of failing the run by @naomi-robusta in #2559
- Sanitize tests by @Avi-Robusta in #2668
- ROB-1735 Freeze the SigV4 clock in the AWS MCP signing test by @Sheeproid in #2664
- ROB-1403 Report each model's limits and their source in the Holmes status by @RoiGlinik in #2656
- ROB-1659 Clarify the k8s MCP server Dex setup guide by @RoiGlinik in #2667
- ROB-1578: Forward toolset subtype to multi-instance children by @Avi-Robusta in #2657
- FRO-630 Link the custom toolsets page relatively from the transformers page by @ezra-robusta in #2680
- ROB-1711 Fix the weekly benchmark's broken evals, harness and reporting by @Sheeproid in #2679
- Weekly Benchmark Results 2026-10-28_10-09 by @github-actions[bot] in #2677
- ROB-1749 Drop gpt-5.5 from the default benchmark models by @Sheeproid in #2683
- ROB-1663 Remote tool calls: keep the images a tool returned in the payload by @naomi-robusta in #2669
- ROB-1679: Log success and error status for tool calls by @Avi-Robusta in #2674
- ROB-1675: Helm values and docs for remediation MCP exec argument allowlists by @Avi-Robusta in #2682
- ROB-1503 Point single-cluster Kubernetes MCP setups at the built-in toolset by @ezra-robusta in #2648
- ROB-1598 Hash the rendered toolset configmap in the Holmes pod checksum by @ezra-robusta in #2647
- FRO-630 docs: fix data source and model provider steps that fail or mislead when followed by @ezra-robusta in #2637
- FRO-599 Document Robusta AI settings, proxies and LLM provider hosts by @ezra-robusta in #2644
- FRO-618 docs: write the deployment tab groups as deployment fences by @ezra-robusta in #2638
- FRO-630 docs: match the Holmes CLI and Helm tabs of the Rovo, OAuth and database groups, and say where the Helm toolset is enabled by default by @ezra-robusta in #2639
- FRO-620 docs: check deployment fences against the Holmes chart and the Kubernetes schemas by @ezra-robusta in #2640
- FRO-620 docs: check what deployment fences pass to Holmes with Holmes's own loaders by @ezra-robusta in #2641
- FRO-630 docs: fix setup steps that fail or mislead on the data source and model provider pages by @ezra-robusta in #2642
- ROB-1738 Fix CVEs in the holmes image by @moshemorad in #2675
- ROB-1743 Don't fetch the Supabase key from SaaS for a self-hosted store by @Avi-Robusta in #2690
- ROB-1795 Serve all AWS accounts from one hosted AWS MCP toolset by @Avi-Robusta in #2694
New Contributors
- @elidrissi-amine made their first contribution in #2598
Full Changelog: 0.45.0...0.46.0-alpha