v1.77.0

New Features

• c3c5284 - gooddata-eval: record why an agentic simulated-user loop stopped (PR #1789 by @Tomkess)
• 1daa2c4 - gooddata-eval: add the agentic anomaly-detection evaluator (PR #1801 by @Tomkess)
• 4fa3949 - gooddata-eval: recognize the report answer part
• aaba134 - gooddata-eval: add the agentic report-skill evaluator
• cb44617 - gooddata-eval: judge the narrative of a drafted report
• 8d8a6fe - gooddata-eval: relay user_context on every agentic kind (commit by @myhoai)
• 38fa9d4 - gooddata-eval: report whether the agent ran the chart it built (commit by @myhoai)
• afb8b79 - gooddata-eval: label gen-ai traces through W3C baggage (commit by @tychtjan)
• 8168136 - eval: capture the detail of every failing run, not just the winning one (PR #1816 by @Tomkess)
• 92fe97e - gooddata-eval: check that gen-ai sends the Document wire names
• a540bcd - gooddata-eval: add --tag to put custom tags on Langfuse traces (commit by @janmatzek)
• 8cefcde - gooddata-eval: put gen-ai turns under the eval item root trace (commit by @tychtjan)
• 524f8d1 - gooddata-eval: keep the SSE error reason on every ChatError (commit by @tychtjan)
• f62aba3 - gooddata-eval: send an item's scores as one Langfuse request (commit by @tychtjan)
• e8eaa31 - gooddata-eval: name gooddata-eval in Langfuse requests and spans (commit by @tychtjan)
• 7fcae3a - gooddata-eval: score v2 Publisher documents

Bug Fixes

• a902b08 - gooddata-eval: log the adhoc-visualization fallback (commit by @myhoai)
• 8152217 - gooddata-eval: delete a metric again when a concurrent write restores it (commit by @myhoai)
• ab84a23 - gooddata-eval: let --gate decide a what-if item, not pass@K alone (PR #1831 by @Tomkess)
• ba0494d - gooddata-eval: wait out the Langfuse rate-limit window on a 429 (commit by @tychtjan)
• 1ec1797 - gooddata-eval: keep sending score batches after one fails (commit by @tychtjan)

Refactors

• 83473b8 - gooddata-eval: name the report skill evaluator for documents

Tests

• e04c1a2 - gooddata-eval: use the real report document in the report-part test
• a1ae1c2 - enable pytest strict mode (commit by @hkad98)
• e22419e - gooddata-eval: extend evaluator for dashboard builder skill (commit by @huyenthanh09)

Chores

• 46daa3a - ask-ai: disable Kapa anonymous user-tracking cookie (commit by @JKgooddata)
• 69269ec - gooddata-api-client: regenerate against staging (commit by @Vojtasii)

v1.76.0

New Features

• 446e774 - gooddata-eval: report what a run actually did (commit by @xvalovic)
• b0135f4 - gooddata-eval: generate eval datasets from a workspace's insights (commit by @xvalovic)
• d12931e - gooddata-eval: evaluate the agentic dashboard-creation skill (commit by @FrankHuynh)
• 9be051b - gooddata-eval: add the agentic what-if-analysis evaluator (PR #1799 by @Tomkess)
• 97ec295 - gooddata-eval: score dashboard editing alongside creation (commit by @FrankHuynh)
• 0e6c832 - gooddata-api-client: add missing workspace permissions (commit by @jeskepetr)
• f9b198e - gooddata-eval: grade conversations on keeping context (commit by @myhoai)

Bug Fixes

• 7f08a54 - gooddata-eval: stop transient gen-ai faults from being scored as content failures (PR #1793 by @Tomkess)
• c73f047 - gooddata-eval: compare attribute-filter elements as a set, not a sequence (PR #1806 by @Tomkess)
• 1ab844c - gooddata-eval: a refusal that offers a safe alternative is a pass (AIS-52) (PR #1807 by @Tomkess)
• 26f2813 - gooddata-eval: do not retry a turn that hit the iteration limit (commit by @myhoai)
• 4a64094 - gooddata-eval: keep metrics the agent only updated during cleanup (commit by @myhoai)

Refactors

• b58ed4e - gooddata-eval: type the generated visualization spec (commit by @xvalovic)

Tests

• c66ac11 - gooddata-eval: cover the clarification judge and simulated user (commit by @myhoai)
• ee39a75 - gooddata-eval: annotate the conversation tests, drop the openai patches (commit by @myhoai)

Chores

• 782acee - gooddata-eval: use neutral identifiers in fixtures, drop a hardcoded path (PR #1827 by @Tomkess)

v1.75.0

New Features

• d61e7a7 - gooddata-eval: extend detail.latency_breakdown to 5 remaining kinds (commit by @Tomkess)
• 79a3d1d - gooddata-eval: register knowledge_question evaluator (commit by @Tomkess)
• 6b9d690 - gooddata-eval: add Langfuse OTLP/experiment foundations and LANGFUSE_BASE_URL (commit by @tychtjan)
• 89d8a89 - gooddata-eval: read Langfuse traces via v2 observations and move the client to core/langfuse (commit by @tychtjan)
• b12210b - gooddata-eval: link agentic runs as Langfuse experiments via OTLP root spans (commit by @tychtjan)
• b47cf1a - gooddata-eval: pass question, conversation and output into experiment spans from every agentic kind (commit by @tychtjan)
• 6488fe0 - gooddata-eval: write single-shot results as Langfuse experiments over OTLP (commit by @tychtjan)
• 334faff - gooddata-eval: make the K verdict a gate, pass@K or pass^K (commit by @myhoai)
• 1c1ae20 - gooddata-api-client: add the CREATE_METRIC workspace permission (commit by @Vojtasii)
• fa82a6a - gooddata-eval: record the turns and steps each agentic run took (commit by @myhoai)

Bug Fixes

• 9aa8d86 - gooddata-eval: return a copy of activated_skills from TurnResult.detail() (commit by @Tomkess)
• 7f04d92 - address review findings on judge faults, run counts and trace linking (commit by @tychtjan)
• 3f9b3f8 - gooddata-eval: persist skill_routing credit across conversation turns (commit by @Tomkess)
• 26d4f52 - gooddata-eval: track set_skills as replacing, not accumulating (commit by @Tomkess)
• 2d64613 - gooddata-eval: keep only the final set_skills declaration per turn (commit by @Tomkess)
• 3e1d84d - gooddata-eval: count skills replaced within a turn toward coverage (commit by @Tomkess)
• 3bd2a29 - gooddata-eval: assert alert group-by attributes (commit by @tychtjan)
• 57aa25f - gooddata-eval: three false-negative sources in eval scoring (commit by @myhoai)
• f5703ce - gooddata-eval: accept a string-encoded rejectedSpans count in OTLP export responses (commit by @tychtjan)
• 18073f8 - gooddata-eval: fall back to the default score retry delay for negative or NaN Retry-After (commit by @tychtjan)
• a7ca20b - gooddata-eval: restore the dataset_run_items compat shim and harden the Langfuse score path (commit by @tychtjan)
• 76470ba - gooddata-eval: keep gate out of the positional signature and in the merged report (commit by @myhoai)
• 454286c - gooddata-eval: report the gate verdict without redefining pass@K (commit by @myhoai)

Performance Improvements

• 1f9babd - take Langfuse trace linking off the eval item critical path (commit by @tychtjan)
• 8125426 - cut gooddata_sdk import cost by importing only what the SDK uses (commit by @hkad98)

Refactors

• 0aa2b77 - gooddata-eval: derive TurnResult.detail() from the model, not a key list (commit by @Tomkess)
• 55e22cd - gooddata-api-client: split LDM object permissions per family (commit by @jeskepetr)
• dc79848 - gooddata-eval: remove the legacy Langfuse ingestion and dataset-run code paths (commit by @tychtjan)
• 99935f9 - gooddata-eval: count turns and steps as int, not float (commit by @myhoai)

Tests

• 316c88a - gooddata-eval: use the real set_skills argument key in conversation tests (commit by @Tomkess)
• bbe1187 - gooddata-eval: add an in-process fake Langfuse server for wire-level tests (commit by @tychtjan)
• 7d8155e - gooddata-eval: cover the Langfuse v4 request choreography against a fake server (commit by @tychtjan)

Chores

• 279dc54 - repo: replace AIDA MCP with tracked AGENTS.md and CLAUDE.md (commit by @tychtjan)

v1.74.0

New Features

• 0dadee6 - gooddata-eval: return AgenticEvalOutcome from evaluate_agentic_guardrail (commit by @Tomkess)
• 2b228c7 - gooddata-eval: return AgenticEvalOutcome from evaluate_agentic_search_tool (commit by @Tomkess)
• 01514ab - gooddata-eval: return AgenticEvalOutcome from evaluate_agentic_general_question (commit by @Tomkess)
• 19dd575 - gooddata-eval: return AgenticEvalOutcome from evaluate_agentic_visualization (commit by @Tomkess)
• 55a3c58 - gooddata-eval: return AgenticEvalOutcome from evaluate_agentic_kda_skill (commit by @Tomkess)
• f25f9fd - support SMTP, default SMTP and in-platform notification channels (commit by @hkad98)

Bug Fixes

• a2beff6 - gooddata-eval: keep retrying create_metric result, not the first failed one (commit by @FrankHuynh)
• a7da556 - gooddata-eval: scope simulated-user pushback to the original request (commit by @FrankHuynh)
• 76344dd - accept every oneOf variant in generated composed models (commit by @hkad98)

v1.73.0

New Features

• cbd27a1 - add new granularities (commit by @okrcmaGD)
• 91a45e3 - add new granularities to gooddata-dbt and gooddata-pipelines (commit by @okrcmaGD)
• 8f83a9e - gooddata-eval: add KDA-skill agentic evaluator (commit by @FrankHuynh)
• 4585e38 - gooddata-eval: support targeting a specific AI Hub agent (commit by @Tomkess)
• f92bc13 - gooddata-eval: capture agent reasoning steps in ChatResult (commit by @Tomkess)
• 7d4496f - gooddata-eval: capture reasoning_steps through the agentic-CLI path (commit by @Tomkess)
• 2b1e3ad - gooddata-eval: capture conversation_id/response_id through the agentic-CLI path (commit by @Tomkess)
• 07d4a96 - gooddata-eval: report compared filters in visualization evaluation detail (commit by @Tomkess)

Bug Fixes

• a84b842 - gooddata-eval: stop alert sim-user drifting trigger and filters (commit by @myhoai)
• 6dd0601 - adapt gooddata-sdk to regenerated api-client (commit by @okrcmaGD)
• 931817c - gooddata-eval: retry KDA on any non-triggering response, drop text classification (commit by @FrankHuynh)
• 208f2e0 - gooddata-eval: set temperature=0 for all simulated users (commit by @myhoai)
• 626ff3f - eval: stop guessing whether the agent asked a question (commit by @myhoai)
• 4e1cbda - gooddata-eval: stop the metric-skill simulated user from dropping MAQL clauses (commit by @Tomkess)
• fd49225 - satisfy ruff format/lint on CI (commit by @Tomkess)
• abe062e - gooddata-eval: check internal_recipients in alert recipients comparison (commit by @Tomkess)
• ba7bd07 - gooddata-eval: escape email before interpolating into RSQL filter (commit by @Tomkess)
• 94dddaa - gooddata-eval: make MAQL comparison case-insensitive for keywords (commit by @Tomkess)
• b946807 - gooddata-eval: avoid sharing mutable filter lists across missing-viz results (commit by @Tomkess)

Performance Improvements

• f896ed9 - gooddata-eval: resolve internal recipient emails in one bulk request (commit by @Tomkess)

Refactors

• 60e8b58 - remove support for Greenplum (commit by @jeskepetr)
• 504b44d - remove support for Drill (commit by @jeskepetr)
• 8e7d622 - gooddata-eval: replace tuple return with AgenticEvalOutcome dataclass (commit by @Tomkess)

Tests

• ecf8718 - gooddata-eval: assert exact MAQL string appears in simulated-user prompt (commit by @Tomkess)
• f2764ac - cover agent_id threading for every agentic test kind (commit by @Tomkess)
• fca3fe1 - pass mock sdk in external-recipients fast-path test (commit by @Tomkess)
• d8d656a - gooddata-eval: cover multi-run reasoning_steps retention (commit by @Tomkess)

Chores

• 3a4164d - regenerate api-client against staging (commit by @okrcmaGD)
• 049bc4a - bump requests to >= 2.33.0 (commit by @no23reason)

v1.72.0

New Features

• 49949a5 - gooddata-eval: let evaluations request a reasoning effort (commit by @hkad98)
• 640d740 - gooddata-sdk: report whether create_or_update created or updated (commit by @vondravl)

Bug Fixes

• e8f35dd - gooddata-eval: treat alert proposal part as confirmation signal (commit by @myhoai)
• b1d5012 - gooddata-eval: drop AFM execution even when it is the only alert field (commit by @myhoai)
• 56cbe09 - ci: bound uv in the tox group so the test image keeps 0.11.x (commit by @myhoai)
• c265cc7 - gooddata-eval: make ranking attribute optional on 1-dim viz (commit by @myhoai)

Chores

• b8993c4 - gooddata-sdk: make UpsertOutcome str() match StrEnum (commit by @vondravl)

v1.71.0

New Features

• 0bd8938 - gooddata-eval: report created metric/automation id from single-turn evaluators (commit by @Tomkess)

Bug Fixes

• e822104 - handle condition waiting better in thread_task_executor (commit by @no23reason)
• 818cdff - prevent potential lock leak in server_methods (commit by @no23reason)
• 295b9e9 - gooddata-pandas: satisfy ty on dynamic pyarrow.compute members (commit by @no23reason)
• 513f430 - guard non-mapping nested result in _extract_automation_id (commit by @Tomkess)

Performance Improvements

• 2776521 - gooddata-pandas: vectorize attribute value type conversion (commit by @no23reason)
• 491379f - gooddata-pandas: resolve catalog attributes once per extract, not per page (commit by @no23reason)
• 3528c88 - gooddata-pandas: parse arrow schema metadata once per build (commit by @no23reason)
• 00f4227 - gooddata-pandas: vectorize grand-total row handling (commit by @no23reason)
• fd700cb - gooddata-pandas: return arrow metric columns as numpy (commit by @no23reason)
• 2c10273 - gooddata-pandas: vectorize arrow date attribute conversion (commit by @no23reason)
• c3abeba - gooddata-pandas: build crosstab primary-label maps from distinct values (commit by @no23reason)

Tests

• 913ab07 - gooddata-pandas: add regression guards for conversion perf changes (commit by @no23reason)

v1.70.0

New Features

• 63a79b6 - eval: conversation tracking and --preserve-failed flag (commit by @macekond)
• d599be3 - gooddata-eval: optional run_metadata_extra on agentic evaluators (commit by @myhoai)

Bug Fixes

• 2ebc8b0 - eval: fix search_tool correctness always scoring 0% (commit by @Tomkess)
• f847679 - eval: harden _args_match against malformed tool-call JSON (commit by @Tomkess)
• 2299bb7 - docs: generate method sub-pages for every class in API reference (commit by @czechian)
• 6c7e272 - eval: treat missing/null alert trigger as ALWAYS default (commit by @henrynguyengooddata)
• 16b8722 - gooddata-eval: delete metrics created during agentic eval runs (commit by @myhoai)

Tests

• ecd1f66 - eval: wire gooddata-eval into tox and fix stale unit tests (commit by @hkad98)

Chores

• 7f97db7 - eval: fix lint violations flagged by ruff (commit by @hkad98)

v1.69.0

New Features

• 95b2cea - add null values to pdm and sql scans (commit by @okrcmaGD)
• 19d3936 - sdk: support multiple MAQL candidates in metric skill evaluation (commit by @zdenekmusil-gd)
• 00331ec - eval: retry transient gen-ai SSE errors in chat client (commit by @zdenekmusil-gd)

Bug Fixes

• 913e7e5 - catalog: Support NUMBER analytics parameters in declarative model (commit by @svagi)
• b526366 - eval: thread-safe progress callbacks for --concurrency (commit by @macekond)

Refactors

• 7f59137 - eval: move all agentic evaluation logic into gooddata_eval SDK (commit by @zdenekmusil-gd)

Tests

• b21a0cc - correct agentic alert parsing and ranking-filter cross-ref (commit by @myhoai)

Chores

• 06634a0 - bump quiver version to v0.418.0 (commit by @okrcmaGD)
• 198404d - regenerate api-client against staging (commit by @okrcmaGD)

v1.68.0

New Features

• 22b6da6 - sdk: add LLM provider test & list-models enablers (commit by @zdenekmusil-gd)
• 0544cac - add support for IP allowlist policies (commit by @DmitriiNikitinGD)
• 6ab0749 - eval: add gooddata-eval model-evaluation CLI (Phase 1 + Phase 2 + Langfuse) (commit by @zdenekmusil-gd)
• c253bdf - eval: multi-model comparison, gd-eval models command, Langfuse enhancements (commit by @zdenekmusil-gd)
• d48428a - eval: add dashboard_summary test kind via /summary endpoint
• fe1ef7f - eval: robust summary scoring + multi-case example dataset
• 360f274 - eval: add --concurrency flag for parallel item evaluation (commit by @zdenekmusil-gd)
• d7f9d97 - eval: carry summary_input through the Langfuse dataset source
• ed3303d - eval: set Langfuse trace version to the model id

Bug Fixes

• 62bf443 - eval: resolve active LLM provider setting by type, not fixed id
• a43d45b - eval: set provider_type in resolve_and_activate default path
• a05b708 - eval: return SummaryInput from the Langfuse source helper
• e06dfc9 - include user and usergroup (commit by @DmitriiNikitinGD)

Chores

• f9639cb - bump ruff to ~=0.15.15 and ty to ~=0.0.40 (commit by @hkad98)
• c6cdda3 - api-client: regenerate against staging (commit by @DmitriiNikitinGD)
• aa60710 - eval: drop example summary dataset (now lives in Langfuse)