Issues

• Plugin tasks run kitaru.task from the PyPI release, not the worker's kitaru

  zenml-io/kitaru#1408 · Filter by author strickvl opened on Oct 30, 2026

• Design: Harbor as a native execution backend for Kitaru experiments

  zenml-io/kitaru#1364 · Filter by author htahir1 opened on Oct 23, 2026

• insight report shows only the latest run, and import analyze suggests an invalid filter

  zenml-io/kitaru#1363 · Filter by author htahir1 opened on Oct 23, 2026

• model-swap samples the first sessions in recorded order, not across the import

  zenml-io/kitaru#1362 · Filter by author htahir1 opened on Oct 23, 2026

• model-swap: a result over 1 MiB fails the task after the spend and loses the run

  zenml-io/kitaru#1361 · Filter by author htahir1 opened on Oct 23, 2026

• Model-swap guide: gaps found by a new-user run

  zenml-io/kitaru#1360 · Filter by author htahir1 opened on Oct 23, 2026

• model-swap: rebuilt calls drop Pydantic AI's required tool choice (allow_text_output=false)

  zenml-io/kitaru#1356 · Filter by author htahir1 opened on Oct 23, 2026
  1

• Warn when replays or experiments wait on a task kind no live worker claims

  Feature

  zenml-io/kitaru#1343 · Filter by author htahir1 opened on Oct 22, 2026

• model-swap: provider rejections on recorded OTel traces (64-char tool-call ids, GPT-5.6 tools)

  Bug

  zenml-io/kitaru#1342 · Filter by author htahir1 opened on Oct 22, 2026

• model-swap: a paid run's result is discarded when the report exceeds the 1 MB task-result limit

  Bug

  zenml-io/kitaru#1341 · Filter by author htahir1 opened on Oct 22, 2026
  1

• Workers stop claiming tasks after one hour: token renewal waits for 401, server returns 403

  Bug

  zenml-io/kitaru#1339 · Filter by author htahir1 opened on Oct 22, 2026
  1

  

• Build insights across trace import batches instead of per import

  Feature

  zenml-io/kitaru#1336 · Filter by author strickvl opened on Oct 22, 2026

• Let users mark insight findings as expected so they stop being flagged

  Feature

  zenml-io/kitaru#1335 · Filter by author strickvl opened on Oct 22, 2026

• Experiment results: intervals, cost per pass, a verdict label, a cross-run comparison, and CLI access

  Feature

  zenml-io/kitaru#1332 · Filter by author htahir1 opened on Oct 21, 2026

• Record and replay agents built directly on the OpenAI and Anthropic client libraries

  Feature

  zenml-io/kitaru#1331 · Filter by author htahir1 opened on Oct 21, 2026

• Honor the proxy environment variables in the HTTP client

  bug
  zenml-io/kitaru#1300 · Filter by author schustmi opened on Oct 8, 2026
  

• Support optional import association for manually created insights

  api-changebacklogenhancement
  zenml-io/kitaru#1293 · Filter by author strickvl opened on Oct 6, 2026
  3

• Add kitaru session export to write sessions and their labels as a JSONL dataset

  backlogenhancement
  zenml-io/kitaru#1289 · Filter by author strickvl opened on Oct 6, 2026

• Design replay from a chosen model call to produce counterfactual step labels

  backlogenhancement
  zenml-io/kitaru#1288 · Filter by author strickvl opened on Oct 6, 2026

• Keep scores and human feedback when importing traces

  enhancement
  zenml-io/kitaru#1287 · Filter by author strickvl opened on Oct 6, 2026

• Let evaluations point at the step they are about

  enhancement
  zenml-io/kitaru#1286 · Filter by author strickvl opened on Oct 6, 2026

• Offer a complete-capture mode in adapters that drop or truncate model inputs

  enhancement
  zenml-io/kitaru#1285 · Filter by author strickvl opened on Oct 6, 2026

• Add a kitaru-pi-importer for pi coding-agent session files

  enhancement
  zenml-io/kitaru#1284 · Filter by author strickvl opened on Oct 6, 2026

• Support fetching nodes for many sessions in one request

  enhancement
  Feature

  zenml-io/kitaru#1211 · Filter by author strickvl opened on Sep 25, 2026
  

• Import ATIF trajectories and Harbor rollout outputs

  enhancement
  zenml-io/kitaru#1070 · Filter by author strickvl opened on Sep 10, 2026