What's Changed

• Add CLI and server version checks with upgrade guidance by @aliasaria in #1677
• Fix github based interactive tasks by @deep1401 in #1679
• fix: apply trackio fixes for experiment ID and modal behavior by @orincolor in #1684
• Fix prettier issue on Trackio modal by @dadmobile in #1685
• Fix name field in add provider modal to show as required by @deep1401 in #1681
• Fix output and files buttons on interact history by @deep1401 in #1686
• Fix common envs for multi-node runs by @deep1401 in #1689
• Fix reuse coroutine errors by @deep1401 in #1690
• Fix misleading 'check your email' message when EMAIL_METHOD is dev by @greninja in #1692
• Change email method defaults on frontend and change healthz call format by @deep1401 in #1694
• Unset active venv during multiuser_setup install by @deep1401 in #1693
• Show provider name in jobs list by @deep1401 in #1691
• Add CLI installer to CI workflows (and add non-interactive mode) by @aliasaria in #1695
• Fix runpod randomly marked complete in small gap when pod is appearing by @deep1401 in #1680
• Improve CLI server install UX for storage and compute prompts by @aliasaria in #1699
• Move one time startup scripts to a prestart script by @deep1401 in #1701
• make sub-routers under compute_provider by @deep1401 in #1696
• Local provider jobs run through sandboxes and add subprocess monitoring by @deep1401 in #1671
• fix e2e test errors by @deep1401 in #1704
• Rename prestart_once to on_server_start by @aliasaria in #1705
• Move compute provider sub-router logic under services by @deep1401 in #1697
• Fix github checkout commands to support older git on skypilot gpu by @deep1401 in #1707
• Documents folder fix and lab access by @deep1401 in #1700
• Fix task exception errors on startup due to remote buckets by @deep1401 in #1710
• Allow deleting jobs in NOT_STARTED state by @deep1401 in #1711
• Fix old eval results viewing from before migration by @deep1401 in #1712
• Add favourite and hide capabilities to jobs list by @aliasaria in #1708
• Fix exp seeding to be after bucket creation by @deep1401 in #1717
• Fix task exception never retrieved due to aiobotocore sessions by @deep1401 in #1714
• Fix trackio logic by @deep1401 in #1713

New Contributors

• @orincolor made their first contribution in #1684
• @greninja made their first contribution in #1692

Full Changelog: v0.33.1...v0.33.2

What's Changed

• Add PyPI metadata to CLI package by @aliasaria in #1641
• Fix broken CLI screenshot URLs on PyPI by @aliasaria in #1642
• Add commits and lines-of-code tracking to pr-stats dashboard by @aliasaria in #1643
• Revamp getting all jobs to have caching for faster calls by @deep1401 in #1644
• Add progress bar for launch text by @deep1401 in #1631
• Warning in QueueTaskModal for tasks imported from gallery by @deep1401 in #1599
• gitignore Claude worktrees by @dadmobile in #1645
• Fix wrong failed warnings for list dir when no jobs by @deep1401 in #1646
• Fix missing org context in local provider executor thread by @aliasaria in #1634
• Improve AGENTS.md: add format:check, env prereqs, condense curl docs by @aliasaria in #1649
• Show more interactive job status and move completed jobs to History by @aliasaria in #1633
• Fix edge cases for requires org context by @deep1401 in #1651
• Replace pathlib instances with os by @deep1401 in #1647
• Fix experiment seeding errors with require org context by @deep1401 in #1653
• Fix circular progress to be hidden after 99 for interactive tasks by @deep1401 in #1654
• Add caching to tasks route, and simplify the check-status calls by @deep1401 in #1625
• Optimize experiments get all by @deep1401 in #1652
• Simplify installer by @deep1401 in #1628
• Cleanup of legacy model code - Part 1 by @dadmobile in #1598
• Fix job stopping for local provider tasks by @deep1401 in #1657
• Fix check-status calls in interactive by @deep1401 in #1659
• Fix env instructions for the agent by @deep1401 in #1666
• Mark runpod complete on no storage backend by @deep1401 in #1665
• Restrict .env file permissions to owner-only (0600) by @aliasaria in #1662
• Validate frontend URL scheme in CLI server install by @aliasaria in #1661
• Validate config before writing .env in CLI server install by @aliasaria in #1663
• Back up .env before overwriting during server install by @aliasaria in #1664
• Fix local interactive issues by @deep1401 in #1672
• Re-add delete running interactive job by @dadmobile in #1668
• Adds force refresh logic for local provider setup by @deep1401 in #1660
• Fix Playwright hello-world-task test for UUID job IDs by @aliasaria in #1676

Full Changelog: v0.33.0...v0.33.1

What's Changed

• Remove datadog import by @dadmobile in #1583
• Auto detect supported accelerators when adding a local provider by @deep1401 in #1589
• Remove code related to run_job by @dadmobile in #1587
• delete plugin sdk and download model by @deep1401 in #1586
• Add setup hooks and make the provider modal nicer by @deep1401 in #1576
• Remove entire interact component and add caching to compute providers list by @deep1401 in #1590
• Fix closing of deletetaskmodal by @deep1401 in #1593
• Fix image and cpu param for cpu machines on runpod by @deep1401 in #1594
• Fix wrong import which broke tasks output by @deep1401 in #1596
• Prompt for server URL during login so users can confirm or change it by @aliasaria in #1591
• Improve CLI and make it more agent friendly by @deep1401 in #1561
• Add comfyui interactive and fix interactive card behaviour by @deep1401 in #1592
• Revamp trackio so you can add multiple runs to the same project by @deep1401 in #1549
• Make TLAB_LOG_LEVEL work properly by @dadmobile in #1595
• Remove Electron legacy code and stubs by @aliasaria in #1572
• Add caching to model list route by @deep1401 in #1588
• Optimize GitHub Actions: consolidate workflows, update versions, add timeouts by @aliasaria in #1573
• Add Playwright E2E tests to CI via GitHub Actions by @aliasaria in #1571
• Make QueueTaskModal more compact (collapsible resource override) by @deep1401 in #1577
• make launches go through a worker queue by @deep1401 in #1581
• Reduce job update rebuilds and make job creation faster by @deep1401 in #1582
• Add autocomplete attributes to enable Chrome password generator by @aliasaria in #1605
• Implement gpu/cpu profiling by @deep1401 in #1505
• Add version update banner to frontend and CLI by @aliasaria in #1584
• Improve CLI code quality and maintainability by @aliasaria in #1606
• Add --format=json support to CLI config command by @aliasaria in #1608
• Add PR activity report generator by @aliasaria in #1609
• Fix version check to show update banner by @deep1401 in #1613
• Change cli config to user 'lab config set' and 'lab config ' by @deep1401 in #1612
• Fix excessive disk usage from redundant .cache dirs in local provider runs by @aliasaria in #1610
• install only prettier for prettier workflow by @deep1401 in #1614
• Add/model dataset group by @mina-parham in #1481
• Fix double v prefix and version comparison in update banner by @aliasaria in #1615
• Add curl authentication guide to AGENTS.md by @aliasaria in #1616
• Cache experiments list by @deep1401 in #1600
• Add transformerlab-cli agent skill by @aliasaria in #1617
• Document --format flag ordering in CLI skill by @aliasaria in #1622
• Small cleanup: Remove TFL_LOCAL_PROVIDER_SETUP from Dockerfile by @dadmobile in #1624
• Fix provider compatibility for Interactive tasks by @deep1401 in #1620
• Fix wrong complete state for jobs by @deep1401 in #1621
• Remove multiuser env var by @deep1401 in #1626
• Add lab server install CLI command by @aliasaria in #1623
• Replace legacy sparse checkout with quiet git clone by @aliasaria in #1619
• Ignore some files during task upload by @deep1401 in #1627
• Add SkyPilot container and cloud settings (global + per-job) by @aliasaria in #1551
• Document SDK installation, subsystem docs, and context vars in AGENTS.md by @aliasaria in #1632
• Disable fields when submitting and fix trash icon for non-terminal jobs by @deep1401 in #1629
• Added 3D artifact visualization support by @ParamThakkar123 in #1585
• Add shared Claude Code project settings by @aliasaria in #1635
• Move jobs within the experiment folder and make calls faster by @deep1401 in #1607
• Fix package-lock json to have the latest packages by @deep1401 in #1637
• Fix frontend call not sending experiment_id to API with job by @deep1401 in #1638
• Replace N+1 serialized job data writes with bulk helper by @aliasaria in #1640
• Add configurable async thread pool and concurrency limits by @aliasaria in #1639
• Fix thread pool exhaustion from local provider launches by @aliasaria in #1636

Full Changelog: v0.32.0...v0.33.0

What's Changed

• Fix references for git related yaml fields by @deep1401 in #1526
• Clean up unused code flagged by knip by @aliasaria in #1527
• Change references from command -> run by @deep1401 in #1502
• Fix 'Task is missing a command to run' for github_repo_url interactive tasks by @aliasaria in #1524
• Add task resources confirmation in GUI and CLI by @deep1401 in #1528
• Remove train components and move the usable ones to tasks by @deep1401 in #1531
• Show stopping state on interactive job cards by @deep1401 in #1533
• Add interactive task support for GitHub-hosted task definitions (Ollama + Gradio) by @aliasaria in #1503
• Add task execution documentation by @aliasaria in #1535
• docs: clarify visual testing tool preference by @aliasaria in #1530
• Add caching on tasks list route by @deep1401 in #1532
• fix: show server unreachable error instead of generic login failure by @aliasaria in #1529
• Make interactive_type optional for gallery-based tasks by @aliasaria in #1538
• Add live setup status when adding local provider by @deep1401 in #1537
• Fix provider logs showing new content in the middle by @aliasaria in #1539
• Add webhook support to notify about jobs in terminal state by @deep1401 in #1536
• Remove all recipe components and their usage by @deep1401 in #1540
• Remove all unused eval components by @deep1401 in #1541
• Fix API Key Auth and move to services by @deep1401 in #1542
• Add --running flag to CLI job list command by @aliasaria in #1545
• Fix embedded images not loading in iframe by @deep1401 in #1544
• Pull ngrok logic out of gallery and put it into the API code by @deep1401 in #1543
• Fix remote status worker to also check and mark dead interactive tasks by @deep1401 in #1546
• Interactive tasks on the CLI by @aliasaria in #1534
• Fix install sh to not use any active venv by @deep1401 in #1547
• fix bug in trackio project name during copy by @deep1401 in #1548
• Add hooks to run fixed commands for each provider before or after every task run by @deep1401 in #1558
• Fix config table to have multiple keys with different teams by @deep1401 in #1557
• Fix slow output rendering in EmbeddedOutputModal by removing line-by-line animation by @aliasaria in #1550
• Add CLI provider management commands (lab provider) by @aliasaria in #1556
• Reduce local provider queue delay with fast-path dispatch by @aliasaria in #1553
• Fix cache in task list by @deep1401 in #1562
• Install uv and install system packages on runpod by @deep1401 in #1564
• Add Documentation section to CLAUDE.md with links to docs/ and .agents/ by @aliasaria in #1570
• Add bulk job data updates to avoid repeated s3 calls by @deep1401 in #1569
• Make the interactive jobs not be interrupted by remote status worker by @deep1401 in #1567
• Hide Interactive Tasks in tasks tab and fix all task list calls by @deep1401 in #1568
• Fix team gallery tasks and team interactive tasks by @deep1401 in #1566
• Mark runpod as non-beta by @deep1401 in #1578
• Move all local provider things within HOME_DIR/local_provider by @deep1401 in #1575
• Remove caching on jobs list by @deep1401 in #1579
• Dont check for provider compat in non-local providers by @deep1401 in #1580

Full Changelog: v0.31.1...v0.32.0

What's Changed

• Exclude api/test/tmp from git and VS Code file watcher by @aliasaria in #1508
• Add Playwright framework starter for demo videos by @aliasaria in #1509
• add/remotion mnist demo by @aliasaria in #1510
• Add ability to list files in tasks by @deep1401 in #1512
• Remove route for v1/models which uses controller by @deep1401 in #1513
• Remove unused header and reclaim grid space by @aliasaria in #1511
• chore: update caniuse-lite browserslist database by @aliasaria in #1517
• docs: add prettier reminder to AGENTS.md by @aliasaria in #1519
• Delete TaskFilesModal and use FileBrowserModal by @deep1401 in #1518
• Remove unused files: HelpVideoModal, ConnectionLostModal, ViewOutputModalStreaming (Eval) by @aliasaria in #1520
• Disable noImplicitAny to focus tsc on actionable errors by @aliasaria in #1522
• Add skipLibCheck to tsconfig.json to suppress node_modules type errors by @aliasaria in #1521
• fix: break import cycle between interactUtils and chat API module by @aliasaria in #1516
• fix: remove unnecessary void operator in ExperimentInfoContext by @aliasaria in #1515
• fix: remove debug console.log from appContextPlugin by @aliasaria in #1514
• Fix artifacts directory function parameter by @deep1401 in #1523
• Remove TransformerLabSettings by @deep1401 in #1525

Full Changelog: v0.31.0...v0.31.1

Key Changes

• New Chart mode for eval runs and ability to compare eval runs
• Azure Storage support for workspace backend
• Introduce caching on multiple endpoints
• Make launching a task more informative
• Major improvements in interactive tasks
• Custom SBATCH flags support for SLURM providers

What's Changed

• Auto add a local provider when a new team is created by @deep1401 in #1382
• Get rid of job queue json and run over and over by @deep1401 in #1385
• Stop fetching pipeline tag when in multiuser mode by @deep1401 in #1388
• initial dockerfile in the test directory -- we can use this to finalize a test container by @aliasaria in #1372
• Delete a bunch of pages by @aliasaria in #1387
• Remove all single user mode logic from sidebar by @deep1401 in #1389
• Make MULTIUSER to be true by default by @deep1401 in #1384
• Fix failing cpu and memory constraints for local provider queue by @deep1401 in #1392
• Fix the default configs when adding a new provider by @deep1401 in #1383
• Add Playwright testing framework by @aliasaria in #1396
• Remove a lot of API routes by @aliasaria in #1391
• Fix Compute Tab for a local provider by @deep1401 in #1390
• Fix job not found errors when running with local providers by @deep1401 in #1397
• AGENTS.md conda environment instructions (for now) by @dadmobile in #1395
• Fix conditions related to os kill in local cluster status so its not marked complete/down while running by @deep1401 in #1402
• Remove components and graphs from header by @deep1401 in #1398
• fix child processes not being killed with local provider by @deep1401 in #1403
• Add/remove unused javascript files and dependencies by @aliasaria in #1401
• Move venv inside the workspace folder of the job dir by @deep1401 in #1400
• Increase number of lines for tunnel info logs by @deep1401 in #1404
• Remove remaining v1 Pages by @aliasaria in #1406
• One-liner: gitignore .claude by @dadmobile in #1408
• Add .coderabbit.yaml by @dadmobile in #1409
• Fix false condition of TFL_REMOTE_STORAGE_ENABLED by @deep1401 in #1412
• Fix deletion of secrets and also not launch tasks with missing secrets by @deep1401 in #1405
• Fix port connection issues on proxy urls by @deep1401 in #1411
• Fix Compatible chip on QueueTaskModal by @deep1401 in #1414
• Close QueueTaskModal only when provider launch request returns by @deep1401 in #1416
• Add Chart mode and Compare mode for evals by @deep1401 in #1415
• Make Interactive Tasks look better by @aliasaria in #1422
• Remove lingering workflows stuff by @deep1401 in #1420
• Don't allow adding multiple local providers by @deep1401 in #1423
• Fix Artifacts dropdown to show options conditionally by @deep1401 in #1425
• Add git workflow instructions to AGENTS.md by @aliasaria in #1429
• First playwright smokescreen test by @aliasaria in #1410
• Update experiment navigation to include experiment name in URLs by @aliasaria in #1417
• Remove RAG reindex functionality from Documents UI and backend by @aliasaria in #1428
• Live status about launch being shown by @deep1401 in #1424
• Add a new terminal state for skypilot by @deep1401 in #1426
• Fix copy_dir in storage.py to do proper remote -> remote copies by @deep1401 in #1431
• fix broken mainapp by @deep1401 in #1433
• rename tabs for consistency: update 'Task Output' to 'Lab SDK Output'… by @aliasaria in #1437
• Fix default image on runpod and vllm execution as well by @deep1401 in #1438
• Update job status to FAILED for a crash detected by tfl remote trap by @deep1401 in #1432
• POC: Try to make storage actually async by @dadmobile in #1436
• Remove coderabbit yaml by @dadmobile in #1442
• Fix ls mode on file browser by @deep1401 in #1440
• Add a base venv for local provider and collect computer information by @deep1401 in #1399
• Get rid of ~/src folder and paste file mounts directly in homedir/sky_workdir by @deep1401 in #1441
• Add background worker to update sweep status by @dadmobile in #1393
• docs: Add Architecture Deep Dives section to AGENTS.md by @aliasaria in #1444
• fix: resolve QueuePool exhaustion by consolidating duplicate SQLAlchemy engines by @aliasaria in #1452
• Fix setting up of provider twice in interactive task by @deep1401 in #1448
• Sweep status 30 seconds default frequency with env by @dadmobile in #1454
• Reduce frequency of sweep status ping by @dadmobile in #1455
• Get rid of TFL_LOCAL_PROVIDER_SETUP and have seed admin user create local provider by default by @deep1401 in #1460
• Fix tasks output and also solve the StopIteration bug by @deep1401 in #1461
• fix: rename 'Provider' to 'Compute Provider' in UI labels by @aliasaria in #1458
• Show datasets files correctly by @deep1401 in #1457
• Set Job to RUNNING when remote trap starts and introduce a lab_init live status by @deep1401 in #1459
• Add ability to leave and delete teams by @deep1401 in #1463
• Add padding in output modal by @deep1401 in #1465
• Fix sweep status inefficiencies by @dadmobile in #1462
• Add Provider to Interactive History List by @dadmobile in #1443
• feat/interactive ports definition by @aliasaria in #1456
• Add support for Azure Blob and ADLS through abfs by @deep1401 in #1450
• Feat/copy task files to job dir by @aliasaria in #1449
• Add custom sbatch flags per-user per slurm provider and edit during runs by @deep1401 in #1451
• fix: local provider job execution improvements by @aliasaria in #1445
• fix launch logic by @deep1401 in #1470
• Add NGROK as special secret and fix things on interactive modal by @deep1401 in #1472
• Remove Save button from interactive task Configure modal by @aliasaria in #1476
• Fix: Cache busting for production frontend assets by @aliasaria in #1471
• Fix refresh interval for Machine Logs by @deep1401 in https://github.com/transformerlab/t...

What's Changed

• Fix frozen branch truthy remote storage by @deep1401 in #1413

Full Changelog: v0.30.1...v0.30.3

What's Changed

• Fix so bump_version and npm install make same pacakge-lock.json by @dadmobile in #1373
• One line: Print an error when datadog init fails. by @dadmobile in #1375
• Migrate some required functions from plugin sdk to lab facade by @deep1401 in #1363
• Interactive Sessions that Run Directly (without ngrok) by @aliasaria in #1371
• Fix failed task resources check on local provider selection by @deep1401 in #1377
• Make provider logs permanent by @deep1401 in #1376
• Fix jobs running with local provider using lab-sdk by @deep1401 in #1379
• Add option to force cuda 13 install by @deep1401 in #1380
• Breakdown interactive runs logic into core,tunnel,tail_logs by @deep1401 in #1378
• add sudo definition in run too by @deep1401 in #1381

Full Changelog: v0.30.0...v0.30.1

• Initial support for Runpod as a provider
• Create tasks entirely in the UI, or from existing code without needing a yaml definition
• GPU Split Strategy support when running inference

What's Changed

• Show a full page loader instead of briefly showing login screen while waiting for auth by @deep1401 in #1328
• Fix random refresh on login page by @deep1401 in #1355
• Remove all server/info calls when in multiuser mode by @deep1401 in #1353
• Fix text and boxes on local datasets/models page in multiuser by @deep1401 in #1339
• Add a loader for the view checkpoint modal instead of static text by @deep1401 in #1359
• add workflow to check if lab-sdk version in api and lab-sdk matches by @deep1401 in #1346
• Fix: Fixed Run model state reset on switching pages by @ParamThakkar123 in #1314
• Add a notifications summary to inform users of any pending notifications by @deep1401 in #1357
• Added GPU Split Strategy support for Inference similar to LM Studio by @ParamThakkar123 in #1323
• Make interactive task modal different in terms of GUI by @deep1401 in #1348
• Remote trap to indicate live status of a command running irrespective of lab-sdk usage by @deep1401 in #1305
• Add an option to access generated datasets via lab facade by @deep1401 in #1340
• Add a third option for creating blank task yaml by @deep1401 in #1360
• Replace REMOTE_WORKSPACE_HOST env var with TFL_STORAGE_PROVIDER by @deep1401 in #1352
• Import ddtrace if Datadog envs are set by @dadmobile in #1320
• Added Support for WAN Image Generation by @ParamThakkar123 in #974
• Add local provider resource validation by @mina-parham in #1337
• Trim down exporter plugin sdk by @deep1401 in #1361
• Make compute providers calls async to not block thread and handle connection errors in SkyPilot requests by @aliasaria in #1362
• Remove provider column from tasks list by @deep1401 in #1366
• Fix provider logs for runpod provider by @deep1401 in #1296
• Fix LostConnectionModal from appearing as a false positive for a timed out healthz request by @deep1401 in #1368
• Avoid fetching plugin file contents when in MULTIUSER mode by @deep1401 in #1370
• Add more async calls to compute_provider by @dadmobile in #1369

Full Changelog: v0.29.2...v0.30.0

What's Changed

• Change docker image workflow sequence to save space by @deep1401 in #1343
• Update package lock to v0.29.1 by @dadmobile in #1344
• fix version issue in package-lock.json by @deep1401 in #1345
• Fix multiuser conditions to show interact tab correctly by @deep1401 in #1347
• Fix error causing Header to crash on bad server info by @dadmobile in #1349
• Fix server info return format in multiuser mode by @deep1401 in #1350
• Clean up .env.example by @dadmobile in #1351

Full Changelog: v0.29.1...v0.29.2