CLI: 0.1.154
What's Changed
- More agent add-ons by @luke-lombardi in #728
- bump sdk version by @luke-lombardi in #730
- add volume path to event by @luke-lombardi in #732
- Feat: Add authorized flag to bots to allow public bot pages by @luke-lombardi in #733
- Bot: Add event history to bot session response by @luke-lombardi in #736
- add public getsession endpoint by @luke-lombardi in #737
- feat: add welcome message to bot sessions by @luke-lombardi in #738
- add user messages to history by @luke-lombardi in #741
- bot: fix input dict serialization by @luke-lombardi in #743
- Feat: Add optional retry_for argument for task queue by @dleviminzi in #731
- Add logging and temporary force update of container status by @dleviminzi in #744
- fix bug with retry for causing taskqueue to spin down by @dleviminzi in #745
- Update blobcache discovery by @luke-lombardi in #746
- Fix: handle cache discovery properly by @luke-lombardi in #747
- Fix: dial timeout to host lookup by @luke-lombardi in #748
- fix: properly use dial context by @luke-lombardi in #749
- Feat: Cedana C/R by @luke-lombardi in #658
- Fix: Patch panic with extractPackageName by @jsun-m in #752
- Feat: Handle generator functions in realtime decorator by @jsun-m in #751
- Fix: Update cedana binaries by @luke-lombardi in #753
- Fix: prevent C/R during serves by @luke-lombardi in #755
- Fix: Add context guard to opening runc events channel by @dleviminzi in #742
- Fix: Pin latest versions and update engine args for vllm by @dleviminzi in #756
- Fix: fix taskqueues getting stopped unexpectedly by @luke-lombardi in #757
- Fix: Dont coalesce pip install commands if they contain flags by @dleviminzi in #758
- Fix: Use indexed stub_id for task query instead of external id by @dleviminzi in #754
- More specific flags to flush on by @dleviminzi in #759
- chore: bump update sdk by @luke-lombardi in #760
- Add missing vllm args for vllm v0.6.4.post1 by @dleviminzi in #761
- update blobcache to enable directIO by @luke-lombardi in #763
- update blobcache again by @luke-lombardi in #764
- update blobcache again for directIO support by @luke-lombardi in #765
- Enable direct mount and expose max write config by @luke-lombardi in #766
- Feat: Multi-gpu workloads by @luke-lombardi in #762
- Don't return any extra flags if direct io is disabled by @luke-lombardi in #768
- Fix: Dont init multiprocessing Value unless it will be used by @dleviminzi in #767
- Fix: Prevent creation of an external machine if the pool does not have a provider by @jsun-m in #769
- feat: sort gpu controllers by priority by @nickpetrovic in #770
- Upgrade runner image to ubuntu 22.04 by @luke-lombardi in #772
- Fix: Add bzip for Mamba and remove distutils for Python 3.12 by @dleviminzi in #775
- Fix: Pin setuptools to working version for Python versions older than 12 by @dleviminzi in #776
- Feat: Add auxiliary state to help check the runtime status of a stub by @jsun-m in #771
- Fix: Use heartbeats to correct for request token drift by @luke-lombardi in #777
- Update README.md by @mernit in #778
- feat: support saving empty files with outputs by @nickpetrovic in #780
- Chore: Remove failed container threshold log by @jsun-m in #781
- Feat: Support multiple cpu pools by @luke-lombardi in #782
- fix: task queue tasks should error when they raise an exception by @nickpetrovic in #783
- fix typo w/ preemptable flag by @luke-lombardi in #784
- Add vm logs config by @dleviminzi in #785
- Feat: Allow users to bring their own Dockerfile by @dleviminzi in #773
- Fix: Add missing enable_prompt_tokens_details flag for vllm by @dleviminzi in #786
- Add gpu_count arg to vllm wrapper by @dleviminzi in #787
- Feat: Structured logging by @dleviminzi in #711
- Exec writer should use slog not old output channel by @dleviminzi in #788
- Increase request buffer connection timeout; bump tailscale by @luke-lombardi in #789
- Chore: Clean up gateway config by @jsun-m in #790
- Fix: Endpoint returns internal service error on exception by @dleviminzi in #791
- Fix: fix task cancellation status bug by @luke-lombardi in #794
- Feat: Build secrets by @dleviminzi in #792
- Fix: Use correct port for GetExternalURL by @jsun-m in #795
- Revert endpoint change by @dleviminzi in #796
- Fix: Write response headers from copy of internal response's headers by @dleviminzi in #797
- Fix: add no wait in endpoint to skip graceful timeout by @jsun-m in #800
- Downgrade tailscale and revert header copy commit by @dleviminzi in #801
- Feat: Export workspace gateway config for configuration in dashboard by @jsun-m in #793
- Fix: Remove instance spec from exec env by @dleviminzi in #802
- fix: Don't volume cache builds by @luke-lombardi in #803
- log retry failures by @luke-lombardi in #804
- Fix: Don't update endpoint tasks in terminal status by @luke-lombardi in #805
- Fix: increase heartbeat frequency by @luke-lombardi in #806
- Downgrade grpcio by @dleviminzi in #807
- Fix: fix memory leak in endpoint buffer by @luke-lombardi in #808
- Hotfix to allow re-registering of ttld machines by @luke-lombardi in #810
- fix tests, add machine state to index by @luke-lombardi in #811
- Chore: Sanitize container requests and add stub id to container events by @jsun-m in #814
- Fix: endpoint throughput improvements by @luke-lombardi in #816
- Fix: Use correct ignore file for beta9 and beam cli commands by @dleviminzi in #818
- Feat: beam shell by @luke-lombardi in #817
- chore: update sdk version by @luke-lombardi in #821
- add ssh run dir by @luke-lombardi in #822
- Fix: Dont timeout during dockerfile builds by @dleviminzi in #823
- Fix: Correct resource reporting for single gpu on shared gpu worker by @jsun-m in #824
- Always return actual error during builds by @dleviminzi in #827
- Fix: Remove task running lock scan by @luke-lombardi in #828
- Fix: move function monitor to process by @luke-lombardi in #829
- Send oom event to es by @dleviminzi in #826
- Fix: clear worker address when deleting container state by @luke-lombardi in #830
- Fix: Add a TTL on build containers for cleanup by @jsun-m in #819
- Feat: Add warm up endpoint to both endpoints and taskqueues by @jsun-m in #813
- Copy custom images to disk instead of memory before unpacking by @dleviminzi in #831
- Increase tmp size limit by @dleviminzi in #832
- Add full request to es oom event by @dleviminzi in #834
- Revert "Fix: Add a TTL on build containers for cleanup (#819)" by @jsun-m in #835
- feat: add multipart uploads and range downloads for volumes by @nickpetrovic in #809
- fix: issue with getting package version for user-agent by @nickpetrovic in #837
- Fix: Patch issue with image build process not spinning down and undo revert for killing expired build containers by @jsun-m in #833
- feat: use logarithmic approach to calculate chunk size by @nickpetrovic in #838
- fix: remove user-agent header by @nickpetrovic in #839
- fix: type error with progress updater by @nickpetrovic in #840
- Fix: Use digest if one is provided in custom images by @dleviminzi in #841
- Fix: Return error and exception when endpoint app raises an exception by @dleviminzi in #820
- Feat: Add the ability to skip outputs and stats retrieval on list query by @jsun-m in #842
- Remove thread local, expose printing retry errors by @nickpetrovic in #844
- fix: multipart uploads and downloads with empty files by @nickpetrovic in #847
- feat: multipart upload/download directories recursively by @nickpetrovic in #848
- Fix: Hotfix build container ttl key trimming by @jsun-m in #849
- fix: panic makeslice len out of range by @nickpetrovic in #850
- fix: show terminal cursor after multipart error by @nickpetrovic in #851
- Add note warning about poetry shell being moved to plugin in v2.0.0 by @dleviminzi in #843
- Fix: Make sure to still sanitize config by @jsun-m in #852
- Fix: Use distinct error value in endpoint by @dleviminzi in #853
- feat: update blobcache to support streaming and prefetch by @luke-lombardi in #854
- fix: fix blobcache panic by @luke-lombardi in #855
- (chore) bump cache again by @luke-lombardi in #857
- Fix: Use correct set of container prefixes in image build service by @jsun-m in #858
- fix: bump blobcache to fix small lock issue by @luke-lombardi in #860
- fix: ensure monitor_task works as expected by @luke-lombardi in #861
- Change log to say base image instead of custom base image by @dleviminzi in #863
- Fix: Remove recursive stat to improve volume api requests by @jsun-m in #864
- Fix: Stuck build containers by @dleviminzi in #836
- Feat: Build with gpu option for images by @dleviminzi in #845
- Fix: Dont sent stop build on success by @dleviminzi in #867
- Feat: Use cloudpickle to run remote functions in jupyter environments by @dleviminzi in #856
- Feat: Add min_containers to QueueDepthAutoscaler and instance controller for loading and reloading instances by @jsun-m in #846
- Revert "Fix: Remove recursive stat to improve volume api requests (#8… by @jsun-m in #869
- bug in notebook env check by @dleviminzi in #870
- fix: fix stop deployment bug by @luke-lombardi in #873
- Fix: fix serve min containers threshold by @luke-lombardi in #874
- Feat: Functions can be invoked in Hex notebooks by @dleviminzi in #875
- Fix: Patch stop deployment http endpoint by @jsun-m in #876
- Add on-prem readme by @mernit in #878
- Update README.md by @mernit in #879
- Update README.md by @mernit in #880
- Update README.md by @mernit in #881
Full Changelog: cli-0.1.124...cli-0.1.154