This release is focused on improving the stability of the SDK.
What's Changed
- Enhanced Database Session Management and Configurable Connection Limits by @movchan74 in #189
- Deployment Health Check and Automatic Restart by @movchan74 in #191
- Support for vLLM 0.6.3 by @HRashidi in #190
- Retry Mechanism for Deployment Requests by @movchan74 in #193
- Fixes for GPU Tests with dstack by @movchan74 in #195
- Task Queue Reliability Enhancements by @movchan74 in #194
- Update Retryable Exceptions by @movchan74 in #196
- Batched whisper integration by @Jiltseb in #197
- Heartbeat Implementation for Task Queue by @movchan74 in #199
- Ensure pytest logs are always extracted in GPU test workflow by @movchan74 in #198
- Adding timestamp related params to batched version by @Jiltseb in #200
- Mitigate Health Check Timeouts for Blocking Operations in Deployments by @movchan74 in #201
- Use Timezone Consistently in Task Repository by @movchan74 in #202
- Removed deprecated route_prefix from Deployment Options by @movchan74 in #203
- Improved Error Handling in WhisperDeployment by @movchan74 in #204
- Update Haystack Dependencies by @movchan74 in #205
- Fix Sampling Parameters Merging by @movchan74 in #206
- Add Ray Dashboard Host and Port Parameters by @movchan74 in #207
- Add Ray Dashboard Host and Port Parameters in Aana CLI by @movchan74 in #208
- Add Torch Cache Clearing to Health Check for vLLM and Whisper Deployments by @movchan74 in #209
- Add Sequential Deployment Option by @movchan74 in #210
Full Changelog: v0.2.2.3...v0.2.3
Contributors
movchan74, Jiltseb, and HRashidi