v0.7.0-beta1

Changelog

• a872309 Abort prediction on a failed healthcheck event
• 5f51a96 Add --await-explicit-shutdown option to HTTP server
• 8aeb7f1 Add --redis-url argument for connecting to redis from URL Signed-off-by: Brandon Berhent bbedward@gmail.com
• 1f80fe8 Add a "starting" state for predictions
• bb2baf1 Add a healthcheck endpoint to cog HTTP server
• 3418f8a Add asynchronous create prediction endpoint
• 4753da9 Add cancel prediction endpoint
• 6c206c9 Add cog.director for coordinating predictions
• d86938f Add missing FileUploadError
• f90078f Add missing continue during health confirmation
• 047567a Add some more logging for cancels and time outs
• 3106496 Add webhook_caller_filtered
• 371a562 Add webhooks to HTTP server predictions
• 0a81808 Allow requesting health status explicitly
• b02b3f6 Attempt to improve cooperative shutdown
• 6fa42f5 Bind context for director logs while handling prediction
• 0068f06 Bump numpy to support Python 3.11
• c4e6920 Bump to 64 cores
• 4645c31 Create event handler in PredictionRunner
• 9b710a5 Dampen down the noise from library loggers
• 6e773bc Decouple webserver lifecycle from director
• fb8f377 Don't allow updates for the wrong prediction
• 8dc78e4 Don't duplicate tests in PRs
• b8cab4b Don't guess at a free port, let the OS pick one
• 3b316bb Don't trust model container's timestamps
• c25d007 Ensure backwards-compatible file uploads
• b3f0ef6 Extend CANCEL_WAIT to 30s
• 0a128c9 Extend prediction create timeout to 5s
• 80a1c30 Fix async endpoint test for start webhooks
• 3aad265 Fix openapi schema generation
• 2a10398 Fix some types and clean up some imports
• 85b0aa8 Fix tests for python 3.7
• 56fa616 Fix up setup run reporting
• 08668a1 Fixed a problem with OpenAPI spec generation
• d16f972 Flesh out director logging
• 9284718 Flesh out model container logging
• 975ad2e Fully invert control between QueueWorker and the webserver
• 12825f9 Handle failure/success reporting correctly
• a9b5c40 Handle file uploads to PUT endpoint from HTTP server
• 996a22f Handle prediction timeouts
• 498ae09 Handle validation errors and failures to cancel
• fa4b3fe Implement PredictionRunner.cancel()
• 85e6caf Invert control of webserver
• 90b5da1 Loosen OpenAPI spec assertion for FastAPI 0.89.0
• 0fe74e2 Make empty logs an empty string, not None
• 347bdd3 Make the abort message required
• 158d663 Move model healthchecking into own thread
• 228ea78 Only check if succeeded if terminal check passes
• 907be82 Only set content type headers if there is a content type to set
• 4bb172c Only set started_at after prediction has started
• 00a0b9e Preserve prediction status after slow cancelation
• ef02a2c Properly crash the container when predict thread errors
• 999f6c7 Reclassify FIXME and TODOs
• a1c588c Refactor redis queue tests
• 2a1cf58 Reimplement max_failure_count in Director
• d31d6e4 Remove Redis XAUTOCLAIM and ack messages whenever possible
• 39c8763 Remove extra response properties from webhook body
• f916be9 Remove unnecessary _last_result variable
• dc1449f Remove unused model_id and log_queue parameters
• 8805a5f Rename QueueWorker to Director
• b0a1e2a Rename shutdown to start_shutdown to avoid name collision
• 73a845e Report setup runs from director
• f17eeae Request and wait for an explicit healthcheck before each prediction
• 8e647e3 Retry file upload requests
• d9edff4 Return 0 runtime duration if prediction hasn't started
• 2372e28 Revert "Add --redis-url argument for connecting to redis from URL"
• 90e34eb Revert "Bump to 64 cores"
• ebc3a2a Revert "Revert "Add --redis-url argument for connecting to redis from URL""
• 9687042 Revert how COG_VERSION is set in cog Makefile target
• 606aac2 Run Python tests in parallel
• df9c3ab Run integration tests in parallel
• bc681c2 SIGKILL when http server receives shutdown call
• 4c2e419 Send webhooks from HTTP runner
• 85ec456 Set default timeouts for file uploads
• 0fab9a8 Shut down model container on graceful termination
• 5751cf7 Split director package
• cbf2088 Standardize on structlog logging
• ccba7c6 Terminate worker before waiting for thread exit
• c7d4f3f Tighten up error handling in PredictionRunner
• bf3649d Unconditionally SIGKILL when shutting down
• f75b695 Update HTTP API docs
• 7f5b598 Update README.md
• f5e5b5b Update instructions for installing cog for development
• d6963a7 Update integration tests to work with director
• 85c7d64 Use JSON logging in production
• ab79aee Use faster GitHub runners
• 8570995 cog: log prediction id
• f0b27cd update .all-contributorsrc [skip ci]
• 08f4c8f update .all-contributorsrc [skip ci]
• ff646be update README.md [skip ci]
• 7bfbb7c update README.md [skip ci]