Changelogs

[2.20.0] - November 8, 2023

Added

  • Support for C++ Version of analytics (experimental), can be built using cmake build flag -DFREMIS_CPP_PLUGIN_EXPERIMENTAL=ON.

  • PaddlePaddle core image

Changed

  • Update libasyik version to 1.0.1 with increased response body limit from 16MB to 64MB

[2.19.0] - Oktober 31, 2023

Added

  • Support for service and pipeline setup from python script, specified in config.json "analytic_setup"

Fixed

  • Coordinator server is not set size limit correctly

Changed

  • Update libasyik version to 0.9.7

[2.18.1] - Oktober 10, 2023

Added

  • Nodes now listen to env variable FREMIS_MAX_HEADER_LIMIT_BYTES and FREMIS_MAX_BODY_LIMIT_BYTES, and it is preferred over server configuration in config.json

[2.18.0] - Oktober 6, 2023

Added

  • Coordinator max size limit for header and body is now optionally configurable via environment variable FREMIS_MAX_HEADER_LIMIT_BYTES and FREMIS_MAX_BODY_LIMIT_BYTES or coordinator's config.yml

[2.17.1] - Oktober 3, 2023

Fixed

  • When there are no nodes available coordinator now will return code 503 - Service Unavailable as oppose to crashing the coordinator

[2.17.0] - September 27, 2023

Added

  • Ability to run AnalyticService in separate process.

  • New module ProcessMonitor to periodically check the state of forked child process.

  • Calls to PythonServices are now pipelined.

  • Ability to run in No License Mode when no keys are entered.

Changed

  • Any PythonService now runs in separate process.

  • Now coordinator accepts keys via env: NODEFLUX_ACCESS_KEY, NODEFLUX_SECRET_KEY, NODEFLUX_DEPLOYMENT_KEY

[2.12.7] - December 20, 2022

Changed

  • Fremis Platform: node/worker config.json now may accept multiple analytic license config

Added

  • Fremis Platform: coordinator mode now has license and qps control enabled

[2.12.6] - December 8, 2022

Changed

  • Fremis Platform: main app now requires subcommand httpserver or coordinator

Added

  • Fremis Platform: new coordinator mode, node configuration via config file supplied via cli argument

  • Fremis Platform: coordinator mode listen to env variable FREMIS_COORDINATOR_HEALTHCHECK_INTERVAL_MS

  • Fremis Platform: coordinator mode listen to env variable FREMIS_COORDINATOR_HEALTHCHECK_TIMEOUT_MS

  • Fremis Platform: coordinator mode listen to env variable FREMIS_COORDINATOR_REQUEST_TIMEOUT_MS

[2.12.5] - November 24, 2022

Added

  • Fremis-N: Add too many request detector, may return 503 when server is overloaded (busy)

  • Fremis-N: Listen to env variable FREMISN_JOB_MONITOR_OVERLOAD_SLOPE which control "too many request" detector

  • Fremis-N: Listen to env variable FREMISN_LATENCY_METER_BUFFER_SIZE

[2.12.4] - November 15, 2022

Changed

  • Fremis Platform: can now run without license

[2.12.3] - October 24, 2022

Added

  • fix the absolute maximum number of candidate results.

[2.12.2] - October 24, 2022

Added

  • add the absolute maximum number of candidate result config to the coordinator.

[2.12.1] - October 22, 2022

Added

  • add a timeout for the coordinator to the worker's request for a search and health check.

  • new env variable to configure request timeout: FREMISN_COORDINATOR_HEALTHCHECK_TIMEOUT_MS & FREMISN_COORDINATOR_SEARCH_TIMEOUT_MS.

  • new env variable to configure max candidate: FREMISN_MAX_CANDIDATE_COUNT.

  • add the absolute maximum number of candidate results.

Changed

  • Not performing searches on unhealthy workers.

Fix

  • fix batch search crash when enrollment is empty.

[2.12.0] - October 14, 2022

Added

  • Fremis-N support batched searching

  • New env variable to configure Fremis-N behavior: FREMISN_BATCH_SEARCH_MAX_SIZE & FREMISN_BATCH_SEARCH_QUEUE_TIMEOUT_MS.

[2.11.1] - October 11, 2022

Changed

  • Fremis-platform enforces new endpoint scheme

[2.11.0] - September 29, 2022

Added

  • Get info API that consists of any instance information, especially license

[2.10.0] - September 21, 2022

Added

  • An initial version of the on-premise platform

[2.9.0] - September 9, 2022

Changed

  • enable qps limiter

  • license is not required in httpserver (worker) subcommand

  • support clustering using license only on the coordinator

  • add signature in coordinator request to worker

  • when running worker without license, request to worker requires signature header from coordinator, otherwise rejected

PreviousPorting Enrollment Database Cluster to DockerNextIntroduction

Last updated