2022.10

• TIM Platform LTS 2022.10
• Released on 31.10.2022

---

TIM API

TIM DB

New features:

• Allowed uploading of data with empty column(s) via
  • POST /datasets/csv and
  • POST /datasets/sql
• Added support of timestamp subformats for timestamps with timezone offset indicator specified at the end
• Expanded logging of endpoints DM06 (DELETE /datasets) and DM09 (DELETE /datasets/{id}) to include a record of each impacted entity (dataset version)
• Added includeWasImputed parameter to DM15 (POST /datasets/{id}/versions/{version}/slice) to allow for leaving out the wasImputed column
• Added timeZoneName property to datasets (full list of names can be found here)
  • Added timeZoneName as optional request body key in
    • POST /datasets/csv,
    • POST /datasets/sql
  • Added timeZoneName key to response body in
    • GET /datasets and
    • GET /datasets/{id}
• Deprecated timeZone property in
  • POST /datasets/csv and
  • POST /datasets/sql
• Added GET /datasets/{id}/versions/{version}/log endpoint for getting dataset version logs
• Added isPanelData filter to
  • GET /datasets and
  • GET /use-cases

TIM Forecast

New features:

• Added endpoints for retrieving production accuracies:
  • GET /forecast-jobs/results/production-table
  • GET /forecast-jobs/results/production-accuracies
• Added sequenceJobId filter to GET /forecast-jobs
• Added the retrain forecasting job type:
  • POST /forecast-jobs/{id}/retrain-model
• Enabled configuration of data alignment for endpoints:
  • POST /forecast-jobs/build-model
  • POST /forecast-jobs/rebuild-model
  • POST /forecast-jobs/retrain-model
  • POST /forecast-jobs/predict
• Deprecated model quality feature and replaced it with target offsets and predictor offsets features
• Automatically determine the step size for the trend feature (only daily trends were supported before, now it is trend from second, minute, hour, day, month or year is supported)

Changes:

• Removed model quality parameter from the TIMForecast Model Zoo and introduced new version v5.6 of the Model Zoo

TIM Detect

New features:

• Added endpoints for retrieving production accuracies:
  • GET /detection-jobs/results/production-table
  • GET /detection-jobs/results/production-accuracies
• Added sequenceJobId filter to GET /detection-jobs
• Added system-driven anomaly detection approach
  • POST /detection-jobs/build-model/system-driven
• Added approach filter to GET /detection-jobs and DELETE /detection-jobs (available options: kpi-driven and system-driven)

Changes:

• Added label column to output of GET /detection-jobs/{id}/results/table
• Deprecated POST /detection-jobs/build-model and replaced it with POST /detection-jobs/build-model/kpi-driven
• Deprecated POST /detection-jobs/{id}/rebuild-model and replaced it with POST /detection-jobs/{id}/rebuild-model/kpi-driven

TIM Explain

New features:

• Added what-if analysis job type for forecasting:
  • POST /forecast-jobs/{id}/what-if
• Added what-if analysis job type for kpi-driven anomaly detection:
  • POST /detection-jobs/{id}/what-if

TIM MLOps

New features:

• Telemetry endpoints (GET /ops/dataset-calls and GET /ops/job-calls)
  • Added dataset version createdAt to responses
  • Added approximated dataset version createdAt to responses for deleted dataset versions
  • GET /ops/dataset-calls
    • Added datasetVersionState filter

Changes:

• Improved authentication
• Telemetry endpoints
  • GET /ops/dataset-calls
    • endpointId filter accepts DatasetManagement endpoints (DM01 to DM16)
    • Removed datasetMicroservice filter
    • Renamed datasetExistsType filter to datasetState
    • Renamed time to madeAt in response
  • GET /ops/job-calls
    • endpointId filter accepts JobManagement endpoints (JM01 to JM47)
    • Removed jobMicroservice filter
    • Renamed jobExistsType filter to jobState
    • Renamed time to madeAt in response

TIM Studio

Changes:

• Configuration cards in both Forecasting & Anomaly Detection have been redesigned & reworked. The API call (job creation) now properly reflects what the configuration specifies in the cards.
• Configuration is properly validated as user modifies it and mutually exclusive options are disabled.
• Options not set/not modified are not passed to API call; thus the default is kept in the server's code.

Python client

Architecture

• Merged User Management and Workflow Management microservices into one (Workflow Management) microservice
• Priority queues for all ML jobs