Skip to main content

Rasa Pro Services Change Log

All notable changes to Rasa Pro Services will be documented in this page. This product adheres to Semantic Versioning starting with version 3.3 (initial version).

[3.6.0] - 2025-10-10​

Rasa Pro Services 3.6.0 (2025-10-10)

Features​

  • Add support for IAM authentication for AWS RDS database in the analytics service. To enable this feature, set the following environment variables:

    • IAM_CLOUD_PROVIDER: set to aws to enable IAM authentication for AWS RDS
    • RASA_ANALYTICS_DB_HOST_NAME: <your-db-hostname> the hostname of the RDS instance
    • RASA_ANALYTICS_DB_PORT: <your-db-port> the port of the RDS instance
    • RASA_ANALYTICS_DB_NAME: <your-db-name> the name of the database
    • RASA_ANALYTICS_DB_USERNAME: <your-db-username> the username to connect to the database
    • AWS_DEFAULT_REGION: <your-aws-region> the AWS region where the RDS instance is hosted

    Additionally, you can also set the following optional environment variables:

    • RASA_ANALYTICS_DB_SSL_MODE: the SSL mode to use when connecting to the database, e.g. verify-full or verify-ca
    • RASA_ANALYTICS_DB_SSL_CA_LOCATION: the path to the SSL root certificate to use when connecting to the database

    When these environment variables are set, the analytics service will use IAM authentication to connect to the AWS RDS database.

  • Add support for IAM authentication for AWS Managed Streaming for Apache Kafka in the analytics service. To enable this feature, set the following environment variables:

    • IAM_CLOUD_PROVIDER: set to aws to enable IAM authentication for AWS MSK
    • AWS_DEFAULT_REGION: <your-aws-region> the AWS region where the MSK instance is hosted
    • KAFKA_SECURITY_PROTOCOL: set to SASL_SSL to use SASL over SSL
    • KAFKA_SASL_MECHANISM: set to OAUTHBEARER to use OAuth Bearer token authentication
    • KAFKA_SSL_CA_LOCATION: the path to the SSL root certificate to use when connecting to the MSK cluster, this can be downloaded from Amazon Trust Services.

    When these environment variables are set, the analytics service will use IAM authentication to generate temporary credentials to connect to AWS MSK.

  • Configure Analytics service to connect to Kafka broker using mTLS. To enable this feature, set the following environment variables:

    • KAFKA_SSL_CERTFILE_LOCATION: Path to the client certificate file.
    • KAFKA_SSL_KEYFILE_LOCATION: Path to the client private key file.

Improvements​

  • Add new environment variables for each AWS service integration (RDS, MSK) that indicates whether to use IAM authentication when connecting to the service:
    • KAFKA_MSK_AWS_IAM_ENABLED - set to true to enable IAM authentication for MSK connections.
    • RDS_SQL_DB_AWS_IAM_ENABLED - set to true to enable IAM authentication for RDS connections.

Bugfixes​

  • Refactor the logic for creating new sessions. It now gets the previous event context directly from the database which makes the logic more robust against cases where the in-memory event stream might not have the previous event.

[3.5.7] - 2025-10-06​

Rasa Pro Services 3.5.7 (2025-10-06)

Bugfixes​

  • Add the installation of zlib OS dependency to Analytics Dockerfile to fix build issues when running the service.

[3.5.6] - 2025-09-11​

Rasa Pro Services 3.5.6 (2025-09-11)

Bugfixes​

  • Upgrade alpine base image to 3.19 in order to a security vulnerability in alpine:3.17

[3.5.5] - 2025-09-03​

Rasa Pro Services 3.5.5 (2025-09-03)

Bugfixes​

  • Upgrade vulnerabilities:
    • requests to version 2.32.5
    • cryptography to version 43.0.3
  • Fix vulnerabilities in outdated setuptools and pip installed via the python:3.9-slim original base image used by the Analytics Docker image build. Replace python:3.9-slim base image with alpine:3.17 which installs python 3.10. Drop python 3.9 support which has reached its EOL.

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.5.4] - 2025-07-29​

Rasa Pro Services 3.5.4 (2025-07-29)

Bugfixes​

  • Upgrade the following dependencies to fix security vulnerabilities in the analytics service:

    • setuptools to version 78.1.1
    • gunicorn to version 23.0.0
    • redshift-connector to version 2.1.8

  • Check if the message has already been processed before processing it. Commit message offset only after it was processed successfully. If message processing fails retry the transaction. If the message offset commit fails, retry the command.

    Added environment variables:

    • RETRY_CONNECTION_COUNT - how many times the connection to the Kafka broker is retried when it fails, Confluent Kafka producer and consumer are recreated on each retry
    • RETRY_DB_TRANSACTION_COUNT - how many times the DB transaction is retried when it fails
    • KAFKA_SOCKET_KEEP_ALIVE_ENABLED - whether the socket keep alive is enabled for the Kafka connection. Corresponds to socket.keepalive.enable in librdkafka.
    • KAFKA_METADATA_MAX_AGE_MS - the maximum age of the Kafka metadata in milliseconds. Corresponds to metadata.max.age.ms in librdkafka.
    • KAFKA_PRODUCER_RETRIES - how many times the librdkafka producer retries sending a message when it fails. Corresponds to retries in librdkafka.
    • KAFKA_PRODUCER_TIMEOUT_MS - the timeout for the librdkafka Kafka producer in milliseconds. Corresponds to request.timeout.ms in librdkafka.
    • KAFKA_PRODUCER_PARTITIONER - the partitioner used by the librdkafka Kafka producer. Corresponds to partitioner in librdkafka.
    • KAFKA_COMPRESSION_CODEC - the compression codec used by the librdkafka Kafka producer. Corresponds to compression.codec in librdkafka.
    • KAFKA_CONSUMER_HEARTBEAT_INTERVAL_MS - the heartbeat interval for the librdkafka Kafka consumer in milliseconds. Corresponds to heartbeat.interval.ms in librdkafka.
    • KAFKA_CONSUMER_SESSION_TIMEOUT_MS - the session timeout for the librdkafka Kafka consumer in milliseconds. Corresponds to session.timeout.ms in librdkafka.
    • KAFKA_CONSUMER_MAX_POLL_INTERVAL_MS - the maximum poll interval for the librdkafka Kafka consumer in milliseconds. Corresponds to max.poll.interval.ms in librdkafka.
    • MAX_MESSAGES_TO_FETCH - how many messages are fetched from the queue at once, default is 10

    For more info about the librdkafka configuration options see https://github.com/confluentinc/librdkafka/blob/master/CONFIGURATION.md.

[3.5.3] - 2025-06-27​

Rasa Pro Services 3.5.3 (2025-06-27)

Bugfixes​

  • Handle JsonPatchConflict exceptions gracefully when loading pre-existing dialogue stack events or when applying current event patches. This prevents the Analytics service from crashing due to conflicts in the JSON patch operations.

[3.5.2] - 2025-05-21​

Rasa Pro Services 3.5.2 (2025-05-21)

Bugfixes​

  • Fixed a bug in Rasa Analytics for Session Creation Logic. A session can be created by events slot event with the slot session_started_metadata followed by action event for action_session_start OR just the event action for action_session_start alone.

[3.5.1] - 2025-05-12​

Rasa Pro Services 3.5.1 (2025-05-12)

Bugfixes​

  • Fix the processing of stack events by the analytics service in the case of multiple parallel conversations whose events are split across several batches consumed by the service.
  • Replace sqlalchemy's DateTime type with postgresql dialect specific TIMESTAMP type in columns that record date and time. This is required to keep fractional seconds precision which is essential when retrieving stack events from the database to reconstruct the dialogue stack for each different conversation id.

[3.5.0] - 2025-03-20​

Rasa Pro Services 3.5.0 (2025-03-20)

Improvements​

  • MTS: Update MTS to train without the need for NFS.
  • MRS: Update MRS to run with bot config and model trained from remote storage instead of using NFS.

Bugfixes​

  • Updated jinja2, werkzeug, idna, requests, zipp and urllib3 to address security vulnerabilities.

[3.4.1] - 2025-05-12​

Rasa Pro Services 3.4.1 (2025-05-12)

Bugfixes​

  • Updated jinja2, werkzeug, idna, requests, zipp and urllib3 to address security vulnerabilities.
  • Fix the processing of stack events by the analytics service in the case of multiple parallel conversations whose events are split across several batches consumed by the service.
  • Replace sqlalchemy's DateTime type with postgresql dialect specific TIMESTAMP type in columns that record date and time. This is required to keep fractional seconds precision which is essential when retrieving stack events from the database to reconstruct the dialogue stack for each different conversation id.

[3.4.0] - 2024-12-12​

Rasa Pro Services 3.4.0 (2024-12-12)

Improvements​

  • Added Python 3.10 and Python 3.11 support to Rasa Analytics

Bugfixes​

  • MTS and MRS: Add retry capability to Kubernetes API calls on 5xx errors.

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.3.5] - 2024-10-29​

Rasa Pro Services 3.3.5 (2024-10-29)

Bugfixes​

  • MTS and MRS: Add retry capability to Kubernetes API calls on 5xx errors.

[3.3.4] - 2024-10-02​

Rasa Pro Services 3.3.4 (2024-10-02)

Bugfixes​

  • Return status code 503 when the Analytics service is unavailable during a healthcheck request. This allows the user to implement liveness probes that could automatically restart the service upon failure.

[3.3.3] - 2024-09-25​

Rasa Pro Services 3.3.3 (2024-09-25)

Bugfixes​

  • MTS: Fix bug (ATO-2257) when training pod's status is not caught when MTS consumer job restarts.
  • [MTS] Update certifi to 2023.7.22 to resolve vulnerability CVE-2023-37920.

[3.3.2] - 2024-05-28​

Rasa Pro Services 3.3.2 (2024-05-28)

Improvements​

  • MRS: Upload rasa pod logs to remote storage for user-friendly access to logs in case the running of the assistant fails.

[3.3.1] - 2024-05-27​

Rasa Pro Services 3.3.1 (2024-05-27)

Bugfixes​

  • Allow users to specify image pull secrets for MTS / MRS

[3.3.0] - 2024-04-03​

Rasa Pro Services 3.3.0 (2024-04-03)

Improvements​

  • Align column names representing the flow_id as defined in the yaml file across rasa_flow_status, rasa_llm_command and rasa_dialogue_stack_frame tables:
    • rasa_llm_command table: flow_name column has been renamed to flow_identifier.
    • rasa_dialogue_stack_frame table: active_flow column has been renamed to active_flow_identifier.
  • MTS: Handle MTS Job Consumer restarts when rasa training pod continues to run. If the TrainingManager finds a running training job when it restarts, it will check the status of the pod:
    • If the pod is pending or running, it will watch the pod until training is complete.
    • If the pod has completed, it will upload the logs and trained model (only if it is present).
  • MTS: Accept nlu in the config data of CALM assistants in order to use nlu_triggers.
  • MTS and MRS: Support debug logs in rasa pods.

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.2.3] - 2023-12-20​

Rasa Pro Services 3.2.3 (2023-12-20)

Improvements​

  • MTS: Setup alembic schema migration mechanism for the database of the model training orchestrator. Add initial table creation migration file.
  • [MTS] Add capability to configure log level for MTS orchestrator. Log level can be configured through LOG_LEVEL environment variable. Default log level is INFO.

Bugfixes​

  • Fix telemetry reporting in shipped Docker images.

[3.2.2] - 2023-12-05​

Rasa Pro Services 3.2.2 (2023-12-05)

Bugfixes​

  • Remove obsolete component RemoteGCSFetcher from MTS orchestrator.

[3.2.1] - 2023-12-01​

Rasa Pro Services 3.2.1 (2023-12-01)

Improvements​

  • Align column names representing the flow_id as defined in the yaml file across rasa_flow_status, rasa_llm_command and rasa_dialogue_stack_frame tables:
    • rasa_llm_command table: flow_name column has been renamed to flow_identifier.
    • rasa_dialogue_stack_frame table: active_flow column has been renamed to active_flow_identifier.
  • Add environment variables to control resource requirements and limits for Rasa pod. MTS and MRS job consumers can now be configured to use specify resource requirements and limits for the Rasa pod. This can be done by setting the following environment variables in the Rasa pod:
    • RASA_REQUESTS_CPU
    • RASA_REQUESTS_MEMORY
    • RASA_LIMITS_CPU
    • RASA_LIMITS_MEMORY

[3.2.0] - 2023-11-22​

Rasa Pro Services 3.2.0 (2023-11-22)

Features​

  • Added new table rasa_dialogue_stack_frame to store active flow names and steps for each event sequence in the conversation.
  • Add new table rasa_llm_command to store LLM generated commands for each user message. Add new column in the _rasa_raw_event table to store the serialized LLM generated commands.
  • Add new table rasa_flow_status to store the transformations of rasa flow events. Add new columns in the _rasa_raw_event table to store the flow_id and step_id of these events where applicable.

[3.1.1] - 2023-07-17​

Rasa Pro Services 3.1.1 (2023-07-17)

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.1.0] - 2023-07-03​

Rasa Pro Services 3.1.0 (2023-07-03)

Features​

  • Added Real Time Processing of Markers. Markers can now be evaluated real time by the Analytics Data Pipeline. We've added event handlers for evaluation all events from Kafka to extract markers. The extracted markers are saved into rasa_marker database table. These markers are evaluated with the patterns stored in rasa_pattern table.

    Added API endpoints to create patterns in rasa_pattern table. This endpoint is used by Rasa Plus for rasa markers upload command.

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.0.2] - 2023-06-13​

Rasa Pro Services 3.0.2 (2023-06-13)

Improvements​

  • Adds an environment variable to control logging level of the application.

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.0.1] - 2022-10-26​

Rasa Pro Services 3.0.1 (2023-10-26)

Miscellaneous internal changes​

Miscellaneous internal changes.

[3.0.0] - 2022-10-24​

Rasa Pro Services 3.0.0 (2023-10-24)

Features​

  • Analytics Data Pipeline helps visualize and process Rasa assistant metrics in the tooling (BI tools, data warehouses) of your choice. Visualizations and analysis of the production assistant and its conversations allow you to assess ROI and improve the performance of the assistant over time.
Last updated on