Changelog¶
All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning.
Fixed¶
- the ansible playbook now creates a valid credentials.yml file for Rasa Core
[0.18.0] - 2019-04-30¶
This is a major update, please read the migration instructions in the Migration Guide.
Added¶
- API endpoint to bulk upload and replace bot responses at
PUT /responses - Instructions how to deploy the Rasa Platform on OpenShift / Kubernetes
Changed¶
- support Core
0.14.0 - nginx container now using image
bitnami/nginx - mongodb container now using image
bitnami/mongodb - rabbitmq container now using image
bitnami/rabbitmq - updated docker file for
event-service,``platform-api`` andtest-integration, app folder is now owned by grouproot. This allow the container to run in Openshift - nginx now listening to port 8080 and 8443
Fixed¶
- the deployment environment configuration is properly read from
environments.yml
[0.18.0b1] - 2019-03-29¶
Added¶
- API endpoint to retrieve NLU training examples by the MD5 hash of their
textfield atGET /projects/<project_id>/data/<md5_hash> - API endpoint to retrieve NLU suggestions by the MD5 hash of their
textfield atGET /projects/<project_id>/logs/<md5_hash> - persistence of lookup tables and entity synonyms in uploaded NLU training data
annotatoraccount role that can annotate conversations- support for temporary intents & user goals
- endpoint to query intents
- support for adding correct labels to conversation messages
- BREAKING support for multiple Rasa Core services running different models
- endpoints to save and retrieve environment configurations at
GET /environmentsandPUT /environments - endpoint to retrieve a combined list of NLU and Core models at
GET /projects/<project_id>/models - add environment variable RABBITMQ_DISK_FREE_LIMIT to docker compose file, this prevents the pod from running out of ephemeral storage
- API endpoint to retrieve roles and their associated permissions at
GET /roles - API endpoint to list all users at
GET /users - API endpoints for user role management:
GET /users/<username>/rolesto list the user’s roles (currently only one role is supported),PUT /users/<username>/roles/<role>to set the user’s role toroleandDELETE /users/<username>/roles/<role>to removerolefrom the user - HTTP endpoint to update the current user’s username at
PATCH /user
Changed¶
- replaced
pytest-pep8withpytest-pycodestyle timestampfield fromGET /models/CORE(NLU)renamed totrained_at, reflecting thetrained_atentry in both Core and NLU model metadata- Tracker dump files can be requested using the
Acceptheader at theGET /conversations/<sender_id>endpoint. For a markdown file, useAccept: text/markdown. For a json file, useAccept: application/json. If unspecified, the endpoint returns the json-formatted tracker (not as a file). - Changed platform-ui port. Now exposes 80:8080. This is to prepare to merge the non-root containers
- renamed HTTP endpoint
GET /users/metoGET /user - renamed HTTP endpoint
POST /users/me/passwordtoPOST /user/password
Removed¶
Fixed¶
- bug in the conversation analytics which displayed the same number of user messages as bot messages
[0.17.1] - 2018-12-17¶
Fixed¶
- regex features are replaced on upload and can be used during training
- users with
useraccount roles are able to log in
[0.17.0] - 2018-12-04¶
This is a major new version which adds support for Rasa Core 0.12. The Rasa Core and NLU containers are now the open source ones available from docker hub. Since the docker-compose.yml and .env files have changed, we recommend you update by fetching and running the latest version of the install script (see platform installation docs). important Since there is a major version change in Rasa Core, you should re-train your model with Rasa Core 0.12 and upload it
Added¶
- User analytics endpoints at
/analytics - query parameters
start, until, policies, minimumUserMessages, policies, maximumConfidenceatGET /conversationsendpoint - Endpoints to fetch conversation metadata on unique actions, intents,
entities and policies. These are
/conversationActions,/conversationIntents,/conversationEntities,/conversationPolicies - support for emojis in training data
- add endpoint to flag user conversations
- Migrations for conversations metadata which did not yet include fields like intents, and actions
- Conversation filtering UI
Changed¶
- Change in docker-compose.yml
- Use open source Docker containers for Rasa Core and Rasa NLU
- Bump rasa_core to v0.12
- Bump rasa_core_sdk to v0.12
- Bump demobot to v0.17
- no-root Dockerfiles
- UI improvements
- Update nginx configurations. Workers set to auto, more connections per worker, tuned connections timeouts.
Removed¶
Changed¶
- Increased worker_connections and worker_rlimit_nofile to 2048 to handle higher traffic.
Added¶
- Creating interactive learning sessions from existing conversations
Changed¶
- Remove tolerance added to “until” timestamp when querying conversations
GET /conversationsendpoint supports new query parameters, allowing filtering by user utterances, intents, entities, actions and whether the conversation appears in the training data. Supports sorting conversations by their minimum action confidence for those that aren’t in the training, and sorting by the latest activity of the conversation.- Use entities endpoint instead of reducing list of possible entities from available data.
Fixed¶
- Handling null intents
- Buggy first line when adding new Suggestions manually
[0.16.6] - 2018-10-23¶
Fixed¶
- Scrolling bug in nlu data table
- Bug where asterisks in passwords were ignored when creating a new user login on the command line.
[0.16.5] - 2018-10-15¶
Added¶
- endpoint at
/projects/<project_id>/entitiesthat returns a list of unique entities in the training data
Changed¶
- all NLU and Core model endpoints are available at
/projects/<project_id>/models/nluand/projects/<project_id>/models/core - updated NLU, Core, Core SDK versions
- Interactive learning UI
Removed¶
Added¶
- users can choose to enable experimental features
- experimental support for interactive learning UI
Changed¶
- uses Rasa Core 0.11.11 which contains fixes for tensorflow compatibility
Fixed¶
- properly migrate events that have been persisted by rasa core <= 0.10
- Minor tensorflow release broke compatibility, and bugfixes were released on Rasa Core master and Rasa Core 0.11 branch. This updates the demo bot to be compatible.
[0.16.3] - 2018-10-04¶
Fixed¶
- migration of old pickeled rasa core events to the json format is now done properly
[0.16.2] - 2018-09-28¶
Added¶
- admin endpoint at
GET /api/logs?api_token=<your_api_token>that returns the logs of the platform instance as a zip file - docs on deploying custom Rasa Core and Rasa NLU code
- added a feature for deleting users through
rasa_commandswithout having to go through the MongoDB container
[0.16.1] - 2018-09-26¶
Changed¶
- updated docker containers to use the latest rasa core bugfix release (0.11.7)
[0.16.0] - 2018-09-19¶
This is a major update, please read the migration instructions in the Migration Guide.
Added¶
- major version change to support Rasa Core
0.11.4+and Rasa NLU0.13.3+. - ability to configure the name of the authenticationDatabase used by mongo
using
MONGO_AUTHDBenv var - core and NLU model APIs at
/models/<project>/coreand/models/<project>/nlu. See the full API docs at Models API - Endpoint that delegates user messages to a specific Rasa Core environment
POST /conversations/<user_id>/messages, with a query parameterenvironment(default value:production). Example:POST /conversations/myid/messages?environment=development.
Changed¶
POST /storiesendpoint accepts plain text (<STORY>) instead of a{'story': <STORY>}objectGET /api/versionendpoint no longer requires auth- Rasa NLU model training can now be initiated at
POST /models/<project>/nlu/jobs - Consistent Core and NLU tagging scheme. The endpoints are (examples given for
- core, replace base with
/models/<project>/nlufor NLU endpoints):
- add tag:
PUT /models/<project>/core/<model_name>/tags/<mytag>- delete tag:
DELETE /models/<project>/core/<model_name>/tags/<mytag>- get model for tag
GET /models/<project>/core/tags/<mytag>- get model by name
GET /models/<project>/core/<model_name>
GET /conversationsreturns a list of conversation metadata containing the fieldssender_id,latest_event_timeandlatest_input_channel
Example request:
$ curl platform-server.com/api/conversations?api_token=<TOKEN> | python -m json.tool
Example response:
[
{
"sender_id": "user_1",
"latest_event_time": 1537345401.394812,
"latest_input_channel": "rasa"
},
{
"sender_id": "8a513b01b4924c4986f424686771a501",
"latest_event_time": 1537345428.457986,
"latest_input_channel": "facebook"
}
]
Removed¶
- support for
until_timeandhistoryquery parameter atGET /statisticsendpoint
Added¶
- URL field queries to include or exclude fields from query result
User can specify which fields they want to include in the returned object, using one or more fields query parameters. A story object has a name key, which means that GET /stories?fields[name]=true would return only the name s of the stories.
To query nested attributes, use syntax similar to python dictionaries.
For example, a log object has the following structure: {"user_input": {"intent": {"name": "greet",...}}}.
So to find all the unique intent values in the data, query:
?fields[user_input][intent][name]=true.
By default this only returns distinct values for the requested field. This can be turned off by passing distinct=false as an extra query parameter.
- supported by endpoints /responses, /stories, /logs, /data
Changed¶
- maximum upload filesize increased to 800Mb
- improved rendering of long messages in
/datatables - improved rendering of long slots in conversations view
- conversations view moved to top, now default view.
/api/versionendpoint no longer requires authentication
Removed¶
Fixed¶
- fixed issue where tensorflow training would hang
- fixed issue where not all intents would show up in dropdown in
/data/new - fixed bug with uploading markdown training data
- fixed description in docs for how to upload a rasa core model
- display user messages with multiple intents linked by OR naturally
[0.15.4] - 2018-08-21¶
Added¶
Changed¶
- Disable save button on model configuration page if there are no changes
Removed¶
- Markdown syntax highlighting (broken for in-word emphasis which we use a lot)
Fixed¶
- Allow setting images as part of bot templates (as allowed in the domain file, specifying the image URL)
- Action bubbles truncate properly in conversations
- Message buttons display again in conversations
[0.15.3] - 2018-08-13¶
Added¶
- endpoint at
GET /stories.mdfor downloading Rasa Core stories in bulk as markdown file POST /storiesendpoint supports uploading bulk stories in anmdfileGET/POST api/responsesendpoints for downloading/uploading nlg templates now supportapi_token- docs on running custom NLU containers
Changed¶
POST /storiesendpoint accepts plain text (<STORY>)
instead of a {'story': <STORY>} object
- GET /api/version endpoint no longer requires auth
- NLU are created from Rasa Core’s event stream
- logs can be manually added py POST``ing a text query as a json payload to ``/api/projects/<project>/logs
- deletion of NLU logs at DELETE /api/projects/<project>/logs/<log_id>
Removed
——-
Fixed¶
--updateoption to change a user’s password in therasa_commands.pyscript- Models view shrinking to nothing on some small screens
- Default secret for JWTs
- Refresh training data is view when you upload new training data
- endpoint at
GET /stories.mdfor downloading Rasa Core stories in bulk as markdown file POST /storiesendpoint supports uploading bulk stories in anmdfileGET/POST api/responsesendpoints for downloading/uploading nlg templates now supportapi_token
Added¶
- API endpoints for uploading, modifying and deleting Rasa Core stories at
/api/stories - Pagination query parameters for bot responses and models (and improved the docs)
- Infinite scrolling for all lists
Changed¶
- No default limit for GETing data, logs and responses
- Uses Rasa Core commit 1e00dae21362b205682f87123045eabf67ab2ec8
- Uses Rasa NLU commit b08ad19fe02368a4d8266d9bb108e406aa07c249
- Improved swagger documentation including documenting authentication in many spec files.
Fixed¶
- Total counts for data and logs
Added¶
Changed¶
Use updated docs theme
Added¶
- Endpoints for managing bot responses.
- X-Total-Count header to many endpoints to help pagination.
- Ability to search and filter on /data and /logs on the server side.
- Endpoint to update training examples.
- Endpoint to find the versions of services running.
Changed¶
- Unified Platform API available at
/api. - Unified and vastly improved API docs.
- Refactored almost all the underlying APIs!
- Moved from generating random tokens to authenticating on JWTs.
- Data is now typed from API to render in the UI
Fixed¶
- Model evaluation crashing when an intent equals None
- Many endpoints being quite esoteric/not-RESTful
- Handle token timeouts transparently in UI
Removed¶
- Dependency of Rasa applications on
rasa_extensions. TheRasaChatInputcomponent is provided byrasa_core==0.10.2.
[0.14.6] - 2018-07-12¶
Added¶
- View conversations as their associated story in preparation for saving them as test cases
Changed¶
Fixed¶
Added¶
- global travis variables for NLU and Core commit, spaCy and tensorflow versions
- get travis to test pep8
- add manage_user script to the NLU API dockerfile
Changed¶
- backend is served from a single repository
- Docker containers have custom requirements
Fixed¶
- login redirects in ui