Changelog

All notable changes to this project will be documented in this file.

[3.2.0] - 2025-03-18

Added

  • (Auto) translate emails inline
  • Add optional TTL for chat history (config.chat.history.pruneOldChatsAfter in days)
  • Service exposes an valid API doc via swagger

Changed

  • Change hostname configuration (add baseURL and devBaseURL, hostname is still supported but deprecated)

Fixed

  • 'Delete all chats' button is not shown if just one period is listed, e.g. Today, instead of just one chat
  • GetMonthlyPlanData respects monthly recurrence
  • Typeahead hint shown too often
  • breaking change Typo in settings string prevents correct consent dialog config
    • Changed dateUsagePolicy to dataUsagePolicy
    • Updated documentation accordingly

[3.1.0] - 2025-02-14

Added

  • PO files for several languages (Japanese, Dutch, Polish, Swedish, Danish, Finish, Turkish, Brazilian Portuguese)
  • As a user I see the AI Tour again in case I never used an AI feature
    • Introduced new setting io.ox/core//ai/remindAfterIdleDays, default -1 (disabled)

Changed

  • All AI calls use streaming now

Fixed

  • Manually triggered AI typeahead can bypass consent check
  • Missing documentation for ai/hostname setting
  • Typeahead has problems with complex mails

[3.0.0] - 2025-01-17

Added

  • Breaking change: As a user, I can switch between configured AI services
    • Service and model configuration changed to a simpler config, please check the helm charts README.MD or the
  • History for AI chat
    • Chats are now stored in a database, making it easy to revisit past topics or pick up where you left off.
  • Simple debug output for the AI config

Changed

  • Breaking change: Database usage is enabled by default now
    • With the recent change around AI Chat (i.e. adding a history), database usage is enabled by default.
    • This switch is now deprecated and will be removed in the near future since more and more features will require a persistence layer.
  • Accounting per service
    • Added migration to add service field to usage records
    • Added service field to usage API

[2.4.0] - 2024-12-18

Added

  • As a user I can use the AI chat on my mobile device
  • As a user I can easily copy AI responses to clipboard
  • As a user I can enable AI text suggestions for my smartphone independently from my desktop setting
  • As a user I can use AI text suggestions on my smartphone
  • Support for App Suite active-active setups
    • Reload settings in a failover scenario
    • Load hostname from settings instead of a module constant

Fixed

  • Text suggestion popup stays visible after mail compose close

[2.3.1] - 2024-11-22

Added

  • As a user I have an chat window to chat with the AI
  • LocalAI can use it's own baseURL
  • As a user AI can help me writing emails by predicting text content
    • New feature which hooks into mail compose and shows text predictions (aka. typeahead)
    • Global feature toggle is io.ox/core//features/ai-typeahead (default: false)

Changed

  • Tooltip should not show the model name

Fixed

  • Add correct OX Count file import and change namespace accordingly
  • Loading issues for AI tour and missing correct capability check
  • Request are now tracked when streaming is enabled
  • Tour is shown even if capability is disabled

[2.2.5] - 2024-10-25

Fixed

  • Fix: Tour is shown even if capability is disabled (upsell case)

[2.2.4] - 2024-10-17

Fixed

  • AI Icon missing in AI Tour

[2.2.3] - 2024-10-10

Changed

  • Better deprecation messages for models.js

Removed

  • Stop parameter for openai chat completions

[2.2.2] - 2024-10-08

Fixed

    • Update cookie to version 0.7.2
  • AI icon is missing in some places
  • Auth for admin routes is not optional
    • Admin route authentication is now optional. If no authentication is configured, admin routes will be disabled.
    • Add test to check if auth is optional for admin routes

[2.2.1] - 2024-09-30

Changed

  • Bump OX Commons

Fixed

  • Fix(deps): update dependency find-my-way to v9.1.0

[2.2.0] - 2024-09-27

Added

  • Support localai.io as provider for self hosted AIs
  • AWS Bedrock Claude support
  • As a user I get a tour about the AI integration in App Suite
  • As a User I can easily identify AI actions through a distinctive icon in the UI
  • AI Service can act as a pure UI source container
    • Added new config option "disableAPI" to just deliver UI source files

Changed

  • Move accounting brand route to admin folder
  • Updated documentation

Fixed

  • Tour shown on mobile devices
  • AI output not shown when opening mails in floating window

[2.1.0] - 2024-08-30

Added

  • Accounting documentation to the Helm chart's README.md
    • Add a section to the README.md file that explains how to use the accounting feature of the chart.
    • Renamed usageTracking to accounting in the helm chart
    • Renamed ENV variable TRACK_USAGE to ACCOUNTING_ENABLED
  • AI actions are counted with OX Count
  • AI-Service has a deployed documentation
  • Supported OpenAI models are now configurable via configMap

Changed

  • Increase log level to 'fatal' for OpenAI API errors to improve visibility

Fixed

  • (Helm Chart) env vars for mysqlSecret are now conditional
  • Issue #50: AI summary only translates the mail in some cases and skips the summary part
    • The used prompt was not clear enough resulting in a pure translation in some cases

[2.0.2] - 2024-08-07

Added

  • updated dependencies
    • moved jose from dev dependencies to regular dependencies

[2.0.1] - 2024-08-07

Added

  • Documentation on database setup and minimal grants

Security

  • switched dependencies to fix vulnerabilities
    • replaced get-jwks with jose to fix reported vulnerabilities in sub-dependency (elliptic)

[2.0.0] - 2024-08-02

Added

  • AI Service frontend integration is an App Suite plugin
    • please make sure to add the AI service to your UI-Middleware base-urls config (see readme for further config details)
  • AI frontend plugin booting is based on feature requirements
    • AI frontend plugin is only launched if all requirements are given
  • Switch to new capability for AI Service
    • capability is now ai-service instead of openai
  • : Integrate ChatGPT 4o model in Ai Service
  • Accounting and user level rate limiting
    • Added mysql db to store account data
    • Added usage plan configuration via config map
    • Added usage tracking for ai requests
    • Added rate limit hooks that check for users plan and tracked usage
    • Feature is configurable via helm chart value usageTracking.enabled which defaults to false
  • AI can do a proofread of written text in mailcompose
    • A new action in the content creation dialog now offers a proofread for written text in mail compose
  • Streaming responses for OpenAI

Fixed

  • "Generate answer" action shown in dropdown for threads even if it should not
  • AI Button missing in Mail Reader Popup in App Suite
  • No AI actions on mobile devices

[1.2.1] - 2024-05-13

Changed

  • Bump version of ox-common

[1.2.0] - 2024-04-30

Added

  • JWKS functionality

Changed

  • Helm: Make security context configuration overwritable

Removed

  • Old fastify routes to avoid users sending custom prompts to the AI service
  • Remove unused service account

Fixed

  • [#9] Helm release broken after shutdown of the cluster

[1.1.0] - 2024-02-09

Added

  • OpenAI Moderation API: Possible harmful in- and output to the AI is filtered

[1.0.1] - 2024-01-12

Fixed

  • : AI service rephrases in english instead of configured language
    • Added user language to rephrase prompt
    • Added unit test for rephrase and summarize
    • Removed async/await for createPrompt since it is not needed

[1.0.0] - 2023-10-09

Added

  • : Add openai endpoints
  • : Tracking AI service API usage
    • More detailed usage logging with OpenAI and Azure APIs
    • Metrics output for with Azure API
    • Requires numericUserId in the JWT issued by switchboard
  • Authentication via JWT
  • Clean metrics logging with existing response hook
  • Integration of Azure API as OpenAI proxy
  • Make certain API params for Azure configurable
  • Support for gpt-4 model ,
  • Swagger UI is now configurable via value in helm chart

Changed

  • Update to node 20

[0.0.1]

Initial release