All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog and this project adheres to Semantic Versioning.
- Change tone task type + provider @janepie #163 #172
- Proofread provider @janepie #169
- ChatWithTools provider @julien-nc #167
- Add ability to summarize by chunks @edward-ly #157
- Admin setting to choose transcription model @julien-nc #154
- Admin setting to set the default image size @julien-nc #135
- Add ability to customize API root path in service URL @edward-ly #143
- Encrypt api keys and basic passwords for app config and user settings @julien-nc #136 #139
- Remove old providers since old APIs can use TaskProcessing providers @julien-nc #140
- Cache the models for 30min in distributed cache @julien-nc #152
- Missuse of user ID in most providers @julien-nc #153
- Use api key in image retrieval request @julien-nc #147
- Simple cache for model list to speed up provider loading @julien-nc #133
- Change default image generation size @kyteinsky #124
- Ability to choose model in the assistant for most providers @julien-nc #130
- Switch from webpack to vite @julien-nc #131
- Fix admin settings not saving the models in some cases @julien-nc #130
- admin-settings: add image model selection dropdown @kyteinsky #122
- bump max NC version to 31, update composer dev dependencies @julien-nc
- Bring back textProcessing, translation, STT, and image generation providers @julien-nc #120
- Avoid ambiguity between old and new providers @julien-nc #121
- fix(TextToText,Translate): Don't throw in getOptionalInputShapeEnumValues @marcelklehr #115
- fix(providers): do no implement useless 'getOptionalOutputShapeDefaults' method, use the size param in text2image provider @julien-nc #119
- Translation provider for the task processing API @julien-nc #111
- Adjust providers to changes in server (enum values and defaults) @julien-nc #111
- Drop the old translation provider @julien-nc #111
- Create all task processing providers that were text processing ones + image generation + audio transcription @julien-nc #102
- ContextWrite provider
- Chat provider
- Add password confirmation when setting sensitive values in user and admin settings
- Drop old image generation, audio transcription and text processing providers
- Use nc/vue 8.14.0
- Fix topics provider output: make it comma separated
- incorrect numeric settings handling on the server side @julien-nc
- move personal and admin settings to AI sections @julien-nc
- update npm pkgs @julien-nc
- support MistralAI API @julien-nc
- only add user param in chat completion request if using openAI @julien-nc
- allow empty string for extra llm params @julien-nc
- safely drop indexes in last migration step @julien-nc #95
- fix mistake when getting extra model params @julien-nc #94
- allow admins to set extra params for completion requests @julien-nc #86
- new "service name" admin setting @julien-nc #87
- mention compatibility with Plusserver in README and app description @julien-nc #87
- make it possible to toggle image, text and stt providers in admin setting @julien-nc #87
- support NC 30
- remove smart pickers that are now provided by the assistant app @julien-nc #85
- use nextcloud/vue components in admin settings @julien-nc #87
- use basic auth to get generated images if necessary @julien-nc #84
- chore: update workflows from templates @skjnldsv
- Update node, npm, nextcloud/vue @MB-Finski
- Update all node deps @kyteinsky
- Remove unnecessary spacing in admin settings @st3iny
- Fix psalm errors @MB-Finski
- Trim suffixed / in base url @kyteinsky
- Use composer bin for dev dependencies @MB-Finski
- Add psalm and lint workflows, cs:fix @MB-Finski
- Add psalm config @MB-Finski
- Support new OCP API with expected runtime for text processing providers @julien-nc
- Calculate expected provider runtimes according to past observed runtimes @MB-Finski
- Specify allow_local_remote_servers in the request options so that it's no longer needed in config.php @jkellerer
- Improve unit test cleanup @MB-Finski
- Implement a text-to-image provider @julien-nc
- Fix image generation with LocalAI (with LocalAi 2.0 the endpoint is no longer fully OpenAI compatible) @MB-Finski
- Fix translation provider @MB-Finski
- Respect LocalAI endpoint setting with providers @MB-Finski
- Hide unsupported settings from image generation dialog with LocalAI @MB-Finski
- Add php-unit tests for providers @MB-Finski
- An attempt at fixing info.xml parsing in the app store
- Optional Basic authentication for self-hosted (LocalAI) API wrapper services @mickenordin
- A comprehensive set of php-unit tests for speeding up development and preventing bugs creeping in @MB-Finski
- Error with params for the completion endpoint @MB-Finski
- Admin setting for choosing the endpoint when using LocalAI @MB-Finski
- Error introduced in transcription for 1.1.0 @MB-Finski
- Enable clearing of prompt history from personal settings page @MB-Finski
- Enable limiting of api usage on per-user basis (user can use an own API key to bypass this) @MB-Finski
- Expose token limit as admin setting for text generation @MB-Finski
- Request no autocomplete for api key (in addition to disabling the field), which has caused issues for some users @MB-Finski
- Replace deprecated query() of IConfig in app container setup @MB-Finski
- Rewrite error handling logic, use proper http error codes @MB-Finski
- try to select gpt-3.5 when default model is not found when getting model list
- use selected default llm (in admin settings ) in the translation provider
- remove dashboard category in info.xml
- dynamically change labels depending if OpenAi or LocalAi is used
- implement text processing providers (summarize, outline, free prompt and custom reformulate) @julien-nc
- admin setting to choose the API request timeout @julien-nc
- fix scrolling in the picker components @julien-nc
- allow usgin LocalAI without an API key @julien-nc
- save new admin default llm model when switching between OpenAI and LocalAI @julien-nc
- do not use OC functions/vars anymore @julien-nc
- App was not working if no admin api key is defined #16 @julien-nc
- Allow using a custom service URL #15 #18 @julien-nc
- New admin settings to toggle each feature #15 @julien-nc
- Bump max Nextcloud version to 28 @julien-nc
- Immediately update prompt history in chat/image picker components @julien-nc
- translation provider @julien-nc
- allow user-specific API keys #16 @julien-nc
- improve style of audio recorder @julien-nc
- use latest nextcloud/vue so the picker modal size can be set by the provider
- change regenerate icon @julien-nc
- disable inputs during generation/regeneration @julien-nc
- set prompt history bubble max width, add title with full text @julien-nc
- implement OCP\SpeechToText\ISpeechToTextProvider @julien-nc
- improve style of the whisper picker component, respect dark theme @julien-nc
- improve style of the GPT and Dall-e picker components @julien-nc
- remember last image size for Dall-e picker @julien-nc
- implement 2 step flow in GPT and Dall-e picker components, preview/adjust before submitting @julien-nc
- add prompt history for text and image generation @julien-nc
- add padding in the picker component because the one in the picker will be removed @julien-nc
- get rid of
@nextcloud/vue-richtextdependency as the build bug has been solved in the nextcloud webpack config @julien-nc
- gpt picker option to set max number of tokens (~words) to generate
- 'include query' option in gpt custom picker
- link preview for Dall-e image generations
- support for gpt-3.5
- admin setting to choose default completion model
- save last used model as default user-specific default completion model
- Whisper speech-to-text smart picker provider to translate/transcode
- show error message of failing API request responses
- improve design
- use more NC components
- the app