feat(declarativeui): add hooks for each of our file actions

[julien-nc]

closes #340

Add declarative UI capabilities for clients.
Add one hook for each of our file actions (summarize, stt and tts).

Currently unknown:

• Which HTTP method is used to reach the hooks
• How is the file ID passed
• How to define multiple filters for one hook

[marcelklehr]

Looks reasonable

[julien-nc]

When combining the capabilities declared by various apps, the server combines the arrays. For objects, all good. For lists, as the keys are preserved and a Php list is almost an object with integer keys, the combining goes wrong and each list overwrites the other ones.

Here is my suggestion for the declarativeui capability structure:

• declarativeui is an object indexed by app IDs. So in each item of this object, an app can do whatever it wants, the content will not be combined with others
• Inside an "appId" object, we can have 2 list of endpoints, one under context-menu and one under create-new
• An endpoint is an object with those keys:
  • name
  • url
  • method (default: GET)
  • mimetype_filters (comma separated list of mimetype filters, for now these are not regexps, clients will accept files which mime type starts with any of the filters)
  • icon is a URL to the icon
  • bodyParams is an object with all the parameters that should be passed in the request body. The {fileId} placeholder is supported there too. Example: "bodyParams": "{\"fileId\": \"{fileid}\"}",

Object in the declarativeui key of the capabilites:

{
  "assistant": {
    "context-menu": [
      {
        "name": "Summarize",
        "url": "/ocs/v2.php/apps/assistant/api/v1/file-action/{fileId}/core:text2text:summary",
        "method": "POST",
        "mimetype_filters": "text/, application/msword, application/vnd.openxmlformats-officedocument.wordprocessingml.document, application/vnd.oasis.opendocument.text, application/pdf",
        "icon": "/apps-extra/assistant/img/declarativeui/summarize.svg"
      },
      {
        "name": "Transcribe audio",
        "url": "/ocs/v2.php/apps/assistant/api/v1/file-action/{fileId}/core:audio2text",
        "method": "POST",
        "mimetype_filters": "audio/",
        "icon": "/apps-extra/assistant/img/declarativeui/speech_to_text.svg"
      },
      {
        "name": "Text to speech",
        "url": "/ocs/v2.php/apps/assistant/api/v1/file-action/{fileId}/core:text2speech",
        "method": "POST",
        "mimetype_filters": "text/, application/msword, application/vnd.openxmlformats-officedocument.wordprocessingml.document, application/vnd.oasis.opendocument.text, application/pdf",
        "icon": "/apps-extra/assistant/img/declarativeui/text_to_speech.svg"
      }
    ]
  },
  "cospend": {
    "context-menu": [
      {
        "name": "cospend fake",
        "url": "hello"
      }
    ]
  }
}

[julien-nc]

@tobiasKaminsky You can test this branch by checking it out and building it:

npm ci
npm run dev
composer i --no-dev

For the capabilities to be declared by the assistant, you need some AI providers. There are some fake ones in the testing app (summarize and STT).