Result structure

GET /api/spaces/{space_id}/audio/{audio_id}/result/ returns the finished result once the response has classified: true.

Sample response (shortened)

{
  "classified": true,
  "success": true,
  "audio_id": 51234,
  "title": "Sales team standup.mp3",
  "duration": 1837,
  "url": "https://storage.yandexcloud.net/aigenda/3f6c1e0a.mp3?X-Amz-Signature=…",
  "is_shared": false,
  "is_uploaded": true,
  "result": {
    "language": "en",
    "protocol": "### November sales plan\n\nWe launch the Start+ plan on November 1…",
    "dialog": [
      {
        "header": "November sales plan",
        "starts": [0, 14, 31],
        "ends": [13, 30, 52],
        "speakers": ["{101}", "{102}", "{101}"],
        "texts": [
          "Let's start with the November plan.",
          "We expect the first sign-ups for the new plan in the second week.",
          "Then we need the landing page by October 25."
        ]
      }
    ],
    "speakers_map": { "101": "Anna", "102": "SPEAKER_01" },
    "speakers_timeline": { "{101}": [[0, 13], [31, 52]], "{102}": [[14, 30]] }
  }
}

Top level

Field	Type	Contents
`classified`	boolean	`true` once processing is finished
`success`	boolean	`true` if processing succeeded
`result`	object \| null	Protocol, transcript, and speakers. `null` while processing runs
`audio_id`	integer	Recording ID
`title`	string	Recording title
`duration`	integer	Recording length in seconds
`url`	string	Temporary link to the source file, valid for 1 hour
`is_shared`	boolean	Whether public access to the result is on
`is_uploaded`	boolean	The recording file has been fully received

The result object

Field	Type	Contents
`protocol`	string	Meeting protocol in Markdown, sections with `###` headings
`dialog`	array	Transcript split into the meeting’s topic blocks
`speakers_map`	object	Speaker names by speaker ID
`speakers_timeline`	object	Speech intervals for each speaker
`language`	string	Recording language, for example `ru`

The response can also contain service fields not listed in this table. Don’t rely on them.

Transcript: dialog

dialog is an array of blocks. Each block matches a topic of the meeting and holds utterances as parallel arrays: the element at index i in every array describes one utterance.

Block field	Type	Contents
`header`	string	Block topic
`starts`	number[]	Start of each utterance, seconds from the beginning of the recording
`ends`	number[]	End of each utterance, seconds
`speakers`	string[]	Speaker of each utterance, an ID in curly braces: `"{101}"`
`texts`	string[]	Text of each utterance

To build a text transcript or subtitles from dialog, see Downloading files.

Speakers

AiGenda separates voices on its own and gives each speaker an ID.

speakers_map maps the ID (without braces) to a name: {"101": "Anna"}. If the voice isn’t labeled in the space yet, you get a label like SPEAKER_01 instead of a name.
speakers_timeline gives speech intervals per speaker: the key is the ID in braces, the value is a list of [start, end] pairs in seconds.

Speakers belong to the space: once you label a voice in the cabinet, AiGenda recognizes it in later recordings of that space.

Protocol

protocol comes as ready-made Markdown. You can show it in your UI right away or save it to a file.

Sample protocol

### November sales plan

- We launch the Start+ plan on November 1.
- The discount for current customers stays until the end of the year.

### Action items

| Task | Owner | Due |
|---|---|---|
| Prepare the plan's landing page | {101} | October 25 |
| Update the call script | {102} | October 28 |

Owners in the protocol appear as speaker IDs in curly braces: {101}. Replace them with names from speakers_map.

Get started

Upload a recording

Processing and results

Reference

Result structure

Top level

The result object

Transcript: dialog

Speakers

Protocol

​Top level

​The result object

​Transcript: dialog

​Speakers

​Protocol

Top level

The result object

Transcript: dialog

Speakers

Protocol