Wispr AI Meetings API
The Meetings API from Wispr AI — 7 operation(s) for meetings.
OpenAPI Specification
openapi: 3.0.3
info:
title: Wispr Backend Analytics Meetings API
description: Wispr Backend API
version: 0.5.2
tags:
- name: Meetings
paths:
/api/v1/meetings/sync:
post:
tags:
- Meetings
summary: Sync Meetings
description: "Reconcile the client's meetings against the server and return the delta.\n\nGated on cloud-sync via the ``require_cloud_sync`` dependency. Acks the\nclient's pushed meetings, pulls the server-side changes the client is\nmissing, and echoes the sync watermark the client should resume from.\n\nArgs:\n sync_request: The client's pushed meetings, watermark, and cursor.\n current_user: The authenticated caller.\n meetings_service: Service that performs the reconciliation.\n storage: Backing store for meeting transcripts.\n\nReturns:\n The acked meetings, the pulled delta, the next watermark, and the\n pagination cursor.\n\nRaises:\n HTTPException:\n 401: the authenticated token carries no user id.\n 422: a pushed field exceeds its size cap.\n 500: reconciliation fails unexpectedly (the original error is\n logged with a truncated message and full traceback, then\n wrapped)."
operationId: sync_meetings
security:
- ApiKeyHeaderPatched: []
parameters:
- name: content-length
in: header
required: false
schema:
anyOf:
- type: integer
- type: 'null'
title: Content-Length
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingSyncRequest'
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingSyncResponse'
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}:
delete:
tags:
- Meetings
summary: Delete Meeting
description: 'Soft-delete a meeting and reap its transcript blobs from S3.
Flips ``is_deleted=true``, then best-effort-deletes everything under
``users/{user_id}/meetings/transcripts/{meeting_id}/`` plus the legacy
flat key. Bumps the sync timestamp so other devices observe the
deletion on their next pull.
Returns 204 on both first-delete and idempotent re-delete. Returns
404 for unknown meetings and meetings owned by another user (we do
not leak existence to non-owners).'
operationId: delete_meeting
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
responses:
'204':
description: Successful Response
'404':
description: Meeting not found or owned by another user
'500':
description: S3 transcript cleanup failed; the meeting row may already be marked deleted. Retry is safe — the cleanup leg is idempotent.
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}/upload-url:
post:
tags:
- Meetings
summary: Create Meeting Upload Url
description: 'Mint a presigned S3 POST for a meeting''s refine audio + create the job.
The ``uploaded`` pipeline: the desktop uploads the encoded audio directly
to S3 with the returned presigned POST, then deletes its local copy. The
reconciliation poller picks the object up (by ``audio_s3_key`` on the
job), submits it to Soniox, and drives completion. Re-minting for a
meeting still awaiting its first upload refreshes the same job row (the
desktop''s pre-upload retry); it won''t disturb an in-flight or completed
job.
Errors map as:
* 404 — meeting doesn''t exist for this user (or is deleted).
* 413 — ``size_bytes`` exceeds :data:`REFINE_MAX_AUDIO_BYTES`.
* 415 — unsupported audio content type (only ``audio/ogg`` / ``audio/wav``).'
operationId: create_meeting_upload_url
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingUploadUrlRequest'
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingUploadUrlResponse'
'429':
description: Too Many Requests
'503':
description: Refine start temporarily unavailable — retry
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}/status:
get:
tags:
- Meetings
summary: Get Meeting Status
description: 'Report a meeting''s refine progress and inline what the client is missing.
The desktop polls this as a recovery channel parallel to sync, passing the
refine status it currently believes (``transcribing`` or ``summarizing``).
The server returns its own job status and inlines the refined transcript
and/or content only when it is ahead of the client. A 404 means the meeting
doesn''t exist on the server, so the client re-pushes it via sync.'
operationId: get_meeting_status
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
- name: client_status
in: query
required: true
schema:
$ref: '#/components/schemas/RefineJobStatus'
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingStatusResponse'
'404':
description: Meeting not found for this user
'429':
description: Too Many Requests
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}/refined-transcript:
get:
tags:
- Meetings
summary: Get Refined Transcript Download Url
description: "Mint a presigned S3 GET for a meeting's refined-transcript NDJSON.\n\nThe desktop calls this when a pull row's ``refined_uploaded_at`` stamp\nbeats its locally persisted fetched-through stamp, then downloads\n``refined.ndjson`` straight from S3 — the bytes never transit the API.\nPresign-only: no ``meeting_refine_jobs`` side effects (same contract as\nthe live-transcript / speaker-observations presigns).\n\nArgs:\n meeting_id: Meeting whose refined transcript to fetch.\n user: Authenticated caller; the row lookup is scoped to their id.\n db: Request-scoped executor (replica read).\n\nReturns:\n The presigned GET URL and its absolute expiry.\n\nRaises:\n HTTPException:\n 403: Cloud Sync is disabled for the user.\n 404: Meeting missing/foreign/deleted, transcript erased, or no\n ``refined_uploaded_at`` stamp yet."
operationId: get_refined_transcript_download_url
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingRefinedTranscriptUrlResponse'
'403':
description: Cloud Sync must be enabled
'404':
description: Meeting not found, deleted, transcript erased, or no refined transcript yet
'429':
description: Too Many Requests
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}/speaker-observations/upload-url:
post:
tags:
- Meetings
summary: Create Speaker Observations Upload Url
description: 'Mint a presigned S3 POST for a meeting''s speaker-observation log.
Sibling to :func:`create_meeting_upload_url` but with no job side
effects: it never touches ``meeting_refine_jobs`` (the audio route''s
upsert revives failed/abandoned jobs and clobbers the audio key — a
stuck observation artifact must not delay or disturb the transcript).
The key is server-composed via :func:`speaker_observations_key`, never
derived from client input. The client PUTs the artifact directly to S3;
Phase-1 transcript persistence reads it from S3 on demand (no confirm leg).
Errors map as:
* 404 — meeting doesn''t exist for this user (or is deleted).
* 413 — ``size_bytes`` exceeds :data:`SPEAKER_OBS_MAX_BYTES`.
* 415 — content type is not ``application/x-ndjson``.'
operationId: create_speaker_observations_upload_url
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/SpeakerObservationsUploadUrlRequest'
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingUploadUrlResponse'
'429':
description: Too Many Requests
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
/api/v1/meetings/{meeting_id}/live-transcript/upload-url:
post:
tags:
- Meetings
summary: Create Live Transcript Upload Url
description: 'Mint a presigned S3 POST for a meeting''s live-transcript NDJSON.
Sibling to :func:`create_speaker_observations_upload_url`: the desktop
PUTs ``live.ndjson`` directly to S3 instead of inlining it in the sync
body. Same no-job-side-effects contract — it never touches
``meeting_refine_jobs``. The key is server-composed via
:func:`live_ndjson_key`, never derived from client input.
Errors map as:
* 404 — meeting doesn''t exist for this user (or is deleted).
* 413 — ``size_bytes`` exceeds :data:`LIVE_TRANSCRIPT_MAX_BYTES`.
* 415 — content type is not ``application/x-ndjson``.'
operationId: create_live_transcript_upload_url
security:
- ApiKeyHeaderPatched: []
parameters:
- name: meeting_id
in: path
required: true
schema:
type: string
format: uuid
title: Meeting Id
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/LiveTranscriptUploadUrlRequest'
responses:
'200':
description: Successful Response
content:
application/json:
schema:
$ref: '#/components/schemas/MeetingUploadUrlResponse'
'429':
description: Too Many Requests
'422':
description: Validation Error
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
components:
schemas:
SpeakerPerson:
properties:
name:
type: string
title: Name
origin:
type: string
enum:
- platform
- calendar
- user
- merged_cal_platform
- self
- llm
title: Origin
platform:
anyOf:
- type: string
maxLength: 120
- type: 'null'
title: Platform
participantId:
anyOf:
- type: string
maxLength: 64
- type: 'null'
title: Participantid
additionalProperties: false
type: object
required:
- name
- origin
title: SpeakerPerson
description: 'One person in :attr:`SpeakerMap.people`, keyed by a plain uuid4 string.
``platform`` is immutable after mint; ``origin`` changes only via the
calendar-merge upgrade (``platform`` → ``merged_cal_platform``); a rename
(any origin, including a machine-minted person) changes only ``name``.
``origin="self"`` represents the meeting owner ("you").
``participantId`` is the server-derived stable calendar-attendee key —
deliberately NOT an email (nothing PII-bearing rides the map) and NOT the
``calendar_participant`` row id (those rows are replaced on every calendar
re-sync); present only on ``calendar`` / ``merged_cal_platform`` people.'
MeetingStatusResponse:
properties:
refine_status:
anyOf:
- $ref: '#/components/schemas/RefineJobStatus'
- type: 'null'
refined_transcript:
anyOf:
- type: string
- type: 'null'
title: Refined Transcript
speaker_map:
anyOf:
- $ref: '#/components/schemas/SpeakerMap'
- type: 'null'
content:
anyOf:
- type: string
- type: 'null'
title: Content
summary:
anyOf:
- type: string
- type: 'null'
title: Summary
notes:
anyOf:
- type: string
- type: 'null'
title: Notes
type: object
required:
- refine_status
title: MeetingStatusResponse
description: 'Refine-progress snapshot the desktop pulls as a backstop to sync.
The client polls ``GET /meetings/{id}/status`` with the refine
status it currently believes, and the server inlines whatever the
client is missing. ``refine_status`` is the server''s job status, or
``None`` when no refine job row exists for the meeting (the client
then has nothing to wait for). ``RefineJobStatus`` is a ``StrEnum``,
so it serializes to its string value on the wire.
``refined_transcript`` / ``content`` are inlined only when the
server is ahead of the client (see ``MeetingsService.get_meeting_status``);
both stay ``None`` otherwise, on a deleted meeting, and whenever the
refined blob is absent, undecodable, or empty.'
MeetingUploadUrlRequest:
properties:
content_type:
type: string
title: Content Type
description: Audio MIME type; one of audio/ogg, audio/wav
size_bytes:
type: integer
exclusiveMinimum: 0.0
title: Size Bytes
description: Encoded audio size in bytes
context:
anyOf:
- type: object
- type: 'null'
title: Context
description: Client meeting context (dictionary, participants, meeting metadata) reused server-side to enrich the refined transcript + summary. The same envelope the client sends on the live transcription WS.
include_google_contact_names:
type: boolean
title: Include Google Contact Names
description: 'Feature payload option: add per-user Google Contacts display names as weak refine terms.'
default: false
include_workspace_directory_names:
type: boolean
title: Include Workspace Directory Names
description: 'Feature payload option: add per-user workspace directory display names as weak refine terms.'
default: false
type: object
required:
- content_type
- size_bytes
title: MeetingUploadUrlRequest
description: Request for a presigned S3 slot to upload a meeting's refine audio.
RefineJobStatus:
type: string
enum:
- awaiting_audio
- transcribing
- summarizing
- complete
- failed
- abandoned
title: RefineJobStatus
MeetingSyncRequest:
properties:
last_sync_time:
anyOf:
- type: string
- type: 'null'
title: Last Sync Time
description: Unix timestamp in ms of the most recent server change the client has seen
hard_refresh:
type: boolean
title: Hard Refresh
description: When true, ignore last_sync_time and return up to 1000 meetings from server regardless of modified_at
default: false
cursor:
anyOf:
- type: string
- type: 'null'
title: Cursor
description: Opaque pagination cursor from the previous response's next_cursor. Refetch with the same last_sync_time and this cursor until next_cursor is null.
meetings:
items:
$ref: '#/components/schemas/SyncMeeting-Input'
type: array
title: Meetings
type: object
required:
- meetings
title: MeetingSyncRequest
description: Request model for meeting synchronization.
SyncMeeting-Input:
properties:
id:
type: string
format: uuid
title: Id
title:
type: string
title: Title
created_at:
type: string
title: Created At
modified_at:
type: string
title: Modified At
content:
type: string
title: Content
is_deleted:
type: boolean
title: Is Deleted
default: false
finalized:
type: boolean
title: Finalized
default: false
calendar_external_id:
anyOf:
- type: string
- type: 'null'
title: Calendar External Id
transcript_deleted_at:
anyOf:
- type: string
- type: 'null'
title: Transcript Deleted At
refined_transcript:
anyOf:
- type: string
- type: 'null'
title: Refined Transcript
readOnly: true
refine_status:
anyOf:
- $ref: '#/components/schemas/RefineJobStatus'
- type: 'null'
readOnly: true
refined_uploaded_at:
anyOf:
- type: string
- type: 'null'
title: Refined Uploaded At
readOnly: true
slug:
anyOf:
- type: string
- type: 'null'
title: Slug
readOnly: true
visibility:
anyOf:
- $ref: '#/components/schemas/NoteVisibility'
- type: 'null'
readOnly: true
notes:
anyOf:
- type: string
- type: 'null'
title: Notes
summary:
anyOf:
- type: string
- type: 'null'
title: Summary
live_transcript_uploaded:
anyOf:
- type: boolean
- type: 'null'
title: Live Transcript Uploaded
calendar_event_id:
anyOf:
- type: string
- type: 'null'
title: Calendar Event Id
ended_at:
anyOf:
- type: string
format: date-time
- type: 'null'
title: Ended At
speaker_map:
anyOf:
- $ref: '#/components/schemas/SpeakerMap'
- type: 'null'
import_source:
anyOf:
- type: string
- type: 'null'
title: Import Source
type: object
required:
- id
- title
- created_at
- modified_at
- content
title: SyncMeeting
description: 'Model for meeting data in sync requests.
The live transcript is uploaded out-of-band via a presigned S3 PUT;
the client then sets ``live_transcript_uploaded`` on its next sync and
the server stamps the server-internal ``live_uploaded_at`` watermark
without the bytes. The live transcript is never inlined on the wire —
cross-device live view is unsupported.
``refined_transcript`` is pull-only. The server inlines the
refined NDJSON body when ``refined_uploaded_at`` advanced past the
client''s watermark. The payload is one ``TranscriptEntry`` per line; the
field is named without a content-type suffix because the client treats it as
"refined transcript", not "refined JSON". Skipped on hard refresh —
a 100-meeting backlog × 200 KB each would blow up the response when
a device catches up after a long offline window.
``refined_uploaded_at`` rides every pull (hard refresh included,
which skips the inline bytes) as the client''s reconciliation
signal: when the row''s stamp beats the client''s locally persisted
fetched-through stamp, it downloads the bytes out-of-band via
``GET /meetings/{id}/refined-transcript``. Response-only — a push
carrying it never writes the column. ``live_uploaded_at`` stays
deliberately NOT on the wire: cross-device live view is
unsupported, so clients have no use for that watermark.'
MeetingRefinedTranscriptUrlResponse:
properties:
url:
type: string
pattern: ^https?://
format: uri
title: Url
expires_at:
type: string
format: date-time
title: Expires At
description: Absolute UTC expiry of the presigned GET
type: object
required:
- url
- expires_at
title: MeetingRefinedTranscriptUrlResponse
description: 'Presigned S3 GET for a meeting''s refined-transcript NDJSON.
The download twin of :class:`MeetingUploadUrlResponse` (same shape as the
notes ``NoteImageKey`` payload): the desktop calls
``GET /meetings/{id}/refined-transcript`` when a pull row''s
``refined_uploaded_at`` stamp beats its locally persisted fetched-through
stamp, then downloads ``refined.ndjson`` straight from S3 — the bytes
never transit the API.'
SyncMeeting-Output:
properties:
id:
type: string
format: uuid
title: Id
title:
type: string
title: Title
created_at:
type: string
title: Created At
modified_at:
type: string
title: Modified At
content:
type: string
title: Content
is_deleted:
type: boolean
title: Is Deleted
default: false
finalized:
type: boolean
title: Finalized
default: false
calendar_external_id:
anyOf:
- type: string
- type: 'null'
title: Calendar External Id
transcript_deleted_at:
anyOf:
- type: string
- type: 'null'
title: Transcript Deleted At
refined_transcript:
anyOf:
- type: string
- type: 'null'
title: Refined Transcript
readOnly: true
refine_status:
anyOf:
- $ref: '#/components/schemas/RefineJobStatus'
- type: 'null'
readOnly: true
refined_uploaded_at:
anyOf:
- type: string
- type: 'null'
title: Refined Uploaded At
readOnly: true
slug:
anyOf:
- type: string
- type: 'null'
title: Slug
readOnly: true
visibility:
anyOf:
- $ref: '#/components/schemas/NoteVisibility'
- type: 'null'
readOnly: true
notes:
anyOf:
- type: string
- type: 'null'
title: Notes
summary:
anyOf:
- type: string
- type: 'null'
title: Summary
live_transcript_uploaded:
anyOf:
- type: boolean
- type: 'null'
title: Live Transcript Uploaded
calendar_event_id:
anyOf:
- type: string
- type: 'null'
title: Calendar Event Id
ended_at:
anyOf:
- type: string
format: date-time
- type: 'null'
title: Ended At
speaker_map:
anyOf:
- $ref: '#/components/schemas/SpeakerMap'
- type: 'null'
import_source:
anyOf:
- type: string
- type: 'null'
title: Import Source
type: object
required:
- id
- title
- created_at
- modified_at
- content
title: SyncMeeting
description: 'Model for meeting data in sync requests.
The live transcript is uploaded out-of-band via a presigned S3 PUT;
the client then sets ``live_transcript_uploaded`` on its next sync and
the server stamps the server-internal ``live_uploaded_at`` watermark
without the bytes. The live transcript is never inlined on the wire —
cross-device live view is unsupported.
``refined_transcript`` is pull-only. The server inlines the
refined NDJSON body when ``refined_uploaded_at`` advanced past the
client''s watermark. The payload is one ``TranscriptEntry`` per line; the
field is named without a content-type suffix because the client treats it as
"refined transcript", not "refined JSON". Skipped on hard refresh —
a 100-meeting backlog × 200 KB each would blow up the response when
a device catches up after a long offline window.
``refined_uploaded_at`` rides every pull (hard refresh included,
which skips the inline bytes) as the client''s reconciliation
signal: when the row''s stamp beats the client''s locally persisted
fetched-through stamp, it downloads the bytes out-of-band via
``GET /meetings/{id}/refined-transcript``. Response-only — a push
carrying it never writes the column. ``live_uploaded_at`` stays
deliberately NOT on the wire: cross-device live view is
unsupported, so clients have no use for that watermark.'
ValidationError:
properties:
loc:
items:
anyOf:
- type: string
- type: integer
type: array
title: Location
msg:
type: string
title: Message
type:
type: string
title: Error Type
input:
title: Input
ctx:
type: object
title: Context
type: object
required:
- loc
- msg
- type
title: ValidationError
SpeakerObservationsUploadUrlRequest:
properties:
content_type:
type: string
title: Content Type
description: Observation-log MIME type; must be application/x-ndjson
size_bytes:
type: integer
exclusiveMinimum: 0.0
title: Size Bytes
description: NDJSON body size in bytes
type: object
required:
- content_type
- size_bytes
title: SpeakerObservationsUploadUrlRequest
description: 'Request for a presigned S3 slot to upload a meeting''s speaker-observation log.
Distinct from :class:`MeetingUploadUrlRequest`: NDJSON not audio, an
8 MiB cap not 500 MiB, and no ``context`` passthrough — this artifact
carries observations the server reduces into canonical speakers, not
refine/summary context.'
SpeakerMap:
properties:
people:
additionalProperties:
$ref: '#/components/schemas/SpeakerPerson'
type: object
maxProperties: 512
title: People
assignments:
additionalProperties:
$ref: '#/components/schemas/SpeakerAssignment'
type: object
maxProperties: 512
title: Assignments
additionalProperties: false
type: object
title: SpeakerMap
description: 'The ``meetings.speaker_map`` blob (nullable JSONB on the meeting row).
``people`` is keyed by person id (a plain uuid4 string, minted by whoever
creates the person); ``assignments`` is keyed by ``refined_id`` (the
transcript ``speaker.id``, serialized as a string key).'
NoteVisibility:
type: string
enum:
- anyone_with_link
- domain
- invited_only
title: NoteVisibility
MeetingUploadUrlResponse:
properties:
url:
type: string
title: Url
fields:
additionalProperties:
type: string
type: object
title: Fields
expires_at:
type: string
title: Expires At
description: ISO-8601 UTC expiry of the presigned POST
type: object
required:
- url
- fields
- expires_at
title: MeetingUploadUrlResponse
description: 'Presigned S3 POST the desktop uses to upload refine audio directly.
The S3 object key is carried inside ``fields[''key'']`` per the S3
presigned-POST contract, so it is not duplicated as a top-level field.'
MeetingSyncResponse:
properties:
acked:
items:
type: string
format: uuid
type: array
title: Acked
pull:
items:
$ref: '#/components/schemas/SyncMeeting-Output'
type: array
title: Pull
sync_time:
type: string
title: Sync Time
next_cursor:
anyOf:
- type: string
- type: 'null'
title: Next Cursor
description: Non-null when more changed meetings remain; refetch with it and the same last_sync_time.
type: object
required:
- acked
- pull
- sync_time
title: MeetingSyncResponse
description: Response model for meeting synchronization.
LiveTranscriptUploadUrlRequest:
properties:
content_type:
type: string
title: Content Type
description: Live-transcript MIME type; must be application/x-ndjson
size_bytes:
type: integer
exclusiveMinimum: 0.0
title: Size Bytes
description: NDJSON body size in bytes
type: object
required:
- content_type
- size_bytes
title: LiveTranscriptUploadUrlRequest
description: 'Request for a presigned S3 slot to upload a meeting''s live-transcript NDJSON.
Sibling to :class:`SpeakerObservationsUploadUrlRequest`: same NDJSON
content type, but the 64 MiB ``LIVE_TRANSCRIPT_MAX_BYTES`` cap and a
distinct server-composed key.'
SpeakerAssignment:
properties:
consensus:
anyOf:
- type: string
maxLength: 64
- type: 'null'
title: Consensus
user:
anyOf:
- type: string
maxLength: 64
- type: 'null'
title: User
mic:
anyOf:
- type: string
maxLength: 64
- type: 'null'
title: Mic
dom:
anyOf:
- type: string
maxLength: 64
- type:
# --- truncated at 32 KB (33 KB total) ---
# Full source: https://raw.githubusercontent.com/api-evangelist/wispr-ai/refs/heads/main/openapi/wispr-ai-meetings-api-openapi.yml