conversations

Retrieve a conversation

Returns the full conversation: metadata, the AI summary, and every attached source with its content rendered as text. Also serves an analysis report (a notebook produced by a scheduled analysis, with no recording): its status is null and its sources are the documents it produced.

GET /api/v1/conversations/{id}

Path parameters

idREQUIRED

string

The conversation ID.

Query parameters

include_content boolean	`false` omits source text: every source's `content` is null. Use it when only the summary, key points, todos or related conversations are needed.
from_seconds number	Only include transcript turns that end at or after this offset in seconds. Other source types are unaffected.
to_seconds number	Only include transcript turns that start at or before this offset in seconds. Must be greater than `from_seconds`.

Response fields

id string
title string · nullable
created_at string · date-time	When the conversation was created in Klang. Use this as the next `created_after` value for incremental sync.
updated_at string · date-time
status "pending" · "ready" · "error" · nullable	Processing state. Wait for `ready` before reading `summary` or transcript content.
termination_reason "not_admitted" · "verification_required" · "captcha_required" · "bot_blocked" · "spawn_rejected" · "spawn_failed" · nullable	When `status` is `error` because a meeting-bot recording was terminated before audio could be captured, this explains why. `null` for any other error (e.g. transcription failure) and for non-error states. - `not_admitted` — the bot was never let into the meeting (host didn't admit it from the lobby/waiting room within the timeout). - `verification_required` — the meeting platform asked the bot to complete email verification, which it can't do. - `captcha_required` — the meeting platform asked the bot to solve a CAPTCHA, which it can't do. - `bot_blocked` — the host organisation blocks meeting bots, so the bot was refused before it could join. - `spawn_rejected` — the meeting platform refused the bot for another reason (e.g. the meeting can't be joined). - `spawn_failed` — the bot couldn't be sent to the meeting.
folder_id string · nullable	ID of the immediate folder containing the conversation. `null` for conversations at the workspace root.
digest string · nullable	AI-generated short summary (≤80 characters) with the key names and topics, optimised to distinguish this conversation from similar ones. Present in both list and retrieve responses.
tags array of object	Tags applied to the conversation (AI suggestions the user hasn't accepted are not included). Empty when the conversation has no tags. Use `GET /tags` for the full catalog including descriptions.
↳id string
↳name string
↳color string · nullable	Hex colour (e.g. `#2E7D32`), or null if the tag has no colour.
summary string · nullable	User-facing summary of the conversation: a short overview, then any topical sections and key principles as markdown bullets.
key_points array of object	The AI key points grouped by topic. Empty until the analysis has run. Older analyses without topics appear as one entry with a null topic.
↳topic string · nullable
↳points array of string
todos array of object	Todos linked to this conversation (AI suggestions excluded).
↳id string
↳title string · nullable
↳details string · nullable
↳status "open" · "done" · "dismissed"	Todo state. `open` = still to do, `done` = completed, `dismissed` = discarded by the user.
↳due_date string · date · nullable	Due date if one was set, otherwise null.
↳created_at string · date-time
↳completed_at string · date-time · nullable	When the todo was marked done; null unless `status` is `done`.
↳conversation_id string · nullable	The conversation this todo belongs to, or null for a personal todo not tied to any conversation.
↳assignee_id string · nullable	User id of the person the todo is assigned to; resolve it with `GET /members`.
↳source "manual" · "ai"	Whether the user wrote the todo (`manual`) or the AI extracted it (`ai`).
related array of object	Other recordings of the same recurring meeting, plus conversations the user has linked to this one (accepted relations only; AI suggestions the user has not confirmed are excluded). Empty when the workspace has relations disabled.
↳id string
↳title string · nullable
↳created_at string · date-time
↳relation_type string · nullable	`series` for another recording of the same recurring meeting, `ai` when the link was suggested by Klang and accepted by the user, `manual` when the user created it.
↳explanation string · nullable	One-sentence rationale for the link.
↳digest string · nullable	The related conversation's short AI digest.
sources array of object	A piece of input attached to a conversation. Klang's summary and analysis are derived from the conversation's full set of sources — so this is what you read when you want to see what the AI was working from. Most conversations have a `transcript` as their primary source plus any extras the user attached. Returned only by `GET /conversations/{id}`; list endpoints return the lighter `SourceLite` shape instead.
↳id string
↳type "transcript" · "document" · "text" · "file"	What kind of input this source is. - `transcript` — what was said in a recorded meeting. The full speaker-labeled, time-stamped text is in `content`. - `document` — a document authored inside Klang (agendas, notes, decision logs, etc.). - `text` — a snippet of plain text the user pasted into the conversation. - `file` — an uploaded file (PDF, slide deck, Word doc). Extracted plain text is in `content`.
↳title string · nullable
↳added_at string · date-time · nullable
↳duration_seconds integer · nullable	Recording length. `null` for non-transcript sources.
↳language string · nullable	ISO 639-3 code (e.g. `eng`, `swe`). `null` for non-transcript sources.
↳meeting_url string · nullable	Join URL of the video meeting the transcript was recorded from (Google Meet, Microsoft Teams, Zoom). `null` for transcripts from uploaded files and for non-transcript sources.
↳attendees array of string · email	Email addresses from the meeting's calendar attendee list. Empty for meetings without calendar attendee data and for non-transcript sources. This can differ from `participants`, which contains speakers detected in the transcript.
↳content string · nullable	Plain-text rendering of the source. For transcripts this is a speaker-labeled, time-stamped flat string. May be `null` if content extraction failed for an uploaded file.
↳participants array of object	Speakers detected on the transcript. Empty array for non-transcript sources. Returns the speaker label as stored — a real name if the user has labeled the speaker in Klang, or a raw label like `Speaker 1` / `Talare 1` if not.
↳name string	The speaker's name if labeled in Klang, otherwise the raw speaker label (e.g. `Speaker 1`).
↳person_id string · uuid	Stable speaker-profile id. Omitted when the participant is not linked to a named person.