Conversations

List conversations, and open one with its turns, findings and scores.

Four GET endpoints, for pulling conversations and their evidence into your own warehouse or dashboard. A read-scoped key, on the same host as everything else.

The endpoints

`GET /v1/threads`	The list, newest first, cursor paged
`GET /v1/threads/{thread_id}`	One conversation's own record
`GET /v1/threads/{thread_id}/sessions`	The calls or chats inside it
`GET /v1/sessions/{session_id}`	One call or chat in full, with its evidence

Listing conversations

bash

curl "https://app.evidova.com/v1/threads?limit=50&channel=voice" \
  -H "authorization: Bearer ak_live_..."

state	The conversation's lifecycle state
client_id	One client. A client-limited key may name only its own
channel	voice or chat
from, to	The window to read
limit	1 to 200. Defaults to 50
cursor	The next_cursor from the previous page

The response is { items, next_cursor }. A null next_cursor is the last page.

What comes back

One conversation

id, contact_key	Ours, and the key that groups repeat contact from one customer
agent_id, client_id	Which agent spoke, and whose customer it was
state, outcome, funnel_stage	Where it ended up, and how far it got
first_event_at, last_event_at	The window it spans
facts	What we extracted from it

One session, in full

GET /v1/sessions/{session_id} is the leaf, and the one worth linking to from your own tooling. It returns the transcript plus everything we concluded from it.

transcript	The turns, in order
findings	Which rule broke on which turn
scores	The numbers, per score area
tool_events	What the agent called, and what came back
latency, latency_flags	Reply times, and which ones we called slow
versions	Which rules version and which instrument produced the above

A key limited to one client gets 404 for a conversation belonging to another, not 403. The surface never confirms that a row it will not serve exists.