See what the agent reads of a chat

See what the chat's agent reads of the chat when its next turn starts: its newest messages word for word, and a summary of the ones before them.

A turn reads as many of the newest messages as fit its share of the model's input; verbatim_from_message_id is the oldest of them, and is null while it reads the whole chat. After a turn ends, a model folds the messages that no longer fit into the summary, and the turn's run is charged for it.

unsummarized_message_count counts messages a turn would neither read nor find in the summary, until the summary catches up. The agent can still read any message back when it needs the exact words.

Required access

Credentials: delegated token, user session, user API key

User API keys use the owner’s current permissions, subject to any key restrictions. Capability scopes below apply to automation account keys and delegated tokens.

Scope: chats:read

Only the member who started the chat, through their user API key or an integration acting for them: a chat is private to them. Another member's chat answers 404, also to a workspace owner or admin, who may only stop its turn or delete it, signed in to the app, and reads none of it. Automation accounts have no chats.

Path Parameters

Header Parameters

Response Body

curl -X GET "https://example.com/api/v1/chats/497f6eca-6276-4993-bfeb-53cbbbba6f08/summary" \  -H "Authorization: Bearer $GUEST_MACHINES_API_KEY"

{  "summary": {    "content": "string",    "through_message_id": 0,    "message_count": 1,    "model": "string",    "updated_at": "2019-08-24T14:15:22Z"  },  "verbatim_from_message_id": 0,  "earlier_message_count": 0,  "unsummarized_message_count": 0,  "total_messages": 0}