agents

Context for a prompt

One block of text ready for a system prompt, in three lowercase sections: what the agent knows, what it may know of the person it works for (only what its settings let it read, each with whose words they are: "they said", "noted" or "inferred"), and what it knows of this user (when user_id is given). Ranked, then trimmed from the bottom to budget_tokens (about four characters a token). A search, so the read scope is enough.

post/v1/agents/{agent}/context

Request

  • Base URL: https://innernet.live/api
  • URL: https://innernet.live/api/v1/agents/{agent}/context
  • Auth: HTTP bearer

Path parameters

agentstring required

The agent's slug, its id, or me (the agent a key made for one agent is bound to). Another agent, by any name, is a 404 to an agent key.

Request body

querystring
user_idstring
run_idstring
budget_tokensinteger

Response

OK

textstring

Empty when there is nothing to add.

personobject[]

The person's facts it carries (same shape as recall).

Changes

No recorded changes to this operation across all 1 revision of this API.