Login Book a demo

Runs

Send runs

Send your agent’s conversations to Ollive to be checked.

POSThttps://api.ollive.ai/api/runs/create/v1
v1

How it works

Use this for agents that don’t run on a connected platform. Send 1 to 10 runs per request, each up to 1,000 messages. Ollive accepts them with a 202 and checks them in the background. When checking is done, read the results with List results and the other monitoring endpoints.

Give each run a run_id: your own ID for it. A run can’t be changed after it is sent. Sending the same run_id again is safe: it comes back as already_sent and isn’t checked twice.

If anything in the request is invalid, none of it is accepted, and the error’s param points to the field to fix. Times can’t be in the future, and an end time can’t come before its start time.

Authorization

  • Authorizationheaderrequired

    Your API key, after the word Bearer, like this: Bearer ollive_sk_…. Create a key.

Request body

application/json

  • agent_idstringrequired

    The agent these runs belong to, from List agents.

  • runsobject[]required1 to 10

    The runs to send.

    • run_idstringrequiredup to 128 characters

      Your ID for this run. Letters, numbers, and . _ : -.

    • channelstring

      Whether it was a chat or a voice conversation.

      chatvoice
    • started_atstring (date-time)

      When the run started. Defaults to the first message’s time.

    • ended_atstring (date-time)

      When it ended. Defaults to the last message’s time.

    • messagesobject[]required1 to 1,000

      The conversation, in order. A new turn starts at each user message that follows an assistant message.

      • rolestringrequired

        Who said it.

        userassistantsystem
      • contentstringrequiredup to 100,000 characters

        What they said.

      • atstring (date-time)

        When it was said.

      • tool_callsobject[]

        The tools the agent used for this message. Assistant messages only.

        • namestringrequired

          The tool’s name.

        • inputany JSONup to 100,000 characters

          What the agent sent to the tool.

        • outputany JSONup to 100,000 characters

          What the tool returned.

        • started_atstring (date-time)

          When the call started.

        • ended_atstring (date-time)

          When it finished.

Response

Accepted. The runs are listed in the order you sent them.

  • dataobject[]required

    One entry per run.

    • run_idstringrequired

      Your ID for the run.

    • statusstringrequired

      queued, or already_sent if this run_id was sent before. It isn’t checked again.

      queuedalready_sent
← Create an agentList runs →