Harness

Run a bounded tool-calling model request

OpenAI-compatible chat-completion transport with Ploid-owned model routing. Requests are capped at 128 KB, 100 messages, 32 tools, 4,096 output tokens, a 90-second total upstream lifetime, and a 1 MB response. stream=true returns SSE and must omit Idempotency-Key. Requires agent:chat.

post/v1/harness/chat/completions

Request

  • Base URL: https://api.ploid.com
  • URL: https://api.ploid.com/v1/harness/chat/completions
  • Auth: one of:
    • HTTP bearer
    • API key in header x-api-key

Headers

Idempotency-Keystring

Recommended on non-streaming POSTs. Reusing the same key and authenticated request within 24 hours replays the original response and bills once. Live SSE responses reject this header; use a durable async run for replayable work.

Request body

model'ploid/atlas' | 'ploid/ox'

Ploid Atlas uses Luna; Ploid Ox uses Sol.

messagesobject[] required
toolsobject[]
parallel_tool_callsboolean
response_formatobject
streamboolean
temperaturenumber
top_pnumber
seedinteger
max_tokensinteger

Requested output budget.

max_completion_tokensinteger

Requested output budget.

Response

Completed JSON response or terminally metered SSE stream.

HarnessChatCompletionResponse required

OpenAI-compatible completion payload with internal model/provider routing metadata removed.

Changes

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