Function buildOpenAiCompatibleBody

src/lib/llm-providers.ts:386–455

(
  config: LlmConfig,
  messages: ChatMessage[],
  overrides?: RequestOverrides,
)

}

function buildOpenAiCompatibleBody(
  config: LlmConfig,
  messages: ChatMessage[],
  overrides?: RequestOverrides,
): Record<string, unknown> {
  assertBigModelImageSupport(config, messages)
  const reasoning = effectiveReasoning(config, overrides)
  const body: Record<string, unknown> = buildOpenAiBody(messages, stripWireAgnosticOverrides(overrides))
  adaptOpenAiStrictCompletionBody(config, body)
  adaptKimiBody(config, body)
  adaptXiaomiMimoBody(config, body, reasoning)

  if (isDeepSeekEndpoint(config)) {
    // DeepSeek V4 thinking mode. `thinking.type=disabled` is the most
    // important path for ingestion/rewrite tasks: it prevents the model
    // from spending the whole response on `reasoning_content` with no
    // final `content`.
    if (supportsDeepSeekThinkingParam(config)) {
      if (reasoning.mode === "off") {
        body.thinking = { type: "disabled" }
      } else if (reasoning.mode !== "auto") {
        body.thinking = { type: "enabled" }
        if (reasoning.mode === "high" || reasoning.mode === "max") {
          body.reasoning_effort = reasoning.mode
        }
      }
    }
    return body
  }

  if (config.provider === "ollama") {
    // Ollama's OpenAI-compatible /v1/chat/completions maps reasoning
    // control onto `reasoning_effort` ("high"|"medium"|"low"|"none";
    // "none" disables thinking). This is the only lever that stops a
    // thinking-capable model — or a non-thinking one Ollama wraps with a
    // thinking template — from spending its entire token budget on
    // chain-of-thought and ending the stream with an empty `content`,
    // which surfaces to the user as the "produced N chars of reasoning,
    // but no actual response content" diagnostic. Until this, callers'
    // `reasoning: { mode: "off" }` (every structured ingest call) was
    // silently dropped on the Ollama path. Non-thinking models (gemma,
    // llama) ignore the field harmlessly. "max" has no Ollama analogue,
    // so it maps to the strongest supported level, "high".
    // See docs.ollama.com/api/openai-compatibility.
    if (reasoning.mode === "off") {
      body.reasoning_effort = "none"
    } else if (
      reasoning.mode === "low" ||
      reasoning.mode === "medium" ||
      reasoning.mode === "high"
    ) {
      body.reasoning_effort = reasoning.mode
    } else if (reasoning.mode === "max") {
      body.reasoning_effort = "high"
    }
    return body
  }