Method chatCompletion

pkg/chatserver/server.go:474–502

chatCompletion runs the agent to completion and replies with one non-streaming OpenAI ChatCompletion object. It returns the agent run error (nil on success) so the caller can decide whether to commit the conversation; the HTTP response — success or error envelope — is always written here.

(c echo.Context, rt runtime.Runtime, sess *session.Session, model string)

// conversation; the HTTP response — success or error envelope — is always
// written here.
func (s *server) chatCompletion(c echo.Context, rt runtime.Runtime, sess *session.Session, model string) error {
	var toolCalls []ToolCallReference
	emit := agentEmit{
		onToolCall: func(tc ToolCallReference) {
			toolCalls = append(toolCalls, tc)
		},
	}
	if err := runAgentLoop(c.Request().Context(), rt, sess, emit); err != nil {
		_ = writeError(c, http.StatusInternalServerError, fmt.Sprintf("agent execution failed: %v", err))
		return err
	}

	return c.JSON(http.StatusOK, ChatCompletionResponse{
		ID:      newChatID(),
		Object:  "chat.completion",
		Created: time.Now().Unix(),
		Model:   model,
		Choices: []ChatCompletionChoice{{
			Index: 0,
			Message: ChatCompletionMessage{
				Role:      "assistant",
				Content:   sess.GetLastAssistantMessageContent(),
				ToolCalls: toolCalls,
			},
			FinishReason: "stop",
		}},
		Usage: sessionUsage(sess),
	})
}

// streamChatCompletion runs the agent and streams its response back to the
// client as Server-Sent Events in OpenAI's chat.completion.chunk format.