CrossRelay/internal/responses/convert.go

package responses

import (
	"crypto/rand"
	"encoding/hex"
	"encoding/json"
	"fmt"
	"strings"
	"time"

	"gitea.rutua.cn/kate522/CrossRelay/internal/chatcmpl"
)

// ConvertRequest converts a Responses request into a Chat Completions
// request addressed to the given upstream model.
func ConvertRequest(req *Request, model string) (*chatcmpl.Request, error) {
	out := &chatcmpl.Request{
		Model:       model,
		MaxTokens:   req.MaxOutputTokens,
		Temperature: req.Temperature,
		TopP:        req.TopP,
		Stream:      req.Stream,
	}
	items, err := ParseInput(req.Input)
	if err != nil {
		return nil, fmt.Errorf("parse input: %w", err)
	}

	var systemParts []string
	if s := strings.TrimSpace(req.Instructions); s != "" {
		systemParts = append(systemParts, s)
	}
	var assistantText []string
	var assistantCalls []chatcmpl.ToolCall
	flushAssistant := func() {
		if len(assistantText) == 0 && len(assistantCalls) == 0 {
			return
		}
		msg := chatcmpl.Message{Role: "assistant"}
		if len(assistantText) > 0 {
			msg.Content = strings.Join(assistantText, "\n\n")
		}
		msg.ToolCalls = assistantCalls
		out.Messages = append(out.Messages, msg)
		assistantText = nil
		assistantCalls = nil
	}

	for _, it := range items {
		switch it.Type {
		case "reasoning":
			// Upstream Chat Completions has no reasoning channel; the
			// item is dropped rather than replayed as spoken text.
			continue
		case "message":
			parts, err := ParseContent(it.Content)
			if err != nil {
				return nil, fmt.Errorf("parse message content: %w", err)
			}
			switch it.Role {
			case "system", "developer":
				flushAssistant()
				for _, p := range parts {
					if isTextPart(p) {
						systemParts = append(systemParts, p.Text)
					}
				}
			case "assistant":
				for _, p := range parts {
					if isTextPart(p) {
						assistantText = append(assistantText, p.Text)
					}
				}
			default:
				flushAssistant()
				out.Messages = append(out.Messages, userMessage(parts))
			}
		case "function_call":
			args := it.Arguments
			if strings.TrimSpace(args) == "" {
				args = "{}"
			}
			assistantCalls = append(assistantCalls, chatcmpl.ToolCall{
				ID:       it.CallID,
				Type:     "function",
				Function: chatcmpl.ToolCallFunction{Name: it.Name, Arguments: args},
			})
		case "function_call_output":
			flushAssistant()
			out.Messages = append(out.Messages, chatcmpl.Message{
				Role:       "tool",
				ToolCallID: it.CallID,
				Content:    OutputText(it.Output),
			})
		}
	}
	flushAssistant()

	if len(systemParts) > 0 {
		system := chatcmpl.Message{Role: "system", Content: strings.Join(systemParts, "\n\n")}
		out.Messages = append([]chatcmpl.Message{system}, out.Messages...)
	}
	for _, t := range req.Tools {
		params := t.Parameters
		if t.Type != "function" {
			// Freeform tool forms (custom, local_shell) become function
			// tools with a permissive schema so upstreams that only
			// speak function calling still work.
			params = json.RawMessage(`{"type":"object","additionalProperties":true}`)
		}
		out.Tools = append(out.Tools, chatcmpl.Tool{
			Type: "function",
			Function: chatcmpl.ToolFunction{
				Name:        t.Name,
				Description: t.Description,
				Parameters:  params,
			},
		})
	}
	choice, err := convertToolChoice(req.ToolChoice)
	if err != nil {
		return nil, err
	}
	out.ToolChoice = choice
	if req.Stream {
		out.StreamOptions = &chatcmpl.StreamOptions{IncludeUsage: true}
	}
	return out, nil
}

func isTextPart(p ContentPart) bool {
	switch p.Type {
	case "input_text", "output_text", "text", "refusal":
		return true
	default:
		return false
	}
}

func userMessage(parts []ContentPart) chatcmpl.Message {
	var cps []chatcmpl.ContentPart
	for _, p := range parts {
		switch {
		case p.Type == "input_image" && p.ImageURL != "":
			cps = append(cps, chatcmpl.ContentPart{Type: "image_url", ImageURL: &chatcmpl.ImageURL{URL: p.ImageURL}})
		case isTextPart(p) && p.Text != "":
			cps = append(cps, chatcmpl.ContentPart{Type: "text", Text: p.Text})
		}
	}
	if len(cps) == 0 {
		return chatcmpl.Message{Role: "user", Content: ""}
	}
	if len(cps) == 1 && cps[0].Type == "text" {
		return chatcmpl.Message{Role: "user", Content: cps[0].Text}
	}
	return chatcmpl.Message{Role: "user", Content: cps}
}

func convertToolChoice(raw json.RawMessage) (any, error) {
	if len(raw) == 0 {
		return nil, nil
	}
	var s string
	if json.Unmarshal(raw, &s) == nil {
		switch s {
		case "auto", "none", "required":
			return s, nil
		default:
			return nil, nil
		}
	}
	var obj struct {
		Type string `json:"type"`
		Name string `json:"name"`
	}
	if json.Unmarshal(raw, &obj) == nil && obj.Type == "function" && obj.Name != "" {
		return map[string]any{
			"type":     "function",
			"function": map[string]any{"name": obj.Name},
		}, nil
	}
	// Unknown tool choice shape: let the upstream decide.
	return nil, nil
}

// ConvertResponse converts a Chat Completions response into a Responses
// response. requestedModel is echoed back to the client.
func ConvertResponse(resp *chatcmpl.Response, requestedModel string) *Response {
	out := &Response{
		ID:        "resp_" + resp.ID,
		Object:    "response",
		CreatedAt: time.Now().Unix(),
		Status:    "completed",
		Model:     requestedModel,
		Output:    []OutItem{},
	}
	var texts []string
	if len(resp.Choices) > 0 {
		c := resp.Choices[0]
		if txt := messageText(c.Message.Content); txt != "" {
			texts = append(texts, txt)
			out.Output = append(out.Output, OutItem{
				ID:      "msg_" + randHex(12),
				Type:    "message",
				Role:    "assistant",
				Status:  "completed",
				Content: []ContentPart{{Type: "output_text", Text: txt}},
			})
		}
		for _, tc := range c.Message.ToolCalls {
			args := tc.Function.Arguments
			if strings.TrimSpace(args) == "" {
				args = "{}"
			}
			out.Output = append(out.Output, OutItem{
				ID:        "fc_" + randHex(12),
				Type:      "function_call",
				Status:    "completed",
				CallID:    tc.ID,
				Name:      tc.Function.Name,
				Arguments: args,
			})
		}
		out.Status, out.IncompleteDetails = MapStatus(c.FinishReason)
	}
	out.OutputText = strings.Join(texts, "\n\n")
	if resp.Usage != nil {
		out.Usage = &Usage{
			InputTokens:  resp.Usage.PromptTokens,
			OutputTokens: resp.Usage.CompletionTokens,
			TotalTokens:  resp.Usage.TotalTokens,
		}
		if out.Usage.TotalTokens == 0 {
			out.Usage.TotalTokens = out.Usage.InputTokens + out.Usage.OutputTokens
		}
	}
	return out
}

func messageText(content any) string {
	switch v := content.(type) {
	case string:
		return v
	case []chatcmpl.ContentPart:
		var sb strings.Builder
		for _, p := range v {
			if p.Type == "text" {
				sb.WriteString(p.Text)
			}
		}
		return sb.String()
	default:
		return ""
	}
}

// MapStatus maps a Chat Completions finish_reason onto a Responses
// status plus incomplete reason.
func MapStatus(finish string) (string, *Incomplete) {
	switch finish {
	case "length":
		return "incomplete", &Incomplete{Reason: "max_output_tokens"}
	case "content_filter":
		return "incomplete", &Incomplete{Reason: "content_filter"}
	default:
		return "completed", nil
	}
}

// EstimateTokens returns a rough input-token estimate for a converted
// request. It fills in usage when the upstream stream carries none.
func EstimateTokens(req *chatcmpl.Request) int {
	total := len(req.Model) / 4
	for _, m := range req.Messages {
		total += len(m.Role) + 4
		total += len(messageText(m.Content)) / 4
		for _, tc := range m.ToolCalls {
			total += len(tc.Function.Name)/4 + len(tc.Function.Arguments)/4
		}
	}
	for _, t := range req.Tools {
		total += len(t.Function.Name)/4 + len(t.Function.Description)/4 + len(t.Function.Parameters)/4
	}
	if total < 1 {
		total = 1
	}
	return total
}

func randHex(n int) string {
	b := make([]byte, n)
	if _, err := rand.Read(b); err != nil {
		return fmt.Sprintf("%x", n)
	}
	return hex.EncodeToString(b)
}