mirror of
https://github.com/router-for-me/CLIProxyAPI.git
synced 2026-02-02 04:20:50 +08:00
260 lines
13 KiB
Go
260 lines
13 KiB
Go
// Package claude provides response translation functionality for Claude Code API compatibility.
|
|
// This package handles the conversion of backend client responses into Claude Code-compatible
|
|
// Server-Sent Events (SSE) format, implementing a sophisticated state machine that manages
|
|
// different response types including text content, thinking processes, and function calls.
|
|
// The translation ensures proper sequencing of SSE events and maintains state across
|
|
// multiple response chunks to provide a seamless streaming experience.
|
|
package claude
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"fmt"
|
|
"time"
|
|
|
|
log "github.com/sirupsen/logrus"
|
|
"github.com/tidwall/gjson"
|
|
"github.com/tidwall/sjson"
|
|
)
|
|
|
|
// Params holds parameters for response conversion and maintains state across streaming chunks.
|
|
// This structure tracks the current state of the response translation process to ensure
|
|
// proper sequencing of SSE events and transitions between different content types.
|
|
type Params struct {
|
|
HasFirstResponse bool // Indicates if the initial message_start event has been sent
|
|
ResponseType int // Current response type: 0=none, 1=content, 2=thinking, 3=function
|
|
ResponseIndex int // Index counter for content blocks in the streaming response
|
|
}
|
|
|
|
// ConvertGeminiCLIResponseToClaude performs sophisticated streaming response format conversion.
|
|
// This function implements a complex state machine that translates backend client responses
|
|
// into Claude Code-compatible Server-Sent Events (SSE) format. It manages different response types
|
|
// and handles state transitions between content blocks, thinking processes, and function calls.
|
|
//
|
|
// Response type states: 0=none, 1=content, 2=thinking, 3=function
|
|
// The function maintains state across multiple calls to ensure proper SSE event sequencing.
|
|
//
|
|
// Parameters:
|
|
// - ctx: The context for the request, used for cancellation and timeout handling
|
|
// - modelName: The name of the model being used for the response (unused in current implementation)
|
|
// - rawJSON: The raw JSON response from the Gemini CLI API
|
|
// - param: A pointer to a parameter object for maintaining state between calls
|
|
//
|
|
// Returns:
|
|
// - []string: A slice of strings, each containing a Claude Code-compatible JSON response
|
|
func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalRequestRawJSON, requestRawJSON, rawJSON []byte, param *any) []string {
|
|
log.Debug("ConvertGeminiCLIResponseToClaude")
|
|
if *param == nil {
|
|
*param = &Params{
|
|
HasFirstResponse: false,
|
|
ResponseType: 0,
|
|
ResponseIndex: 0,
|
|
}
|
|
}
|
|
|
|
if bytes.Equal(rawJSON, []byte("[DONE]")) {
|
|
return []string{
|
|
"event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n",
|
|
}
|
|
}
|
|
|
|
// Track whether tools are being used in this response chunk
|
|
usedTool := false
|
|
output := ""
|
|
|
|
// Initialize the streaming session with a message_start event
|
|
// This is only sent for the very first response chunk to establish the streaming session
|
|
if !(*param).(*Params).HasFirstResponse {
|
|
output = "event: message_start\n"
|
|
|
|
// Create the initial message structure with default values according to Claude Code API specification
|
|
// This follows the Claude Code API specification for streaming message initialization
|
|
messageStartTemplate := `{"type": "message_start", "message": {"id": "msg_1nZdL29xx5MUA1yADyHTEsnR8uuvGzszyY", "type": "message", "role": "assistant", "content": [], "model": "claude-3-5-sonnet-20241022", "stop_reason": null, "stop_sequence": null, "usage": {"input_tokens": 0, "output_tokens": 0}}}`
|
|
|
|
// Override default values with actual response metadata if available from the Gemini CLI response
|
|
if modelVersionResult := gjson.GetBytes(rawJSON, "response.modelVersion"); modelVersionResult.Exists() {
|
|
messageStartTemplate, _ = sjson.Set(messageStartTemplate, "message.model", modelVersionResult.String())
|
|
}
|
|
if responseIDResult := gjson.GetBytes(rawJSON, "response.responseId"); responseIDResult.Exists() {
|
|
messageStartTemplate, _ = sjson.Set(messageStartTemplate, "message.id", responseIDResult.String())
|
|
}
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", messageStartTemplate)
|
|
|
|
(*param).(*Params).HasFirstResponse = true
|
|
}
|
|
|
|
// Process the response parts array from the backend client
|
|
// Each part can contain text content, thinking content, or function calls
|
|
partsResult := gjson.GetBytes(rawJSON, "response.candidates.0.content.parts")
|
|
if partsResult.IsArray() {
|
|
partResults := partsResult.Array()
|
|
for i := 0; i < len(partResults); i++ {
|
|
partResult := partResults[i]
|
|
|
|
// Extract the different types of content from each part
|
|
partTextResult := partResult.Get("text")
|
|
functionCallResult := partResult.Get("functionCall")
|
|
|
|
// Handle text content (both regular content and thinking)
|
|
if partTextResult.Exists() {
|
|
// Process thinking content (internal reasoning)
|
|
if partResult.Get("thought").Bool() {
|
|
// Continue existing thinking block if already in thinking state
|
|
if (*param).(*Params).ResponseType == 2 {
|
|
output = output + "event: content_block_delta\n"
|
|
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
} else {
|
|
// Transition from another state to thinking
|
|
// First, close any existing content block
|
|
if (*param).(*Params).ResponseType != 0 {
|
|
if (*param).(*Params).ResponseType == 2 {
|
|
// output = output + "event: content_block_delta\n"
|
|
// output = output + fmt.Sprintf(`data: {"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":null}}`, (*param).(*Params).ResponseIndex)
|
|
// output = output + "\n\n\n"
|
|
}
|
|
output = output + "event: content_block_stop\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
(*param).(*Params).ResponseIndex++
|
|
}
|
|
|
|
// Start a new thinking content block
|
|
output = output + "event: content_block_start\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_start","index":%d,"content_block":{"type":"thinking","thinking":""}}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
output = output + "event: content_block_delta\n"
|
|
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
(*param).(*Params).ResponseType = 2 // Set state to thinking
|
|
}
|
|
} else {
|
|
// Process regular text content (user-visible output)
|
|
// Continue existing text block if already in content state
|
|
if (*param).(*Params).ResponseType == 1 {
|
|
output = output + "event: content_block_delta\n"
|
|
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
} else {
|
|
// Transition from another state to text content
|
|
// First, close any existing content block
|
|
if (*param).(*Params).ResponseType != 0 {
|
|
if (*param).(*Params).ResponseType == 2 {
|
|
// output = output + "event: content_block_delta\n"
|
|
// output = output + fmt.Sprintf(`data: {"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":null}}`, (*param).(*Params).ResponseIndex)
|
|
// output = output + "\n\n\n"
|
|
}
|
|
output = output + "event: content_block_stop\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
(*param).(*Params).ResponseIndex++
|
|
}
|
|
|
|
// Start a new text content block
|
|
output = output + "event: content_block_start\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_start","index":%d,"content_block":{"type":"text","text":""}}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
output = output + "event: content_block_delta\n"
|
|
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
(*param).(*Params).ResponseType = 1 // Set state to content
|
|
}
|
|
}
|
|
} else if functionCallResult.Exists() {
|
|
// Handle function/tool calls from the AI model
|
|
// This processes tool usage requests and formats them for Claude Code API compatibility
|
|
usedTool = true
|
|
fcName := functionCallResult.Get("name").String()
|
|
|
|
// Handle state transitions when switching to function calls
|
|
// Close any existing function call block first
|
|
if (*param).(*Params).ResponseType == 3 {
|
|
output = output + "event: content_block_stop\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
(*param).(*Params).ResponseIndex++
|
|
(*param).(*Params).ResponseType = 0
|
|
}
|
|
|
|
// Special handling for thinking state transition
|
|
if (*param).(*Params).ResponseType == 2 {
|
|
// output = output + "event: content_block_delta\n"
|
|
// output = output + fmt.Sprintf(`data: {"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":null}}`, (*param).(*Params).ResponseIndex)
|
|
// output = output + "\n\n\n"
|
|
}
|
|
|
|
// Close any other existing content block
|
|
if (*param).(*Params).ResponseType != 0 {
|
|
output = output + "event: content_block_stop\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
(*param).(*Params).ResponseIndex++
|
|
}
|
|
|
|
// Start a new tool use content block
|
|
// This creates the structure for a function call in Claude Code format
|
|
output = output + "event: content_block_start\n"
|
|
|
|
// Create the tool use block with unique ID and function details
|
|
data := fmt.Sprintf(`{"type":"content_block_start","index":%d,"content_block":{"type":"tool_use","id":"","name":"","input":{}}}`, (*param).(*Params).ResponseIndex)
|
|
data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
|
|
data, _ = sjson.Set(data, "content_block.name", fcName)
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
|
|
if fcArgsResult := functionCallResult.Get("args"); fcArgsResult.Exists() {
|
|
output = output + "event: content_block_delta\n"
|
|
data, _ = sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"input_json_delta","partial_json":""}}`, (*param).(*Params).ResponseIndex), "delta.partial_json", fcArgsResult.Raw)
|
|
output = output + fmt.Sprintf("data: %s\n\n\n", data)
|
|
}
|
|
(*param).(*Params).ResponseType = 3
|
|
}
|
|
}
|
|
}
|
|
|
|
usageResult := gjson.GetBytes(rawJSON, "response.usageMetadata")
|
|
// Process usage metadata and finish reason when present in the response
|
|
if usageResult.Exists() && bytes.Contains(rawJSON, []byte(`"finishReason"`)) {
|
|
if candidatesTokenCountResult := usageResult.Get("candidatesTokenCount"); candidatesTokenCountResult.Exists() {
|
|
// Close the final content block
|
|
output = output + "event: content_block_stop\n"
|
|
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
|
|
output = output + "\n\n\n"
|
|
|
|
// Send the final message delta with usage information and stop reason
|
|
output = output + "event: message_delta\n"
|
|
output = output + `data: `
|
|
|
|
// Create the message delta template with appropriate stop reason
|
|
template := `{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
|
|
// Set tool_use stop reason if tools were used in this response
|
|
if usedTool {
|
|
template = `{"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
|
|
}
|
|
|
|
// Include thinking tokens in output token count if present
|
|
thoughtsTokenCount := usageResult.Get("thoughtsTokenCount").Int()
|
|
template, _ = sjson.Set(template, "usage.output_tokens", candidatesTokenCountResult.Int()+thoughtsTokenCount)
|
|
template, _ = sjson.Set(template, "usage.input_tokens", usageResult.Get("promptTokenCount").Int())
|
|
|
|
output = output + template + "\n\n\n"
|
|
}
|
|
}
|
|
|
|
return []string{output}
|
|
}
|
|
|
|
// ConvertGeminiCLIResponseToClaudeNonStream converts a non-streaming Gemini CLI response to a non-streaming Claude response.
|
|
//
|
|
// Parameters:
|
|
// - ctx: The context for the request.
|
|
// - modelName: The name of the model.
|
|
// - rawJSON: The raw JSON response from the Gemini CLI API.
|
|
// - param: A pointer to a parameter object for the conversion.
|
|
//
|
|
// Returns:
|
|
// - string: A Claude-compatible JSON response.
|
|
func ConvertGeminiCLIResponseToClaudeNonStream(_ context.Context, _ string, originalRequestRawJSON, requestRawJSON, _ []byte, _ *any) string {
|
|
log.Debug("ConvertGeminiCLIResponseToClaudeNonStream")
|
|
return ""
|
|
}
|