mirror of
https://github.com/router-for-me/CLIProxyAPI.git
synced 2026-02-03 04:50:52 +08:00
Add openai codex support
This commit is contained in:
Luis Pater
@@ -1,208 +0,0 @@
|
||||
// Package claude provides HTTP handlers for Claude API code-related functionality.
|
||||
// This package implements Claude-compatible streaming chat completions with sophisticated
|
||||
// client rotation and quota management systems to ensure high availability and optimal
|
||||
// resource utilization across multiple backend clients. It handles request translation
|
||||
// between Claude API format and the underlying Gemini backend, providing seamless
|
||||
// API compatibility while maintaining robust error handling and connection management.
|
||||
package claude
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/translator/claude/code"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
)
|
||||
|
||||
// ClaudeCodeAPIHandlers contains the handlers for Claude API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type ClaudeCodeAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewClaudeCodeAPIHandlers creates a new Claude API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a ClaudeCodeAPIHandlers.
|
||||
func NewClaudeCodeAPIHandlers(apiHandlers *handlers.APIHandlers) *ClaudeCodeAPIHandlers {
|
||||
return &ClaudeCodeAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// ClaudeMessages handles Claude-compatible streaming chat completions.
|
||||
// This function implements a sophisticated client rotation and quota management system
|
||||
// to ensure high availability and optimal resource utilization across multiple backend clients.
|
||||
func (h *ClaudeCodeAPIHandlers) ClaudeMessages(c *gin.Context) {
|
||||
// Extract raw JSON data from the incoming request
|
||||
rawJSON, err := c.GetRawData()
|
||||
// If data retrieval fails, return a 400 Bad Request error.
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Set up Server-Sent Events (SSE) headers for streaming response
|
||||
// These headers are essential for maintaining a persistent connection
|
||||
// and enabling real-time streaming of chat completions
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
// This is crucial for streaming as it allows immediate sending of data chunks
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Parse and prepare the Claude request, extracting model name, system instructions,
|
||||
// conversation contents, and available tools from the raw JSON
|
||||
modelName, systemInstruction, contents, tools := code.PrepareClaudeRequest(rawJSON)
|
||||
|
||||
// Map Claude model names to corresponding Gemini models
|
||||
// This allows the proxy to handle Claude API calls using Gemini backend
|
||||
if modelName == "claude-sonnet-4-20250514" {
|
||||
modelName = "gemini-2.5-pro"
|
||||
} else if modelName == "claude-3-5-haiku-20241022" {
|
||||
modelName = "gemini-2.5-flash"
|
||||
}
|
||||
|
||||
// Create a cancellable context for the backend client request
|
||||
// This allows proper cleanup and cancellation of ongoing requests
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
// This prevents deadlocks and ensures proper resource cleanup
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
// Main client rotation loop with quota management
|
||||
// This loop implements a sophisticated load balancing and failover mechanism
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
// Determine the authentication method being used by the selected client
|
||||
// This affects how responses are formatted and logged
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
// Initiate streaming communication with the backend client
|
||||
// This returns two channels: one for response chunks and one for errors
|
||||
|
||||
includeThoughts := false
|
||||
if userAgent, hasKey := c.Request.Header["User-Agent"]; hasKey {
|
||||
includeThoughts = !strings.Contains(userAgent[0], "claude-cli")
|
||||
}
|
||||
|
||||
respChan, errChan := cliClient.SendMessageStream(cliCtx, rawJSON, modelName, systemInstruction, contents, tools, includeThoughts)
|
||||
|
||||
// Track response state for proper Claude format conversion
|
||||
hasFirstResponse := false
|
||||
responseType := 0
|
||||
responseIndex := 0
|
||||
|
||||
// Main streaming loop - handles multiple concurrent events using Go channels
|
||||
// This select statement manages four different types of events simultaneously
|
||||
for {
|
||||
select {
|
||||
// Case 1: Handle client disconnection
|
||||
// Detects when the HTTP client has disconnected and cleans up resources
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("Client disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request to prevent resource leaks
|
||||
return
|
||||
}
|
||||
|
||||
// Case 2: Process incoming response chunks from the backend
|
||||
// This handles the actual streaming data from the AI model
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
// Stream has ended - send the final message_stop event
|
||||
// This follows the Claude API specification for stream termination
|
||||
_, _ = c.Writer.Write([]byte(`event: message_stop`))
|
||||
_, _ = c.Writer.Write([]byte("\n"))
|
||||
_, _ = c.Writer.Write([]byte(`data: {"type":"message_stop"}`))
|
||||
_, _ = c.Writer.Write([]byte("\n\n\n"))
|
||||
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// Convert the backend response to Claude-compatible format
|
||||
// This translation layer ensures API compatibility
|
||||
claudeFormat := code.ConvertCliToClaude(chunk, isGlAPIKey, hasFirstResponse, &responseType, &responseIndex)
|
||||
if claudeFormat != "" {
|
||||
_, _ = c.Writer.Write([]byte(claudeFormat))
|
||||
flusher.Flush() // Immediately send the chunk to the client
|
||||
}
|
||||
hasFirstResponse = true
|
||||
|
||||
// Case 3: Handle errors from the backend
|
||||
// This manages various error conditions and implements retry logic
|
||||
case errInfo, okError := <-errChan:
|
||||
if okError {
|
||||
// Special handling for quota exceeded errors
|
||||
// If configured, attempt to switch to a different project/client
|
||||
if errInfo.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop // Restart the client selection process
|
||||
} else {
|
||||
// Forward other errors directly to the client
|
||||
c.Status(errInfo.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errInfo.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Case 4: Send periodic keep-alive signals
|
||||
// Prevents connection timeouts during long-running requests
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
// Send a ping event to maintain the connection
|
||||
// This is especially important for slow AI model responses
|
||||
output := "event: ping\n"
|
||||
output = output + `data: {"type": "ping"}`
|
||||
output = output + "\n\n\n"
|
||||
_, _ = c.Writer.Write([]byte(output))
|
||||
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
382
internal/api/handlers/claude/code_handlers.go
Normal file
382
internal/api/handlers/claude/code_handlers.go
Normal file
@@ -0,0 +1,382 @@
|
||||
// Package claude provides HTTP handlers for Claude API code-related functionality.
|
||||
// This package implements Claude-compatible streaming chat completions with sophisticated
|
||||
// client rotation and quota management systems to ensure high availability and optimal
|
||||
// resource utilization across multiple backend clients. It handles request translation
|
||||
// between Claude API format and the underlying Gemini backend, providing seamless
|
||||
// API compatibility while maintaining robust error handling and connection management.
|
||||
package claude
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
translatorClaudeCodeToCodex "github.com/luispater/CLIProxyAPI/internal/translator/codex/claude/code"
|
||||
translatorClaudeCodeToGeminiCli "github.com/luispater/CLIProxyAPI/internal/translator/gemini-cli/claude/code"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
)
|
||||
|
||||
// ClaudeCodeAPIHandlers contains the handlers for Claude API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type ClaudeCodeAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewClaudeCodeAPIHandlers creates a new Claude API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a ClaudeCodeAPIHandlers.
|
||||
func NewClaudeCodeAPIHandlers(apiHandlers *handlers.APIHandlers) *ClaudeCodeAPIHandlers {
|
||||
return &ClaudeCodeAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// ClaudeMessages handles Claude-compatible streaming chat completions.
|
||||
// This function implements a sophisticated client rotation and quota management system
|
||||
// to ensure high availability and optimal resource utilization across multiple backend clients.
|
||||
func (h *ClaudeCodeAPIHandlers) ClaudeMessages(c *gin.Context) {
|
||||
// Extract raw JSON data from the incoming request
|
||||
rawJSON, err := c.GetRawData()
|
||||
// If data retrieval fails, return a 400 Bad Request error.
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// h.handleGeminiStreamingResponse(c, rawJSON)
|
||||
// h.handleCodexStreamingResponse(c, rawJSON)
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
provider := util.GetProviderName(modelName.String())
|
||||
if provider == "gemini" {
|
||||
h.handleGeminiStreamingResponse(c, rawJSON)
|
||||
} else if provider == "gpt" {
|
||||
h.handleCodexStreamingResponse(c, rawJSON)
|
||||
} else {
|
||||
h.handleGeminiStreamingResponse(c, rawJSON)
|
||||
}
|
||||
}
|
||||
|
||||
// handleGeminiStreamingResponse streams Claude-compatible responses backed by Gemini.
|
||||
// It sets up SSE, selects a backend client with rotation/quota logic,
|
||||
// forwards chunks, and translates them to Claude CLI format.
|
||||
func (h *ClaudeCodeAPIHandlers) handleGeminiStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
// Set up Server-Sent Events (SSE) headers for streaming response
|
||||
// These headers are essential for maintaining a persistent connection
|
||||
// and enabling real-time streaming of chat completions
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
// This is crucial for streaming as it allows immediate sending of data chunks
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Parse and prepare the Claude request, extracting model name, system instructions,
|
||||
// conversation contents, and available tools from the raw JSON
|
||||
modelName, systemInstruction, contents, tools := translatorClaudeCodeToGeminiCli.ConvertClaudeCodeRequestToCli(rawJSON)
|
||||
|
||||
// Map Claude model names to corresponding Gemini models
|
||||
// This allows the proxy to handle Claude API calls using Gemini backend
|
||||
if modelName == "claude-sonnet-4-20250514" {
|
||||
modelName = "gemini-2.5-pro"
|
||||
} else if modelName == "claude-3-5-haiku-20241022" {
|
||||
modelName = "gemini-2.5-flash"
|
||||
}
|
||||
|
||||
// Create a cancellable context for the backend client request
|
||||
// This allows proper cleanup and cancellation of ongoing requests
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
cliClient = client.NewGeminiClient(nil, nil, nil)
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
// This prevents deadlocks and ensures proper resource cleanup
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
// Main client rotation loop with quota management
|
||||
// This loop implements a sophisticated load balancing and failover mechanism
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
// Determine the authentication method being used by the selected client
|
||||
// This affects how responses are formatted and logged
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use gemini generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request use gemini account: %s, project id: %s", cliClient.GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
// Initiate streaming communication with the backend client
|
||||
// This returns two channels: one for response chunks and one for errors
|
||||
|
||||
includeThoughts := false
|
||||
if userAgent, hasKey := c.Request.Header["User-Agent"]; hasKey {
|
||||
includeThoughts = !strings.Contains(userAgent[0], "claude-cli")
|
||||
}
|
||||
|
||||
respChan, errChan := cliClient.SendMessageStream(cliCtx, rawJSON, modelName, systemInstruction, contents, tools, includeThoughts)
|
||||
|
||||
// Track response state for proper Claude format conversion
|
||||
hasFirstResponse := false
|
||||
responseType := 0
|
||||
responseIndex := 0
|
||||
|
||||
// Main streaming loop - handles multiple concurrent events using Go channels
|
||||
// This select statement manages four different types of events simultaneously
|
||||
for {
|
||||
select {
|
||||
// Case 1: Handle client disconnection
|
||||
// Detects when the HTTP client has disconnected and cleans up resources
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request to prevent resource leaks
|
||||
return
|
||||
}
|
||||
|
||||
// Case 2: Process incoming response chunks from the backend
|
||||
// This handles the actual streaming data from the AI model
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
// Stream has ended - send the final message_stop event
|
||||
// This follows the Claude API specification for stream termination
|
||||
_, _ = c.Writer.Write([]byte(`event: message_stop`))
|
||||
_, _ = c.Writer.Write([]byte("\n"))
|
||||
_, _ = c.Writer.Write([]byte(`data: {"type":"message_stop"}`))
|
||||
_, _ = c.Writer.Write([]byte("\n\n\n"))
|
||||
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// Convert the backend response to Claude-compatible format
|
||||
// This translation layer ensures API compatibility
|
||||
claudeFormat := translatorClaudeCodeToGeminiCli.ConvertCliResponseToClaudeCode(chunk, isGlAPIKey, hasFirstResponse, &responseType, &responseIndex)
|
||||
if claudeFormat != "" {
|
||||
_, _ = c.Writer.Write([]byte(claudeFormat))
|
||||
flusher.Flush() // Immediately send the chunk to the client
|
||||
}
|
||||
hasFirstResponse = true
|
||||
|
||||
// Case 3: Handle errors from the backend
|
||||
// This manages various error conditions and implements retry logic
|
||||
case errInfo, okError := <-errChan:
|
||||
if okError {
|
||||
// Special handling for quota exceeded errors
|
||||
// If configured, attempt to switch to a different project/client
|
||||
if errInfo.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop // Restart the client selection process
|
||||
} else {
|
||||
// Forward other errors directly to the client
|
||||
c.Status(errInfo.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errInfo.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Case 4: Send periodic keep-alive signals
|
||||
// Prevents connection timeouts during long-running requests
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
// Send a ping event to maintain the connection
|
||||
// This is especially important for slow AI model responses
|
||||
output := "event: ping\n"
|
||||
output = output + `data: {"type": "ping"}`
|
||||
output = output + "\n\n\n"
|
||||
_, _ = c.Writer.Write([]byte(output))
|
||||
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleCodexStreamingResponse streams Claude-compatible responses backed by OpenAI.
|
||||
// It converts the Claude request into Codex/OpenAI responses format, establishes SSE,
|
||||
// and translates streaming chunks back into Claude CLI events.
|
||||
func (h *ClaudeCodeAPIHandlers) handleCodexStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
// Set up Server-Sent Events (SSE) headers for streaming response
|
||||
// These headers are essential for maintaining a persistent connection
|
||||
// and enabling real-time streaming of chat completions
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
// This is crucial for streaming as it allows immediate sending of data chunks
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Parse and prepare the Claude request, extracting model name, system instructions,
|
||||
// conversation contents, and available tools from the raw JSON
|
||||
newRequestJSON := translatorClaudeCodeToCodex.ConvertClaudeCodeRequestToCodex(rawJSON)
|
||||
modelName := gjson.GetBytes(rawJSON, "model").String()
|
||||
// Map Claude model names to corresponding Gemini models
|
||||
// This allows the proxy to handle Claude API calls using Gemini backend
|
||||
if modelName == "claude-sonnet-4-20250514" {
|
||||
modelName = "gpt-5"
|
||||
} else if modelName == "claude-3-5-haiku-20241022" {
|
||||
modelName = "gpt-5"
|
||||
}
|
||||
newRequestJSON, _ = sjson.Set(newRequestJSON, "model", modelName)
|
||||
// log.Debugf(string(rawJSON))
|
||||
// log.Debugf(newRequestJSON)
|
||||
// return
|
||||
// Create a cancellable context for the backend client request
|
||||
// This allows proper cleanup and cancellation of ongoing requests
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
// This prevents deadlocks and ensures proper resource cleanup
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
// Main client rotation loop with quota management
|
||||
// This loop implements a sophisticated load balancing and failover mechanism
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request use codex account: %s", cliClient.GetEmail())
|
||||
|
||||
// Initiate streaming communication with the backend client
|
||||
// This returns two channels: one for response chunks and one for errors
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
|
||||
// Track response state for proper Claude format conversion
|
||||
hasFirstResponse := false
|
||||
hasToolCall := false
|
||||
|
||||
// Main streaming loop - handles multiple concurrent events using Go channels
|
||||
// This select statement manages four different types of events simultaneously
|
||||
for {
|
||||
select {
|
||||
// Case 1: Handle client disconnection
|
||||
// Detects when the HTTP client has disconnected and cleans up resources
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request to prevent resource leaks
|
||||
return
|
||||
}
|
||||
|
||||
// Case 2: Process incoming response chunks from the backend
|
||||
// This handles the actual streaming data from the AI model
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// Convert the backend response to Claude-compatible format
|
||||
// This translation layer ensures API compatibility
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
var claudeFormat string
|
||||
claudeFormat, hasToolCall = translatorClaudeCodeToCodex.ConvertCodexResponseToClaude(jsonData, hasToolCall)
|
||||
// log.Debugf("claudeFormat: %s", claudeFormat)
|
||||
if claudeFormat != "" {
|
||||
_, _ = c.Writer.Write([]byte(claudeFormat))
|
||||
_, _ = c.Writer.Write([]byte("\n"))
|
||||
}
|
||||
flusher.Flush() // Immediately send the chunk to the client
|
||||
hasFirstResponse = true
|
||||
} else {
|
||||
// log.Debugf("chunk: %s", string(chunk))
|
||||
}
|
||||
// Case 3: Handle errors from the backend
|
||||
// This manages various error conditions and implements retry logic
|
||||
case errInfo, okError := <-errChan:
|
||||
if okError {
|
||||
// log.Debugf("Code: %d, Error: %v", errInfo.StatusCode, errInfo.Error)
|
||||
// Special handling for quota exceeded errors
|
||||
// If configured, attempt to switch to a different project/client
|
||||
if errInfo.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
log.Debugf("quota exceeded, switch client")
|
||||
continue outLoop // Restart the client selection process
|
||||
} else {
|
||||
// Forward other errors directly to the client
|
||||
c.Status(errInfo.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errInfo.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Case 4: Send periodic keep-alive signals
|
||||
// Prevents connection timeouts during long-running requests
|
||||
case <-time.After(3000 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
// Send a ping event to maintain the connection
|
||||
// This is especially important for slow AI model responses
|
||||
output := "event: ping\n"
|
||||
output = output + `data: {"type": "ping"}`
|
||||
output = output + "\n\n"
|
||||
_, _ = c.Writer.Write([]byte(output))
|
||||
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -1,268 +0,0 @@
|
||||
// Package cli provides HTTP handlers for Gemini CLI API functionality.
|
||||
// This package implements handlers that process CLI-specific requests for Gemini API operations,
|
||||
// including content generation and streaming content generation endpoints.
|
||||
// The handlers restrict access to localhost only and manage communication with the backend service.
|
||||
package cli
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
"io"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
)
|
||||
|
||||
// GeminiCLIAPIHandlers contains the handlers for Gemini CLI API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type GeminiCLIAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewGeminiCLIAPIHandlers creates a new Gemini CLI API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a GeminiCLIAPIHandlers.
|
||||
func NewGeminiCLIAPIHandlers(apiHandlers *handlers.APIHandlers) *GeminiCLIAPIHandlers {
|
||||
return &GeminiCLIAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// CLIHandler handles CLI-specific requests for Gemini API operations.
|
||||
// It restricts access to localhost only and routes requests to appropriate internal handlers.
|
||||
func (h *GeminiCLIAPIHandlers) CLIHandler(c *gin.Context) {
|
||||
if !strings.HasPrefix(c.Request.RemoteAddr, "127.0.0.1:") {
|
||||
c.JSON(http.StatusForbidden, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "CLI reply only allow local access",
|
||||
Type: "forbidden",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
rawJSON, _ := c.GetRawData()
|
||||
requestRawURI := c.Request.URL.Path
|
||||
if requestRawURI == "/v1internal:generateContent" {
|
||||
h.internalGenerateContent(c, rawJSON)
|
||||
} else if requestRawURI == "/v1internal:streamGenerateContent" {
|
||||
h.internalStreamGenerateContent(c, rawJSON)
|
||||
} else {
|
||||
reqBody := bytes.NewBuffer(rawJSON)
|
||||
req, err := http.NewRequest("POST", fmt.Sprintf("https://cloudcode-pa.googleapis.com%s", c.Request.URL.RequestURI()), reqBody)
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
for key, value := range c.Request.Header {
|
||||
req.Header[key] = value
|
||||
}
|
||||
|
||||
httpClient, err := util.SetProxy(h.Cfg, &http.Client{})
|
||||
if err != nil {
|
||||
log.Fatalf("set proxy failed: %v", err)
|
||||
}
|
||||
|
||||
resp, err := httpClient.Do(req)
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
defer func() {
|
||||
if err = resp.Body.Close(); err != nil {
|
||||
log.Printf("warn: failed to close response body: %v", err)
|
||||
}
|
||||
}()
|
||||
bodyBytes, _ := io.ReadAll(resp.Body)
|
||||
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: string(bodyBytes),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
defer func() {
|
||||
_ = resp.Body.Close()
|
||||
}()
|
||||
|
||||
for key, value := range resp.Header {
|
||||
c.Header(key, value[0])
|
||||
}
|
||||
output, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
log.Errorf("Failed to read response body: %v", err)
|
||||
return
|
||||
}
|
||||
_, _ = c.Writer.Write(output)
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) internalStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
if alt == "" {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
}
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, rawJSON, "")
|
||||
hasFirstResponse := false
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("Client disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
hasFirstResponse = true
|
||||
if cliClient.GetGenerativeLanguageAPIKey() != "" {
|
||||
chunk, _ = sjson.SetRawBytes(chunk, "response", chunk)
|
||||
}
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
_, _ = c.Writer.Write([]byte("\n"))
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) internalGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendRawMessage(cliCtx, rawJSON, "")
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
491
internal/api/handlers/gemini/cli/cli_handlers.go
Normal file
491
internal/api/handlers/gemini/cli/cli_handlers.go
Normal file
@@ -0,0 +1,491 @@
|
||||
// Package cli provides HTTP handlers for Gemini CLI API functionality.
|
||||
// This package implements handlers that process CLI-specific requests for Gemini API operations,
|
||||
// including content generation and streaming content generation endpoints.
|
||||
// The handlers restrict access to localhost only and manage communication with the backend service.
|
||||
package cli
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"io"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
translatorGeminiToCodex "github.com/luispater/CLIProxyAPI/internal/translator/codex/gemini"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
)
|
||||
|
||||
// GeminiCLIAPIHandlers contains the handlers for Gemini CLI API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type GeminiCLIAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewGeminiCLIAPIHandlers creates a new Gemini CLI API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a GeminiCLIAPIHandlers.
|
||||
func NewGeminiCLIAPIHandlers(apiHandlers *handlers.APIHandlers) *GeminiCLIAPIHandlers {
|
||||
return &GeminiCLIAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// CLIHandler handles CLI-specific requests for Gemini API operations.
|
||||
// It restricts access to localhost only and routes requests to appropriate internal handlers.
|
||||
func (h *GeminiCLIAPIHandlers) CLIHandler(c *gin.Context) {
|
||||
if !strings.HasPrefix(c.Request.RemoteAddr, "127.0.0.1:") {
|
||||
c.JSON(http.StatusForbidden, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "CLI reply only allow local access",
|
||||
Type: "forbidden",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
rawJSON, _ := c.GetRawData()
|
||||
requestRawURI := c.Request.URL.Path
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
provider := util.GetProviderName(modelName.String())
|
||||
|
||||
if requestRawURI == "/v1internal:generateContent" {
|
||||
if provider == "gemini" || provider == "unknow" {
|
||||
h.handleInternalGenerateContent(c, rawJSON)
|
||||
} else if provider == "gpt" {
|
||||
h.handleCodexInternalGenerateContent(c, rawJSON)
|
||||
}
|
||||
} else if requestRawURI == "/v1internal:streamGenerateContent" {
|
||||
if provider == "gemini" || provider == "unknow" {
|
||||
h.handleInternalStreamGenerateContent(c, rawJSON)
|
||||
} else if provider == "gpt" {
|
||||
h.handleCodexInternalStreamGenerateContent(c, rawJSON)
|
||||
}
|
||||
} else {
|
||||
reqBody := bytes.NewBuffer(rawJSON)
|
||||
req, err := http.NewRequest("POST", fmt.Sprintf("https://cloudcode-pa.googleapis.com%s", c.Request.URL.RequestURI()), reqBody)
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
for key, value := range c.Request.Header {
|
||||
req.Header[key] = value
|
||||
}
|
||||
|
||||
httpClient := util.SetProxy(h.Cfg, &http.Client{})
|
||||
|
||||
resp, err := httpClient.Do(req)
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
defer func() {
|
||||
if err = resp.Body.Close(); err != nil {
|
||||
log.Printf("warn: failed to close response body: %v", err)
|
||||
}
|
||||
}()
|
||||
bodyBytes, _ := io.ReadAll(resp.Body)
|
||||
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: string(bodyBytes),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
defer func() {
|
||||
_ = resp.Body.Close()
|
||||
}()
|
||||
|
||||
for key, value := range resp.Header {
|
||||
c.Header(key, value[0])
|
||||
}
|
||||
output, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
log.Errorf("Failed to read response body: %v", err)
|
||||
return
|
||||
}
|
||||
_, _ = c.Writer.Write(output)
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) handleInternalStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
if alt == "" {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
}
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, rawJSON, "")
|
||||
hasFirstResponse := false
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
hasFirstResponse = true
|
||||
if cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey() != "" {
|
||||
chunk, _ = sjson.SetRawBytes(chunk, "response", chunk)
|
||||
}
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
_, _ = c.Writer.Write([]byte("\n"))
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) handleInternalGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
// log.Debugf("GenerateContent: %s", string(rawJSON))
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendRawMessage(cliCtx, rawJSON, "")
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
log.Debugf("code: %d, error: %s", err.StatusCode, err.Error.Error())
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) handleCodexInternalStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
rawJSON = []byte(gjson.GetBytes(rawJSON, "request").Raw)
|
||||
rawJSON, _ = sjson.SetBytes(rawJSON, "model", modelResult.String())
|
||||
rawJSON, _ = sjson.SetRawBytes(rawJSON, "system_instruction", []byte(gjson.GetBytes(rawJSON, "systemInstruction").Raw))
|
||||
rawJSON, _ = sjson.DeleteBytes(rawJSON, "systemInstruction")
|
||||
|
||||
// log.Debugf("Request: %s", string(rawJSON))
|
||||
// return
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
newRequestJSON := translatorGeminiToCodex.ConvertGeminiRequestToCodex(rawJSON)
|
||||
// log.Debugf("Request: %s", newRequestJSON)
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
|
||||
params := &translatorGeminiToCodex.ConvertCodexResponseToGeminiParams{
|
||||
Model: modelName.String(),
|
||||
CreatedAt: 0,
|
||||
ResponseID: "",
|
||||
LastStorageOutput: "",
|
||||
}
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// _, _ = logFile.Write(chunk)
|
||||
// _, _ = logFile.Write([]byte("\n"))
|
||||
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() != "" {
|
||||
outputs := translatorGeminiToCodex.ConvertCodexResponseToGemini(jsonData, params)
|
||||
if len(outputs) > 0 {
|
||||
for i := 0; i < len(outputs); i++ {
|
||||
outputs[i], _ = sjson.SetRaw("{}", "response", outputs[i])
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write([]byte(outputs[i]))
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case errMessage, okError := <-errChan:
|
||||
if okError {
|
||||
if errMessage.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
log.Debugf("code: %d, error: %s", errMessage.StatusCode, errMessage.Error.Error())
|
||||
c.Status(errMessage.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errMessage.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiCLIAPIHandlers) handleCodexInternalGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
orgRawJSON := rawJSON
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
rawJSON = []byte(gjson.GetBytes(rawJSON, "request").Raw)
|
||||
rawJSON, _ = sjson.SetBytes(rawJSON, "model", modelResult.String())
|
||||
rawJSON, _ = sjson.SetRawBytes(rawJSON, "system_instruction", []byte(gjson.GetBytes(rawJSON, "systemInstruction").Raw))
|
||||
rawJSON, _ = sjson.DeleteBytes(rawJSON, "systemInstruction")
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
newRequestJSON := translatorGeminiToCodex.ConvertGeminiRequestToCodex(rawJSON)
|
||||
// log.Debugf("Request: %s", newRequestJSON)
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() != "" {
|
||||
var geminiStr string
|
||||
geminiStr = translatorGeminiToCodex.ConvertCodexResponseToGeminiNonStream(jsonData, modelName.String())
|
||||
if geminiStr != "" {
|
||||
_, _ = c.Writer.Write([]byte(geminiStr))
|
||||
}
|
||||
}
|
||||
}
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
log.Debugf("org: %s", string(orgRawJSON))
|
||||
log.Debugf("raw: %s", string(rawJSON))
|
||||
log.Debugf("newRequestJSON: %s", newRequestJSON)
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -1,437 +0,0 @@
|
||||
// Package gemini provides HTTP handlers for Gemini API endpoints.
|
||||
// This package implements handlers for managing Gemini model operations including
|
||||
// model listing, content generation, streaming content generation, and token counting.
|
||||
// It serves as a proxy layer between clients and the Gemini backend service,
|
||||
// handling request translation, client management, and response processing.
|
||||
package gemini
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/translator/gemini/cli"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
)
|
||||
|
||||
// GeminiAPIHandlers contains the handlers for Gemini API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type GeminiAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewGeminiAPIHandlers creates a new Gemini API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a GeminiAPIHandlers.
|
||||
func NewGeminiAPIHandlers(apiHandlers *handlers.APIHandlers) *GeminiAPIHandlers {
|
||||
return &GeminiAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// GeminiModels handles the Gemini models listing endpoint.
|
||||
// It returns a JSON response containing available Gemini models and their specifications.
|
||||
func (h *GeminiAPIHandlers) GeminiModels(c *gin.Context) {
|
||||
c.Status(http.StatusOK)
|
||||
c.Header("Content-Type", "application/json; charset=UTF-8")
|
||||
_, _ = c.Writer.Write([]byte(`{"models":[{"name":"models/gemini-2.5-flash","version":"001","displayName":"Gemini `))
|
||||
_, _ = c.Writer.Write([]byte(`2.5 Flash","description":"Stable version of Gemini 2.5 Flash, our mid-size multimod`))
|
||||
_, _ = c.Writer.Write([]byte(`al model that supports up to 1 million tokens, released in June of 2025.","inputTok`))
|
||||
_, _ = c.Writer.Write([]byte(`enLimit":1048576,"outputTokenLimit":65536,"supportedGenerationMethods":["generateCo`))
|
||||
_, _ = c.Writer.Write([]byte(`ntent","countTokens","createCachedContent","batchGenerateContent"],"temperature":1,`))
|
||||
_, _ = c.Writer.Write([]byte(`"topP":0.95,"topK":64,"maxTemperature":2,"thinking":true},{"name":"models/gemini-2.`))
|
||||
_, _ = c.Writer.Write([]byte(`5-pro","version":"2.5","displayName":"Gemini 2.5 Pro","description":"Stable release`))
|
||||
_, _ = c.Writer.Write([]byte(` (June 17th, 2025) of Gemini 2.5 Pro","inputTokenLimit":1048576,"outputTokenLimit":`))
|
||||
_, _ = c.Writer.Write([]byte(`65536,"supportedGenerationMethods":["generateContent","countTokens","createCachedCo`))
|
||||
_, _ = c.Writer.Write([]byte(`ntent","batchGenerateContent"],"temperature":1,"topP":0.95,"topK":64,"maxTemperatur`))
|
||||
_, _ = c.Writer.Write([]byte(`e":2,"thinking":true}],"nextPageToken":""}`))
|
||||
}
|
||||
|
||||
// GeminiGetHandler handles GET requests for specific Gemini model information.
|
||||
// It returns detailed information about a specific Gemini model based on the action parameter.
|
||||
func (h *GeminiAPIHandlers) GeminiGetHandler(c *gin.Context) {
|
||||
var request struct {
|
||||
Action string `uri:"action" binding:"required"`
|
||||
}
|
||||
if err := c.ShouldBindUri(&request); err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
if request.Action == "gemini-2.5-pro" {
|
||||
c.Status(http.StatusOK)
|
||||
c.Header("Content-Type", "application/json; charset=UTF-8")
|
||||
_, _ = c.Writer.Write([]byte(`{"name":"models/gemini-2.5-pro","version":"2.5","displayName":"Gemini 2.5 Pro",`))
|
||||
_, _ = c.Writer.Write([]byte(`"description":"Stable release (June 17th, 2025) of Gemini 2.5 Pro","inputTokenL`))
|
||||
_, _ = c.Writer.Write([]byte(`imit":1048576,"outputTokenLimit":65536,"supportedGenerationMethods":["generateC`))
|
||||
_, _ = c.Writer.Write([]byte(`ontent","countTokens","createCachedContent","batchGenerateContent"],"temperatur`))
|
||||
_, _ = c.Writer.Write([]byte(`e":1,"topP":0.95,"topK":64,"maxTemperature":2,"thinking":true}`))
|
||||
} else if request.Action == "gemini-2.5-flash" {
|
||||
c.Status(http.StatusOK)
|
||||
c.Header("Content-Type", "application/json; charset=UTF-8")
|
||||
_, _ = c.Writer.Write([]byte(`{"name":"models/gemini-2.5-flash","version":"001","displayName":"Gemini 2.5 Fla`))
|
||||
_, _ = c.Writer.Write([]byte(`sh","description":"Stable version of Gemini 2.5 Flash, our mid-size multimodal `))
|
||||
_, _ = c.Writer.Write([]byte(`model that supports up to 1 million tokens, released in June of 2025.","inputTo`))
|
||||
_, _ = c.Writer.Write([]byte(`kenLimit":1048576,"outputTokenLimit":65536,"supportedGenerationMethods":["gener`))
|
||||
_, _ = c.Writer.Write([]byte(`ateContent","countTokens","createCachedContent","batchGenerateContent"],"temper`))
|
||||
_, _ = c.Writer.Write([]byte(`ature":1,"topP":0.95,"topK":64,"maxTemperature":2,"thinking":true}`))
|
||||
} else {
|
||||
c.Status(http.StatusNotFound)
|
||||
_, _ = c.Writer.Write([]byte(
|
||||
`{"error":{"message":"Not Found","code":404,"status":"NOT_FOUND"}}`,
|
||||
))
|
||||
}
|
||||
}
|
||||
|
||||
// GeminiHandler handles POST requests for Gemini API operations.
|
||||
// It routes requests to appropriate handlers based on the action parameter (model:method format).
|
||||
func (h *GeminiAPIHandlers) GeminiHandler(c *gin.Context) {
|
||||
var request struct {
|
||||
Action string `uri:"action" binding:"required"`
|
||||
}
|
||||
if err := c.ShouldBindUri(&request); err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
action := strings.Split(request.Action, ":")
|
||||
if len(action) != 2 {
|
||||
c.JSON(http.StatusNotFound, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("%s not found.", c.Request.URL.Path),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelName := action[0]
|
||||
method := action[1]
|
||||
rawJSON, _ := c.GetRawData()
|
||||
rawJSON, _ = sjson.SetBytes(rawJSON, "model", []byte(modelName))
|
||||
|
||||
if method == "generateContent" {
|
||||
h.geminiGenerateContent(c, rawJSON)
|
||||
} else if method == "streamGenerateContent" {
|
||||
h.geminiStreamGenerateContent(c, rawJSON)
|
||||
} else if method == "countTokens" {
|
||||
h.geminiCountTokens(c, rawJSON)
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) geminiStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
if alt == "" {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
}
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
template := ""
|
||||
parsed := gjson.Parse(string(rawJSON))
|
||||
contents := parsed.Get("request.contents")
|
||||
if contents.Exists() {
|
||||
template = string(rawJSON)
|
||||
} else {
|
||||
template = `{"project":"","request":{},"model":""}`
|
||||
template, _ = sjson.SetRaw(template, "request", string(rawJSON))
|
||||
template, _ = sjson.Set(template, "model", gjson.Get(template, "request.model").String())
|
||||
template, _ = sjson.Delete(template, "request.model")
|
||||
}
|
||||
|
||||
template, errFixCLIToolResponse := cli.FixCLIToolResponse(template)
|
||||
if errFixCLIToolResponse != nil {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: errFixCLIToolResponse.Error(),
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
systemInstructionResult := gjson.Get(template, "request.system_instruction")
|
||||
if systemInstructionResult.Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.systemInstruction", systemInstructionResult.Raw)
|
||||
template, _ = sjson.Delete(template, "request.system_instruction")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, rawJSON, alt)
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("Client disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
if cliClient.GetGenerativeLanguageAPIKey() == "" {
|
||||
if alt == "" {
|
||||
responseResult := gjson.GetBytes(chunk, "response")
|
||||
if responseResult.Exists() {
|
||||
chunk = []byte(responseResult.Raw)
|
||||
}
|
||||
} else {
|
||||
chunkTemplate := "[]"
|
||||
responseResult := gjson.ParseBytes(chunk)
|
||||
if responseResult.IsArray() {
|
||||
responseResultItems := responseResult.Array()
|
||||
for i := 0; i < len(responseResultItems); i++ {
|
||||
responseResultItem := responseResultItems[i]
|
||||
if responseResultItem.Get("response").Exists() {
|
||||
chunkTemplate, _ = sjson.SetRaw(chunkTemplate, "-1", responseResultItem.Get("response").Raw)
|
||||
}
|
||||
}
|
||||
}
|
||||
chunk = []byte(chunkTemplate)
|
||||
}
|
||||
}
|
||||
if alt == "" {
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
} else {
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
}
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
log.Debugf("quota exceeded, switch client")
|
||||
continue outLoop
|
||||
} else {
|
||||
log.Debugf("error code :%d, error: %v", err.StatusCode, err.Error.Error())
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) geminiCountTokens(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
alt := h.GetAlt(c)
|
||||
// orgrawJSON := rawJSON
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName, false)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
|
||||
template := `{"request":{}}`
|
||||
if gjson.GetBytes(rawJSON, "generateContentRequest").Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request", gjson.GetBytes(rawJSON, "generateContentRequest").Raw)
|
||||
template, _ = sjson.Delete(template, "generateContentRequest")
|
||||
} else if gjson.GetBytes(rawJSON, "contents").Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.contents", gjson.GetBytes(rawJSON, "contents").Raw)
|
||||
template, _ = sjson.Delete(template, "contents")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendRawTokenCount(cliCtx, rawJSON, alt)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
// log.Debugf(err.Error.Error())
|
||||
// log.Debugf(string(rawJSON))
|
||||
// log.Debugf(string(orgrawJSON))
|
||||
}
|
||||
break
|
||||
} else {
|
||||
if cliClient.GetGenerativeLanguageAPIKey() == "" {
|
||||
responseResult := gjson.GetBytes(resp, "response")
|
||||
if responseResult.Exists() {
|
||||
resp = []byte(responseResult.Raw)
|
||||
}
|
||||
}
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) geminiGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
template := ""
|
||||
parsed := gjson.Parse(string(rawJSON))
|
||||
contents := parsed.Get("request.contents")
|
||||
if contents.Exists() {
|
||||
template = string(rawJSON)
|
||||
} else {
|
||||
template = `{"project":"","request":{},"model":""}`
|
||||
template, _ = sjson.SetRaw(template, "request", string(rawJSON))
|
||||
template, _ = sjson.Set(template, "model", gjson.Get(template, "request.model").String())
|
||||
template, _ = sjson.Delete(template, "request.model")
|
||||
}
|
||||
|
||||
template, errFixCLIToolResponse := cli.FixCLIToolResponse(template)
|
||||
if errFixCLIToolResponse != nil {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: errFixCLIToolResponse.Error(),
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
systemInstructionResult := gjson.Get(template, "request.system_instruction")
|
||||
if systemInstructionResult.Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.systemInstruction", systemInstructionResult.Raw)
|
||||
template, _ = sjson.Delete(template, "request.system_instruction")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
resp, err := cliClient.SendRawMessage(cliCtx, rawJSON, alt)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
if cliClient.GetGenerativeLanguageAPIKey() == "" {
|
||||
responseResult := gjson.GetBytes(resp, "response")
|
||||
if responseResult.Exists() {
|
||||
resp = []byte(responseResult.Raw)
|
||||
}
|
||||
}
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
735
internal/api/handlers/gemini/gemini_handlers.go
Normal file
735
internal/api/handlers/gemini/gemini_handlers.go
Normal file
@@ -0,0 +1,735 @@
|
||||
// Package gemini provides HTTP handlers for Gemini API endpoints.
|
||||
// This package implements handlers for managing Gemini model operations including
|
||||
// model listing, content generation, streaming content generation, and token counting.
|
||||
// It serves as a proxy layer between clients and the Gemini backend service,
|
||||
// handling request translation, client management, and response processing.
|
||||
package gemini
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"net/http"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
translatorGeminiToCodex "github.com/luispater/CLIProxyAPI/internal/translator/codex/gemini"
|
||||
translatorGeminiToGeminiCli "github.com/luispater/CLIProxyAPI/internal/translator/gemini-cli/gemini/cli"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/tidwall/sjson"
|
||||
)
|
||||
|
||||
// GeminiAPIHandlers contains the handlers for Gemini API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type GeminiAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewGeminiAPIHandlers creates a new Gemini API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns a GeminiAPIHandlers.
|
||||
func NewGeminiAPIHandlers(apiHandlers *handlers.APIHandlers) *GeminiAPIHandlers {
|
||||
return &GeminiAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// GeminiModels handles the Gemini models listing endpoint.
|
||||
// It returns a JSON response containing available Gemini models and their specifications.
|
||||
func (h *GeminiAPIHandlers) GeminiModels(c *gin.Context) {
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"data": []map[string]any{
|
||||
{
|
||||
"id": "gemini-2.5-flash",
|
||||
"object": "model",
|
||||
"version": "001",
|
||||
"name": "Gemini 2.5 Flash",
|
||||
"description": "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
{
|
||||
"id": "gemini-2.5-pro",
|
||||
"object": "model",
|
||||
"version": "2.5",
|
||||
"name": "Gemini 2.5 Pro",
|
||||
"description": "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
{
|
||||
"id": "gpt-5",
|
||||
"object": "model",
|
||||
"version": "gpt-5-2025-08-07",
|
||||
"name": "GPT 5",
|
||||
"description": "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
|
||||
"context_length": 400_000,
|
||||
"max_completion_tokens": 128_000,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
},
|
||||
})
|
||||
}
|
||||
|
||||
// GeminiGetHandler handles GET requests for specific Gemini model information.
|
||||
// It returns detailed information about a specific Gemini model based on the action parameter.
|
||||
func (h *GeminiAPIHandlers) GeminiGetHandler(c *gin.Context) {
|
||||
var request struct {
|
||||
Action string `uri:"action" binding:"required"`
|
||||
}
|
||||
if err := c.ShouldBindUri(&request); err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
switch request.Action {
|
||||
case "gemini-2.5-pro":
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"id": "gemini-2.5-pro",
|
||||
"object": "model",
|
||||
"version": "2.5",
|
||||
"name": "Gemini 2.5 Pro",
|
||||
"description": "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
})
|
||||
case "gemini-2.5-flash":
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"id": "gemini-2.5-flash",
|
||||
"object": "model",
|
||||
"version": "001",
|
||||
"name": "Gemini 2.5 Flash",
|
||||
"description": "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
})
|
||||
case "gpt-5":
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"id": "gpt-5",
|
||||
"object": "model",
|
||||
"version": "gpt-5-2025-08-07",
|
||||
"name": "GPT 5",
|
||||
"description": "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
|
||||
"context_length": 400_000,
|
||||
"max_completion_tokens": 128_000,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
})
|
||||
default:
|
||||
c.JSON(http.StatusNotFound, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Not Found",
|
||||
Type: "not_found",
|
||||
},
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
// GeminiHandler handles POST requests for Gemini API operations.
|
||||
// It routes requests to appropriate handlers based on the action parameter (model:method format).
|
||||
func (h *GeminiAPIHandlers) GeminiHandler(c *gin.Context) {
|
||||
var request struct {
|
||||
Action string `uri:"action" binding:"required"`
|
||||
}
|
||||
if err := c.ShouldBindUri(&request); err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
action := strings.Split(request.Action, ":")
|
||||
if len(action) != 2 {
|
||||
c.JSON(http.StatusNotFound, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("%s not found.", c.Request.URL.Path),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelName := action[0]
|
||||
method := action[1]
|
||||
rawJSON, _ := c.GetRawData()
|
||||
rawJSON, _ = sjson.SetBytes(rawJSON, "model", []byte(modelName))
|
||||
|
||||
provider := util.GetProviderName(modelName)
|
||||
if provider == "gemini" || provider == "unknow" {
|
||||
switch method {
|
||||
case "generateContent":
|
||||
h.handleGeminiGenerateContent(c, rawJSON)
|
||||
case "streamGenerateContent":
|
||||
h.handleGeminiStreamGenerateContent(c, rawJSON)
|
||||
case "countTokens":
|
||||
h.handleGeminiCountTokens(c, rawJSON)
|
||||
}
|
||||
} else if provider == "gpt" {
|
||||
switch method {
|
||||
case "generateContent":
|
||||
h.handleCodexGenerateContent(c, rawJSON)
|
||||
case "streamGenerateContent":
|
||||
h.handleCodexStreamGenerateContent(c, rawJSON)
|
||||
}
|
||||
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) handleGeminiStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
if alt == "" {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
}
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
template := ""
|
||||
parsed := gjson.Parse(string(rawJSON))
|
||||
contents := parsed.Get("request.contents")
|
||||
if contents.Exists() {
|
||||
template = string(rawJSON)
|
||||
} else {
|
||||
template = `{"project":"","request":{},"model":""}`
|
||||
template, _ = sjson.SetRaw(template, "request", string(rawJSON))
|
||||
template, _ = sjson.Set(template, "model", gjson.Get(template, "request.model").String())
|
||||
template, _ = sjson.Delete(template, "request.model")
|
||||
}
|
||||
|
||||
template, errFixCLIToolResponse := translatorGeminiToGeminiCli.FixCLIToolResponse(template)
|
||||
if errFixCLIToolResponse != nil {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: errFixCLIToolResponse.Error(),
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
systemInstructionResult := gjson.Get(template, "request.system_instruction")
|
||||
if systemInstructionResult.Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.systemInstruction", systemInstructionResult.Raw)
|
||||
template, _ = sjson.Delete(template, "request.system_instruction")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, rawJSON, alt)
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
if cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey() == "" {
|
||||
if alt == "" {
|
||||
responseResult := gjson.GetBytes(chunk, "response")
|
||||
if responseResult.Exists() {
|
||||
chunk = []byte(responseResult.Raw)
|
||||
}
|
||||
} else {
|
||||
chunkTemplate := "[]"
|
||||
responseResult := gjson.ParseBytes(chunk)
|
||||
if responseResult.IsArray() {
|
||||
responseResultItems := responseResult.Array()
|
||||
for i := 0; i < len(responseResultItems); i++ {
|
||||
responseResultItem := responseResultItems[i]
|
||||
if responseResultItem.Get("response").Exists() {
|
||||
chunkTemplate, _ = sjson.SetRaw(chunkTemplate, "-1", responseResultItem.Get("response").Raw)
|
||||
}
|
||||
}
|
||||
}
|
||||
chunk = []byte(chunkTemplate)
|
||||
}
|
||||
}
|
||||
if alt == "" {
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
} else {
|
||||
_, _ = c.Writer.Write(chunk)
|
||||
}
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
log.Debugf("quota exceeded, switch client")
|
||||
continue outLoop
|
||||
} else {
|
||||
log.Debugf("error code :%d, error: %v", err.StatusCode, err.Error.Error())
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) handleGeminiCountTokens(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
alt := h.GetAlt(c)
|
||||
// orgrawJSON := rawJSON
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName, false)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
|
||||
template := `{"request":{}}`
|
||||
if gjson.GetBytes(rawJSON, "generateContentRequest").Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request", gjson.GetBytes(rawJSON, "generateContentRequest").Raw)
|
||||
template, _ = sjson.Delete(template, "generateContentRequest")
|
||||
} else if gjson.GetBytes(rawJSON, "contents").Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.contents", gjson.GetBytes(rawJSON, "contents").Raw)
|
||||
template, _ = sjson.Delete(template, "contents")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendRawTokenCount(cliCtx, rawJSON, alt)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
// log.Debugf(err.Error.Error())
|
||||
// log.Debugf(string(rawJSON))
|
||||
// log.Debugf(string(orgrawJSON))
|
||||
}
|
||||
break
|
||||
} else {
|
||||
if cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey() == "" {
|
||||
responseResult := gjson.GetBytes(resp, "response")
|
||||
if responseResult.Exists() {
|
||||
resp = []byte(responseResult.Raw)
|
||||
}
|
||||
}
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) handleGeminiGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
alt := h.GetAlt(c)
|
||||
|
||||
modelResult := gjson.GetBytes(rawJSON, "model")
|
||||
modelName := modelResult.String()
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
template := ""
|
||||
parsed := gjson.Parse(string(rawJSON))
|
||||
contents := parsed.Get("request.contents")
|
||||
if contents.Exists() {
|
||||
template = string(rawJSON)
|
||||
} else {
|
||||
template = `{"project":"","request":{},"model":""}`
|
||||
template, _ = sjson.SetRaw(template, "request", string(rawJSON))
|
||||
template, _ = sjson.Set(template, "model", gjson.Get(template, "request.model").String())
|
||||
template, _ = sjson.Delete(template, "request.model")
|
||||
}
|
||||
|
||||
template, errFixCLIToolResponse := translatorGeminiToGeminiCli.FixCLIToolResponse(template)
|
||||
if errFixCLIToolResponse != nil {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: errFixCLIToolResponse.Error(),
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
systemInstructionResult := gjson.Get(template, "request.system_instruction")
|
||||
if systemInstructionResult.Exists() {
|
||||
template, _ = sjson.SetRaw(template, "request.systemInstruction", systemInstructionResult.Raw)
|
||||
template, _ = sjson.Delete(template, "request.system_instruction")
|
||||
}
|
||||
rawJSON = []byte(template)
|
||||
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
resp, err := cliClient.SendRawMessage(cliCtx, rawJSON, alt)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
if cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey() == "" {
|
||||
responseResult := gjson.GetBytes(resp, "response")
|
||||
if responseResult.Exists() {
|
||||
resp = []byte(responseResult.Raw)
|
||||
}
|
||||
}
|
||||
_, _ = c.Writer.Write(resp)
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) handleCodexStreamGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
newRequestJSON := translatorGeminiToCodex.ConvertGeminiRequestToCodex(rawJSON)
|
||||
// log.Debugf("Request: %s", newRequestJSON)
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
params := &translatorGeminiToCodex.ConvertCodexResponseToGeminiParams{
|
||||
Model: modelName.String(),
|
||||
CreatedAt: 0,
|
||||
ResponseID: "",
|
||||
LastStorageOutput: "",
|
||||
}
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() != "" {
|
||||
outputs := translatorGeminiToCodex.ConvertCodexResponseToGemini(jsonData, params)
|
||||
if len(outputs) > 0 {
|
||||
for i := 0; i < len(outputs); i++ {
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write([]byte(outputs[i]))
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
}
|
||||
}
|
||||
}
|
||||
// log.Debugf(string(jsonData))
|
||||
}
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (h *GeminiAPIHandlers) handleCodexGenerateContent(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
newRequestJSON := translatorGeminiToCodex.ConvertGeminiRequestToCodex(rawJSON)
|
||||
// log.Debugf("Request: %s", newRequestJSON)
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() != "" {
|
||||
var geminiStr string
|
||||
geminiStr = translatorGeminiToCodex.ConvertCodexResponseToGeminiNonStream(jsonData, modelName.String())
|
||||
if geminiStr != "" {
|
||||
_, _ = c.Writer.Write([]byte(geminiStr))
|
||||
}
|
||||
}
|
||||
}
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -5,52 +5,78 @@ package handlers
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"sync"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
"github.com/luispater/CLIProxyAPI/internal/config"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"sync"
|
||||
)
|
||||
|
||||
// ErrorResponse represents a standard error response format for the API.
|
||||
// It contains a single ErrorDetail field.
|
||||
type ErrorResponse struct {
|
||||
// Error contains detailed information about the error that occurred.
|
||||
Error ErrorDetail `json:"error"`
|
||||
}
|
||||
|
||||
// ErrorDetail provides specific information about an error that occurred.
|
||||
// It includes a human-readable message, an error type, and an optional error code.
|
||||
type ErrorDetail struct {
|
||||
// A human-readable message providing more details about the error.
|
||||
// Message is a human-readable message providing more details about the error.
|
||||
Message string `json:"message"`
|
||||
// The type of error that occurred (e.g., "invalid_request_error").
|
||||
|
||||
// Type is the category of error that occurred (e.g., "invalid_request_error").
|
||||
Type string `json:"type"`
|
||||
// A short code identifying the error, if applicable.
|
||||
|
||||
// Code is a short code identifying the error, if applicable.
|
||||
Code string `json:"code,omitempty"`
|
||||
}
|
||||
|
||||
// APIHandlers contains the handlers for API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
// It holds a pool of clients to interact with the backend service and manages
|
||||
// load balancing, client selection, and configuration.
|
||||
type APIHandlers struct {
|
||||
CliClients []*client.Client
|
||||
Cfg *config.Config
|
||||
Mutex *sync.Mutex
|
||||
LastUsedClientIndex int
|
||||
// CliClients is the pool of available AI service clients.
|
||||
CliClients []client.Client
|
||||
|
||||
// Cfg holds the current application configuration.
|
||||
Cfg *config.Config
|
||||
|
||||
// Mutex ensures thread-safe access to shared resources.
|
||||
Mutex *sync.Mutex
|
||||
|
||||
// LastUsedClientIndex tracks the last used client index for each provider
|
||||
// to implement round-robin load balancing.
|
||||
LastUsedClientIndex map[string]int
|
||||
}
|
||||
|
||||
// NewAPIHandlers creates a new API handlers instance.
|
||||
// It takes a slice of clients and a debug flag as input.
|
||||
func NewAPIHandlers(cliClients []*client.Client, cfg *config.Config) *APIHandlers {
|
||||
// It takes a slice of clients and configuration as input.
|
||||
//
|
||||
// Parameters:
|
||||
// - cliClients: A slice of AI service clients
|
||||
// - cfg: The application configuration
|
||||
//
|
||||
// Returns:
|
||||
// - *APIHandlers: A new API handlers instance
|
||||
func NewAPIHandlers(cliClients []client.Client, cfg *config.Config) *APIHandlers {
|
||||
return &APIHandlers{
|
||||
CliClients: cliClients,
|
||||
Cfg: cfg,
|
||||
Mutex: &sync.Mutex{},
|
||||
LastUsedClientIndex: 0,
|
||||
LastUsedClientIndex: make(map[string]int),
|
||||
}
|
||||
}
|
||||
|
||||
// UpdateClients updates the handlers' client list and configuration
|
||||
func (h *APIHandlers) UpdateClients(clients []*client.Client, cfg *config.Config) {
|
||||
// UpdateClients updates the handlers' client list and configuration.
|
||||
// This method is called when the configuration or authentication tokens change.
|
||||
//
|
||||
// Parameters:
|
||||
// - clients: The new slice of AI service clients
|
||||
// - cfg: The new application configuration
|
||||
func (h *APIHandlers) UpdateClients(clients []client.Client, cfg *config.Config) {
|
||||
h.CliClients = clients
|
||||
h.Cfg = cfg
|
||||
}
|
||||
@@ -58,30 +84,63 @@ func (h *APIHandlers) UpdateClients(clients []*client.Client, cfg *config.Config
|
||||
// GetClient returns an available client from the pool using round-robin load balancing.
|
||||
// It checks for quota limits and tries to find an unlocked client for immediate use.
|
||||
// The modelName parameter is used to check quota status for specific models.
|
||||
func (h *APIHandlers) GetClient(modelName string, isGenerateContent ...bool) (*client.Client, *client.ErrorMessage) {
|
||||
if len(h.CliClients) == 0 {
|
||||
//
|
||||
// Parameters:
|
||||
// - modelName: The name of the model to be used
|
||||
// - isGenerateContent: Optional parameter to indicate if this is for content generation
|
||||
//
|
||||
// Returns:
|
||||
// - client.Client: An available client for the requested model
|
||||
// - *client.ErrorMessage: An error message if no client is available
|
||||
func (h *APIHandlers) GetClient(modelName string, isGenerateContent ...bool) (client.Client, *client.ErrorMessage) {
|
||||
provider := util.GetProviderName(modelName)
|
||||
clients := make([]client.Client, 0)
|
||||
if provider == "gemini" {
|
||||
for i := 0; i < len(h.CliClients); i++ {
|
||||
if cli, ok := h.CliClients[i].(*client.GeminiClient); ok {
|
||||
clients = append(clients, cli)
|
||||
}
|
||||
}
|
||||
} else if provider == "gpt" {
|
||||
for i := 0; i < len(h.CliClients); i++ {
|
||||
if cli, ok := h.CliClients[i].(*client.CodexClient); ok {
|
||||
clients = append(clients, cli)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if _, hasKey := h.LastUsedClientIndex[provider]; !hasKey {
|
||||
h.LastUsedClientIndex[provider] = 0
|
||||
}
|
||||
|
||||
if len(clients) == 0 {
|
||||
return nil, &client.ErrorMessage{StatusCode: 500, Error: fmt.Errorf("no clients available")}
|
||||
}
|
||||
|
||||
var cliClient *client.Client
|
||||
var cliClient client.Client
|
||||
|
||||
// Lock the mutex to update the last used client index
|
||||
h.Mutex.Lock()
|
||||
startIndex := h.LastUsedClientIndex
|
||||
startIndex := h.LastUsedClientIndex[provider]
|
||||
if (len(isGenerateContent) > 0 && isGenerateContent[0]) || len(isGenerateContent) == 0 {
|
||||
currentIndex := (startIndex + 1) % len(h.CliClients)
|
||||
h.LastUsedClientIndex = currentIndex
|
||||
currentIndex := (startIndex + 1) % len(clients)
|
||||
h.LastUsedClientIndex[provider] = currentIndex
|
||||
}
|
||||
h.Mutex.Unlock()
|
||||
|
||||
// Reorder the client to start from the last used index
|
||||
reorderedClients := make([]*client.Client, 0)
|
||||
for i := 0; i < len(h.CliClients); i++ {
|
||||
cliClient = h.CliClients[(startIndex+1+i)%len(h.CliClients)]
|
||||
reorderedClients := make([]client.Client, 0)
|
||||
for i := 0; i < len(clients); i++ {
|
||||
cliClient = clients[(startIndex+1+i)%len(clients)]
|
||||
if cliClient.IsModelQuotaExceeded(modelName) {
|
||||
log.Debugf("Model %s is quota exceeded for account %s, project id: %s", modelName, cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
if provider == "gemini" {
|
||||
log.Debugf("Gemini Model %s is quota exceeded for account %s, project id: %s", modelName, cliClient.GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
} else if provider == "gpt" {
|
||||
log.Debugf("Codex Model %s is quota exceeded for account %s", modelName, cliClient.GetEmail())
|
||||
}
|
||||
cliClient = nil
|
||||
continue
|
||||
|
||||
}
|
||||
reorderedClients = append(reorderedClients, cliClient)
|
||||
}
|
||||
@@ -93,14 +152,14 @@ func (h *APIHandlers) GetClient(modelName string, isGenerateContent ...bool) (*c
|
||||
locked := false
|
||||
for i := 0; i < len(reorderedClients); i++ {
|
||||
cliClient = reorderedClients[i]
|
||||
if cliClient.RequestMutex.TryLock() {
|
||||
if cliClient.GetRequestMutex().TryLock() {
|
||||
locked = true
|
||||
break
|
||||
}
|
||||
}
|
||||
if !locked {
|
||||
cliClient = h.CliClients[0]
|
||||
cliClient.RequestMutex.Lock()
|
||||
cliClient = clients[0]
|
||||
cliClient.GetRequestMutex().Lock()
|
||||
}
|
||||
|
||||
return cliClient, nil
|
||||
@@ -108,6 +167,12 @@ func (h *APIHandlers) GetClient(modelName string, isGenerateContent ...bool) (*c
|
||||
|
||||
// GetAlt extracts the 'alt' parameter from the request query string.
|
||||
// It checks both 'alt' and '$alt' parameters and returns the appropriate value.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request
|
||||
//
|
||||
// Returns:
|
||||
// - string: The alt parameter value, or empty string if it's "sse"
|
||||
func (h *APIHandlers) GetAlt(c *gin.Context) string {
|
||||
var alt string
|
||||
var hasAlt bool
|
||||
|
||||
@@ -1,264 +0,0 @@
|
||||
// Package openai provides HTTP handlers for OpenAI API endpoints.
|
||||
// This package implements the OpenAI-compatible API interface, including model listing
|
||||
// and chat completion functionality. It supports both streaming and non-streaming responses,
|
||||
// and manages a pool of clients to interact with backend services.
|
||||
// The handlers translate OpenAI API requests to the appropriate backend format and
|
||||
// convert responses back to OpenAI-compatible format.
|
||||
package openai
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/translator/openai"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
"net/http"
|
||||
"time"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
)
|
||||
|
||||
// OpenAIAPIHandlers contains the handlers for OpenAI API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type OpenAIAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewOpenAIAPIHandlers creates a new OpenAI API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns an OpenAIAPIHandlers.
|
||||
func NewOpenAIAPIHandlers(apiHandlers *handlers.APIHandlers) *OpenAIAPIHandlers {
|
||||
return &OpenAIAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// Models handles the /v1/models endpoint.
|
||||
// It returns a hardcoded list of available AI models.
|
||||
func (h *OpenAIAPIHandlers) Models(c *gin.Context) {
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"data": []map[string]any{
|
||||
{
|
||||
"id": "gemini-2.5-pro",
|
||||
"object": "model",
|
||||
"version": "2.5",
|
||||
"name": "Gemini 2.5 Pro",
|
||||
"description": "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
|
||||
"context_length": 1048576,
|
||||
"max_completion_tokens": 65536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
{
|
||||
"id": "gemini-2.5-flash",
|
||||
"object": "model",
|
||||
"version": "001",
|
||||
"name": "Gemini 2.5 Flash",
|
||||
"description": "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
|
||||
"context_length": 1048576,
|
||||
"max_completion_tokens": 65536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
},
|
||||
})
|
||||
}
|
||||
|
||||
// ChatCompletions handles the /v1/chat/completions endpoint.
|
||||
// It determines whether the request is for a streaming or non-streaming response
|
||||
// and calls the appropriate handler.
|
||||
func (h *OpenAIAPIHandlers) ChatCompletions(c *gin.Context) {
|
||||
rawJSON, err := c.GetRawData()
|
||||
// If data retrieval fails, return a 400 Bad Request error.
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Check if the client requested a streaming response.
|
||||
streamResult := gjson.GetBytes(rawJSON, "stream")
|
||||
if streamResult.Type == gjson.True {
|
||||
h.handleStreamingResponse(c, rawJSON)
|
||||
} else {
|
||||
h.handleNonStreamingResponse(c, rawJSON)
|
||||
}
|
||||
}
|
||||
|
||||
// handleNonStreamingResponse handles non-streaming chat completion responses.
|
||||
// It selects a client from the pool, sends the request, and aggregates the response
|
||||
// before sending it back to the client.
|
||||
func (h *OpenAIAPIHandlers) handleNonStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
modelName, systemInstruction, contents, tools := openai.PrepareRequest(rawJSON)
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendMessage(cliCtx, rawJSON, modelName, systemInstruction, contents, tools)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
openAIFormat := openai.ConvertCliToOpenAINonStream(resp, time.Now().Unix(), isGlAPIKey)
|
||||
if openAIFormat != "" {
|
||||
_, _ = c.Writer.Write([]byte(openAIFormat))
|
||||
}
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleStreamingResponse handles streaming responses
|
||||
func (h *OpenAIAPIHandlers) handleStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
modelName, systemInstruction, contents, tools := openai.PrepareRequest(rawJSON)
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient *client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.RequestMutex.Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request use account: %s, project id: %s", cliClient.GetEmail(), cliClient.GetProjectID())
|
||||
}
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendMessageStream(cliCtx, rawJSON, modelName, systemInstruction, contents, tools)
|
||||
hasFirstResponse := false
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("Client disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
// Stream is closed, send the final [DONE] message.
|
||||
_, _ = fmt.Fprintf(c.Writer, "data: [DONE]\n\n")
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// Convert the chunk to OpenAI format and send it to the client.
|
||||
hasFirstResponse = true
|
||||
openAIFormat := openai.ConvertCliToOpenAI(chunk, time.Now().Unix(), isGlAPIKey)
|
||||
if openAIFormat != "" {
|
||||
_, _ = fmt.Fprintf(c.Writer, "data: %s\n\n", openAIFormat)
|
||||
flusher.Flush()
|
||||
}
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
_, _ = c.Writer.Write([]byte(": CLI-PROXY-API PROCESSING\n\n"))
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
506
internal/api/handlers/openai/openai_handlers.go
Normal file
506
internal/api/handlers/openai/openai_handlers.go
Normal file
@@ -0,0 +1,506 @@
|
||||
// Package openai provides HTTP handlers for OpenAI API endpoints.
|
||||
// This package implements the OpenAI-compatible API interface, including model listing
|
||||
// and chat completion functionality. It supports both streaming and non-streaming responses,
|
||||
// and manages a pool of clients to interact with backend services.
|
||||
// The handlers translate OpenAI API requests to the appropriate backend format and
|
||||
// convert responses back to OpenAI-compatible format.
|
||||
package openai
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"net/http"
|
||||
"time"
|
||||
|
||||
"github.com/luispater/CLIProxyAPI/internal/api/handlers"
|
||||
"github.com/luispater/CLIProxyAPI/internal/client"
|
||||
translatorOpenAIToCodex "github.com/luispater/CLIProxyAPI/internal/translator/codex/openai"
|
||||
translatorOpenAIToGeminiCli "github.com/luispater/CLIProxyAPI/internal/translator/gemini-cli/openai"
|
||||
"github.com/luispater/CLIProxyAPI/internal/util"
|
||||
log "github.com/sirupsen/logrus"
|
||||
"github.com/tidwall/gjson"
|
||||
|
||||
"github.com/gin-gonic/gin"
|
||||
)
|
||||
|
||||
// OpenAIAPIHandlers contains the handlers for OpenAI API endpoints.
|
||||
// It holds a pool of clients to interact with the backend service.
|
||||
type OpenAIAPIHandlers struct {
|
||||
*handlers.APIHandlers
|
||||
}
|
||||
|
||||
// NewOpenAIAPIHandlers creates a new OpenAI API handlers instance.
|
||||
// It takes an APIHandlers instance as input and returns an OpenAIAPIHandlers.
|
||||
//
|
||||
// Parameters:
|
||||
// - apiHandlers: The base API handlers instance
|
||||
//
|
||||
// Returns:
|
||||
// - *OpenAIAPIHandlers: A new OpenAI API handlers instance
|
||||
func NewOpenAIAPIHandlers(apiHandlers *handlers.APIHandlers) *OpenAIAPIHandlers {
|
||||
return &OpenAIAPIHandlers{
|
||||
APIHandlers: apiHandlers,
|
||||
}
|
||||
}
|
||||
|
||||
// Models handles the /v1/models endpoint.
|
||||
// It returns a hardcoded list of available AI models with their capabilities
|
||||
// and specifications in OpenAI-compatible format.
|
||||
func (h *OpenAIAPIHandlers) Models(c *gin.Context) {
|
||||
c.JSON(http.StatusOK, gin.H{
|
||||
"data": []map[string]any{
|
||||
{
|
||||
"id": "gemini-2.5-pro",
|
||||
"object": "model",
|
||||
"version": "2.5",
|
||||
"name": "Gemini 2.5 Pro",
|
||||
"description": "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
{
|
||||
"id": "gemini-2.5-flash",
|
||||
"object": "model",
|
||||
"version": "001",
|
||||
"name": "Gemini 2.5 Flash",
|
||||
"description": "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
|
||||
"context_length": 1_048_576,
|
||||
"max_completion_tokens": 65_536,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
"temperature",
|
||||
"top_p",
|
||||
"top_k",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
{
|
||||
"id": "gpt-5",
|
||||
"object": "model",
|
||||
"version": "gpt-5-2025-08-07",
|
||||
"name": "GPT 5",
|
||||
"description": "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
|
||||
"context_length": 400_000,
|
||||
"max_completion_tokens": 128_000,
|
||||
"supported_parameters": []string{
|
||||
"tools",
|
||||
},
|
||||
"temperature": 1,
|
||||
"topP": 0.95,
|
||||
"topK": 64,
|
||||
"maxTemperature": 2,
|
||||
"thinking": true,
|
||||
},
|
||||
},
|
||||
})
|
||||
}
|
||||
|
||||
// ChatCompletions handles the /v1/chat/completions endpoint.
|
||||
// It determines whether the request is for a streaming or non-streaming response
|
||||
// and calls the appropriate handler based on the model provider.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request and response
|
||||
func (h *OpenAIAPIHandlers) ChatCompletions(c *gin.Context) {
|
||||
rawJSON, err := c.GetRawData()
|
||||
// If data retrieval fails, return a 400 Bad Request error.
|
||||
if err != nil {
|
||||
c.JSON(http.StatusBadRequest, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: fmt.Sprintf("Invalid request: %v", err),
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Check if the client requested a streaming response.
|
||||
streamResult := gjson.GetBytes(rawJSON, "stream")
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
provider := util.GetProviderName(modelName.String())
|
||||
if provider == "gemini" {
|
||||
if streamResult.Type == gjson.True {
|
||||
h.handleGeminiStreamingResponse(c, rawJSON)
|
||||
} else {
|
||||
h.handleGeminiNonStreamingResponse(c, rawJSON)
|
||||
}
|
||||
} else if provider == "gpt" {
|
||||
if streamResult.Type == gjson.True {
|
||||
h.handleCodexStreamingResponse(c, rawJSON)
|
||||
} else {
|
||||
h.handleCodexNonStreamingResponse(c, rawJSON)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleGeminiNonStreamingResponse handles non-streaming chat completion responses
|
||||
// for Gemini models. It selects a client from the pool, sends the request, and
|
||||
// aggregates the response before sending it back to the client in OpenAI format.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request and response
|
||||
// - rawJSON: The raw JSON bytes of the OpenAI-compatible request
|
||||
func (h *OpenAIAPIHandlers) handleGeminiNonStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
modelName, systemInstruction, contents, tools := translatorOpenAIToGeminiCli.ConvertOpenAIChatRequestToCli(rawJSON)
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.(*client.GeminiClient).GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
|
||||
resp, err := cliClient.SendMessage(cliCtx, rawJSON, modelName, systemInstruction, contents, tools)
|
||||
if err != nil {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
break
|
||||
} else {
|
||||
openAIFormat := translatorOpenAIToGeminiCli.ConvertCliResponseToOpenAIChatNonStream(resp, time.Now().Unix(), isGlAPIKey)
|
||||
if openAIFormat != "" {
|
||||
_, _ = c.Writer.Write([]byte(openAIFormat))
|
||||
}
|
||||
cliCancel()
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleGeminiStreamingResponse handles streaming responses for Gemini models.
|
||||
// It establishes a streaming connection with the backend service and forwards
|
||||
// the response chunks to the client in real-time using Server-Sent Events.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request and response
|
||||
// - rawJSON: The raw JSON bytes of the OpenAI-compatible request
|
||||
func (h *OpenAIAPIHandlers) handleGeminiStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
modelName, systemInstruction, contents, tools := translatorOpenAIToGeminiCli.ConvertOpenAIChatRequestToCli(rawJSON)
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName)
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
isGlAPIKey := false
|
||||
if glAPIKey := cliClient.(*client.GeminiClient).GetGenerativeLanguageAPIKey(); glAPIKey != "" {
|
||||
log.Debugf("Request use generative language API Key: %s", glAPIKey)
|
||||
isGlAPIKey = true
|
||||
} else {
|
||||
log.Debugf("Request cli use account: %s, project id: %s", cliClient.GetEmail(), cliClient.(*client.GeminiClient).GetProjectID())
|
||||
}
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendMessageStream(cliCtx, rawJSON, modelName, systemInstruction, contents, tools)
|
||||
hasFirstResponse := false
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("GeminiClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
// Stream is closed, send the final [DONE] message.
|
||||
_, _ = fmt.Fprintf(c.Writer, "data: [DONE]\n\n")
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// Convert the chunk to OpenAI format and send it to the client.
|
||||
hasFirstResponse = true
|
||||
openAIFormat := translatorOpenAIToGeminiCli.ConvertCliResponseToOpenAIChat(chunk, time.Now().Unix(), isGlAPIKey)
|
||||
if openAIFormat != "" {
|
||||
_, _ = fmt.Fprintf(c.Writer, "data: %s\n\n", openAIFormat)
|
||||
flusher.Flush()
|
||||
}
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
if hasFirstResponse {
|
||||
_, _ = c.Writer.Write([]byte(": CLI-PROXY-API PROCESSING\n\n"))
|
||||
flusher.Flush()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleCodexNonStreamingResponse handles non-streaming chat completion responses
|
||||
// for OpenAI models. It selects a client from the pool, sends the request, and
|
||||
// aggregates the response before sending it back to the client in OpenAI format.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request and response
|
||||
// - rawJSON: The raw JSON bytes of the OpenAI-compatible request
|
||||
func (h *OpenAIAPIHandlers) handleCodexNonStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "application/json")
|
||||
|
||||
newRequestJSON := translatorOpenAIToCodex.ConvertOpenAIChatRequestToCodex(rawJSON)
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(errorResponse.Error.Error()))
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() == "response.completed" {
|
||||
responseResult := data.Get("response")
|
||||
openaiStr := translatorOpenAIToCodex.ConvertCodexResponseToOpenAIChatNonStream(responseResult.Raw, time.Now().Unix())
|
||||
_, _ = c.Writer.Write([]byte(openaiStr))
|
||||
}
|
||||
}
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = c.Writer.Write([]byte(err.Error.Error()))
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// handleCodexStreamingResponse handles streaming responses for OpenAI models.
|
||||
// It establishes a streaming connection with the backend service and forwards
|
||||
// the response chunks to the client in real-time using Server-Sent Events.
|
||||
//
|
||||
// Parameters:
|
||||
// - c: The Gin context containing the HTTP request and response
|
||||
// - rawJSON: The raw JSON bytes of the OpenAI-compatible request
|
||||
func (h *OpenAIAPIHandlers) handleCodexStreamingResponse(c *gin.Context, rawJSON []byte) {
|
||||
c.Header("Content-Type", "text/event-stream")
|
||||
c.Header("Cache-Control", "no-cache")
|
||||
c.Header("Connection", "keep-alive")
|
||||
c.Header("Access-Control-Allow-Origin", "*")
|
||||
|
||||
// Get the http.Flusher interface to manually flush the response.
|
||||
flusher, ok := c.Writer.(http.Flusher)
|
||||
if !ok {
|
||||
c.JSON(http.StatusInternalServerError, handlers.ErrorResponse{
|
||||
Error: handlers.ErrorDetail{
|
||||
Message: "Streaming not supported",
|
||||
Type: "server_error",
|
||||
},
|
||||
})
|
||||
return
|
||||
}
|
||||
|
||||
// Prepare the request for the backend client.
|
||||
newRequestJSON := translatorOpenAIToCodex.ConvertOpenAIChatRequestToCodex(rawJSON)
|
||||
// log.Debugf("Request: %s", newRequestJSON)
|
||||
|
||||
modelName := gjson.GetBytes(rawJSON, "model")
|
||||
|
||||
cliCtx, cliCancel := context.WithCancel(context.Background())
|
||||
var cliClient client.Client
|
||||
defer func() {
|
||||
// Ensure the client's mutex is unlocked on function exit.
|
||||
if cliClient != nil {
|
||||
cliClient.GetRequestMutex().Unlock()
|
||||
}
|
||||
}()
|
||||
|
||||
outLoop:
|
||||
for {
|
||||
var errorResponse *client.ErrorMessage
|
||||
cliClient, errorResponse = h.GetClient(modelName.String())
|
||||
if errorResponse != nil {
|
||||
c.Status(errorResponse.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, errorResponse.Error)
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
|
||||
log.Debugf("Request codex use account: %s", cliClient.GetEmail())
|
||||
|
||||
// Send the message and receive response chunks and errors via channels.
|
||||
var params *translatorOpenAIToCodex.ConvertCliToOpenAIParams
|
||||
respChan, errChan := cliClient.SendRawMessageStream(cliCtx, []byte(newRequestJSON), "")
|
||||
for {
|
||||
select {
|
||||
// Handle client disconnection.
|
||||
case <-c.Request.Context().Done():
|
||||
if c.Request.Context().Err().Error() == "context canceled" {
|
||||
log.Debugf("CodexClient disconnected: %v", c.Request.Context().Err())
|
||||
cliCancel() // Cancel the backend request.
|
||||
return
|
||||
}
|
||||
// Process incoming response chunks.
|
||||
case chunk, okStream := <-respChan:
|
||||
if !okStream {
|
||||
_, _ = c.Writer.Write([]byte("[done]\n\n"))
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
return
|
||||
}
|
||||
// log.Debugf("Response: %s\n", string(chunk))
|
||||
// Convert the chunk to OpenAI format and send it to the client.
|
||||
if bytes.HasPrefix(chunk, []byte("data: ")) {
|
||||
jsonData := chunk[6:]
|
||||
data := gjson.ParseBytes(jsonData)
|
||||
typeResult := data.Get("type")
|
||||
if typeResult.String() != "" {
|
||||
var openaiStr string
|
||||
params, openaiStr = translatorOpenAIToCodex.ConvertCodexResponseToOpenAIChat(jsonData, params)
|
||||
if openaiStr != "" {
|
||||
_, _ = c.Writer.Write([]byte("data: "))
|
||||
_, _ = c.Writer.Write([]byte(openaiStr))
|
||||
_, _ = c.Writer.Write([]byte("\n\n"))
|
||||
}
|
||||
}
|
||||
// log.Debugf(string(jsonData))
|
||||
}
|
||||
flusher.Flush()
|
||||
// Handle errors from the backend.
|
||||
case err, okError := <-errChan:
|
||||
if okError {
|
||||
if err.StatusCode == 429 && h.Cfg.QuotaExceeded.SwitchProject {
|
||||
continue outLoop
|
||||
} else {
|
||||
c.Status(err.StatusCode)
|
||||
_, _ = fmt.Fprint(c.Writer, err.Error.Error())
|
||||
flusher.Flush()
|
||||
cliCancel()
|
||||
}
|
||||
return
|
||||
}
|
||||
// Send a keep-alive signal to the client.
|
||||
case <-time.After(500 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user