proxx/handler.go
Franz Kafka 1c3eccce10 Use shared HTTP client for upstream requests
- Create package-level httpClient with 300s timeout\n- Reuse client instead of creating new one per request\n- Prevents resource exhaustion under load\n- Reduces connection overhead
2026-04-15 09:02:17 +00:00

248 lines
7.5 KiB
Go

package main
import (
"encoding/json"
"fmt"
"io"
"log"
"math/rand"
"net/http"
"strings"
"time"
)
// Config holds the application configuration
type Config struct {
Port int `yaml:"port"`
UpstreamURL string `yaml:"upstream_url"`
}
var config *Config
// httpClient is a shared HTTP client for all upstream requests
var httpClient = &http.Client{Timeout: 300 * time.Second}
// blockedHeaders are headers that should never be forwarded to upstream
// for security/privacy reasons. These headers could leak internal URLs,
// session information, or other sensitive data.
var blockedHeaders = map[string]bool{
"Referer": true, // Don't leak internal URLs to external API
"Cookie": true, // Don't forward session cookies
"Authorization": true, // Already extracted and sent as x-api-key
"X-Forwarded-For": true, // Don't leak client IP
"X-Real-Ip": true, // Don't leak client IP
"X-Forwarded-Host": true, // Don't leak internal hostnames
}
// ClaudeCodeHeaders returns the headers to mimic claude-code CLI
func ClaudeCodeHeaders(apiKey, sessionID string) map[string]string {
return map[string]string{
"User-Agent": "claude-cli/1.0.18 (pro, cli)",
"x-api-key": apiKey,
"x-app": "cli",
"anthropic-version": "2023-06-01",
"anthropic-beta": "claude-code-20250219,interleaved-thinking-2025-05-14,prompt-caching-scope-2026-01-05,context-management-2025-06-27,redact-thinking-2026-02-12",
"X-Claude-Code-Session-Id": sessionID,
"content-type": "application/json",
}
}
func handleModels(w http.ResponseWriter, r *http.Request) {
if r.Method != http.MethodGet {
writeError(w, http.StatusMethodNotAllowed, "Method not allowed", "invalid_request_error", "method_not_allowed")
return
}
models := []map[string]interface{}{
{"id": "glm-4.7", "object": "model", "created": 1234567890, "owned_by": "zhipu"},
{"id": "glm-4.6", "object": "model", "created": 1234567890, "owned_by": "zhipu"},
}
response := map[string]interface{}{
"object": "list",
"data": models,
}
w.Header().Set("content-type", "application/json")
json.NewEncoder(w).Encode(response)
}
func handleChatCompletions(w http.ResponseWriter, r *http.Request) {
if r.Method != http.MethodPost {
writeError(w, http.StatusMethodNotAllowed, "Method not allowed", "invalid_request_error", "method_not_allowed")
return
}
// Extract Bearer token
authHeader := r.Header.Get("Authorization")
if authHeader == "" || !strings.HasPrefix(authHeader, "Bearer ") {
writeError(w, http.StatusUnauthorized, "Missing or invalid Authorization header", "authentication_error", "missing_authorization")
return
}
apiKey := strings.TrimPrefix(authHeader, "Bearer ")
// Read body
body, err := io.ReadAll(r.Body)
if err != nil {
writeError(w, http.StatusBadRequest, "Failed to read request body", "invalid_request_error", "body_read_error")
return
}
// Decode request
var req ChatCompletionRequest
if err := json.Unmarshal(body, &req); err != nil {
writeError(w, http.StatusBadRequest, "Invalid JSON in request body", "invalid_request_error", "json_decode_error")
return
}
// Get session ID from context (set by main)
sessionID := r.Context().Value(sessionIDKey).(string)
// Convert to Anthropic format — always non-streaming to upstream
// (ZAI's streaming returns empty for GLM models)
anthropicReq := ConvertOpenAIRequest(&req)
anthropicReq.Stream = false
reqBody, _ := json.Marshal(anthropicReq)
log.Printf("[debug] Sending to upstream %s, model=%s, body=%s", config.UpstreamURL, req.Model, string(reqBody))
// Non-streaming request to upstream
upstreamResp, err := callUpstream(anthropicReq, apiKey, sessionID)
if err != nil {
writeError(w, http.StatusBadGateway, fmt.Sprintf("Upstream request failed: %v", err), "upstream_error", "proxy_error")
return
}
defer upstreamResp.Body.Close()
if upstreamResp.StatusCode != http.StatusOK {
respBody, _ := io.ReadAll(upstreamResp.Body)
log.Printf("[debug] Upstream error status %d: %s", upstreamResp.StatusCode, string(respBody))
writeError(w, http.StatusBadGateway, fmt.Sprintf("Upstream returned error: %s", string(respBody)), "upstream_error", fmt.Sprintf("status_%d", upstreamResp.StatusCode))
return
}
// Read the full Anthropic response
respBody, err := io.ReadAll(upstreamResp.Body)
if err != nil {
writeError(w, http.StatusBadGateway, "Failed to read upstream response", "upstream_error", "body_read_error")
return
}
log.Printf("[debug] Upstream response: %s", string(respBody))
var anthropicResp AnthropicResponse
if err := json.Unmarshal(respBody, &anthropicResp); err != nil {
writeError(w, http.StatusBadGateway, "Failed to parse upstream response", "upstream_error", "json_decode_error")
return
}
isStream := req.Stream != nil && *req.Stream
if isStream {
// Convert the non-streaming response to SSE chunks for the client
w.Header().Set("content-type", "text/event-stream")
w.Header().Set("cache-control", "no-cache")
w.Header().Set("connection", "keep-alive")
created := time.Now().Unix()
chunkID := "chatcmpl-" + randomString(8)
// Extract text content
var textContent string
for _, block := range anthropicResp.Content {
if block.Type == "text" {
textContent += block.Text
}
}
// Send text as chunks (simulate streaming)
if textContent != "" {
chunk := StreamChunk{
ID: chunkID,
Object: "chat.completion.chunk",
Created: created,
Model: req.Model,
Choices: []StreamChoice{
{
Index: 0,
Delta: Delta{
Role: "assistant",
Content: textContent,
},
},
},
}
data, _ := json.Marshal(chunk)
fmt.Fprintf(w, "data: %s\n\n", data)
w.(http.Flusher).Flush()
}
// Send finish chunk
finishChunk := StreamChunk{
ID: chunkID,
Object: "chat.completion.chunk",
Created: created,
Model: req.Model,
Choices: []StreamChoice{
{
Index: 0,
Delta: Delta{},
FinishReason: mapStopReason(anthropicResp.StopReason),
},
},
}
data, _ := json.Marshal(finishChunk)
fmt.Fprintf(w, "data: %s\n\n", data)
w.(http.Flusher).Flush()
fmt.Fprintf(w, "data: [DONE]\n\n")
w.(http.Flusher).Flush()
} else {
// Non-streaming: convert directly
openAIResp := ConvertAnthropicResponse(&anthropicResp, req.Model)
w.Header().Set("content-type", "application/json")
json.NewEncoder(w).Encode(openAIResp)
}
}
func callUpstream(req *AnthropicRequest, apiKey, sessionID string) (*http.Response, error) {
bodyBytes, err := json.Marshal(req)
if err != nil {
return nil, fmt.Errorf("failed to marshal request: %w", err)
}
upstreamURL := config.UpstreamURL + "/v1/messages"
httpReq, err := http.NewRequest(http.MethodPost, upstreamURL, strings.NewReader(string(bodyBytes)))
if err != nil {
return nil, fmt.Errorf("failed to create request: %w", err)
}
headers := ClaudeCodeHeaders(apiKey, sessionID)
for k, v := range headers {
httpReq.Header.Set(k, v)
}
return httpClient.Do(httpReq)
}
func writeError(w http.ResponseWriter, code int, message, errType, errCode string) {
w.Header().Set("content-type", "application/json")
w.WriteHeader(code)
resp := map[string]interface{}{
"error": map[string]string{
"message": message,
"type": errType,
"code": errCode,
},
}
json.NewEncoder(w).Encode(resp)
}
func randomString(n int) string {
const letters = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
b := make([]byte, n)
r := rand.New(rand.NewSource(time.Now().UnixNano()))
for i := range b {
b[i] = letters[r.Intn(len(letters))]
}
return string(b)
}