claude-code-proxy/proxy/internal/middleware/logging.go

267 lines
6.8 KiB
Go
Raw Normal View History

2025-06-29 19:27:00 -04:00
package middleware
import (
"bytes"
"context"
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
"encoding/json"
2025-08-03 22:30:13 -04:00
"fmt"
2025-06-29 19:27:00 -04:00
"io"
"log"
"net/http"
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
"strings"
2025-06-29 19:27:00 -04:00
"time"
"github.com/seifghazi/claude-code-monitor/internal/model"
)
func Logging(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
start := time.Now()
2025-08-03 22:30:13 -04:00
// For POST requests with body, read and store the bytes
2025-06-29 19:27:00 -04:00
var bodyBytes []byte
2025-08-03 22:30:13 -04:00
if r.Body != nil && (r.Method == "POST" || r.Method == "PUT" || r.Method == "PATCH") {
2025-06-29 19:27:00 -04:00
var err error
bodyBytes, err = io.ReadAll(r.Body)
if err != nil {
log.Printf("❌ Error reading request body: %v", err)
http.Error(w, "Error reading request body", http.StatusBadRequest)
return
}
r.Body.Close()
r.Body = io.NopCloser(bytes.NewReader(bodyBytes))
2025-08-03 22:30:13 -04:00
// Store raw bytes in context for handler to use
ctx := context.WithValue(r.Context(), model.BodyBytesKey, bodyBytes)
r = r.WithContext(ctx)
}
2025-06-29 19:27:00 -04:00
wrapped := &responseWriter{ResponseWriter: w, statusCode: http.StatusOK}
next.ServeHTTP(wrapped, r)
duration := time.Since(start)
2025-08-03 22:30:13 -04:00
statusColor := getStatusColor(wrapped.statusCode)
2025-06-29 19:27:00 -04:00
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
// Build a richer log line for proxy requests
if isProxyRequest(r.URL.Path) {
details := buildProxyLogDetails(r, bodyBytes, wrapped, duration)
log.Printf("%s%s%s %s",
statusColor, details, colorReset,
colorDim+formatDuration(duration)+colorReset)
} else {
log.Printf("%s %s %s%d%s (%s)",
r.Method,
r.URL.Path,
statusColor,
wrapped.statusCode,
colorReset,
formatDuration(duration))
}
2025-08-03 22:30:13 -04:00
})
2025-06-29 19:27:00 -04:00
}
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
// isProxyRequest returns true for /v1/* API proxy paths
func isProxyRequest(path string) bool {
return strings.HasPrefix(path, "/v1/")
}
// buildProxyLogDetails creates a rich single-line log for proxy requests
func buildProxyLogDetails(r *http.Request, bodyBytes []byte, w *responseWriter, duration time.Duration) string {
var parts []string
// Status
parts = append(parts, fmt.Sprintf("%d", w.statusCode))
// Method + path
parts = append(parts, fmt.Sprintf("%s %s", r.Method, r.URL.Path))
// Extract model and stream flag from request body
if len(bodyBytes) > 0 {
var body struct {
Model string `json:"model"`
Stream bool `json:"stream"`
}
if err := json.Unmarshal(bodyBytes, &body); err == nil {
if body.Model != "" {
// Shorten model name for readability
modelShort := shortenModel(body.Model)
parts = append(parts, colorCyan+modelShort+colorReset)
}
if body.Stream {
parts = append(parts, "stream")
}
}
}
// Client info — use X-Forwarded-For if behind proxy, else RemoteAddr
clientIP := r.Header.Get("X-Forwarded-For")
if clientIP == "" {
clientIP = r.Header.Get("X-Real-Ip")
}
if clientIP == "" {
clientIP = r.RemoteAddr
}
// Strip port from IP
if host, _, err := splitHostPort(clientIP); err == nil {
clientIP = host
}
// User-Agent — extract just the tool/client name
ua := r.Header.Get("User-Agent")
if clientName := extractClientName(ua); clientName != "" {
parts = append(parts, colorDim+clientName+colorReset)
}
if clientIP != "" {
parts = append(parts, colorDim+clientIP+colorReset)
}
return strings.Join(parts, " ")
}
// shortenModel turns "claude-sonnet-4-20250514" into "sonnet-4"
func shortenModel(model string) string {
lower := strings.ToLower(model)
for _, family := range []string{"opus", "sonnet", "haiku"} {
if strings.Contains(lower, family) {
// Find version number after family name
idx := strings.Index(lower, family)
rest := lower[idx+len(family):]
// Extract version like "-4" or "-4-20250514"
rest = strings.TrimLeft(rest, "-")
if dashIdx := strings.Index(rest, "-"); dashIdx > 0 {
// Keep just the version number (e.g. "4" from "4-20250514")
return family + "-" + rest[:dashIdx]
}
if rest != "" {
return family + "-" + rest
}
return family
}
}
// For non-Claude models, take first two segments
segs := strings.SplitN(model, "-", 3)
if len(segs) >= 2 {
return segs[0] + "-" + segs[1]
}
return model
}
// extractClientName pulls a recognizable client name from User-Agent
func extractClientName(ua string) string {
if ua == "" {
return ""
}
lower := strings.ToLower(ua)
switch {
case strings.Contains(lower, "claude-code"):
return "claude-code"
case strings.Contains(lower, "cursor"):
return "cursor"
case strings.Contains(lower, "continue"):
return "continue"
case strings.Contains(lower, "anthropic-sdk"):
return "sdk"
case strings.Contains(lower, "python"):
return "python"
case strings.Contains(lower, "node"):
return "node"
case strings.Contains(lower, "curl"):
return "curl"
default:
// Take first token
if spaceIdx := strings.IndexByte(ua, ' '); spaceIdx > 0 {
first := ua[:spaceIdx]
if len(first) > 20 {
return first[:20]
}
return first
}
if len(ua) > 20 {
return ua[:20]
}
return ua
}
}
// splitHostPort is a simple wrapper that handles IPs without ports
func splitHostPort(addr string) (string, string, error) {
// If there's no colon or it's an IPv6 without port, just return the addr
if !strings.Contains(addr, ":") {
return addr, "", nil
}
// Handle comma-separated X-Forwarded-For
if commaIdx := strings.IndexByte(addr, ','); commaIdx > 0 {
addr = strings.TrimSpace(addr[:commaIdx])
}
host, port, err := splitAddr(addr)
return host, port, err
}
func splitAddr(addr string) (string, string, error) {
if strings.Count(addr, ":") > 1 {
// IPv6 — may or may not have brackets
return addr, "", nil
}
idx := strings.LastIndexByte(addr, ':')
if idx < 0 {
return addr, "", nil
}
return addr[:idx], addr[idx+1:], nil
}
2025-06-29 19:27:00 -04:00
type responseWriter struct {
http.ResponseWriter
statusCode int
}
func (rw *responseWriter) WriteHeader(code int) {
rw.statusCode = code
rw.ResponseWriter.WriteHeader(code)
}
2025-08-03 22:30:13 -04:00
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
// Flush propagates to the underlying http.Flusher. Without this, embedding
// http.ResponseWriter (an interface) silently drops Flush(), so SSE writes
// buffer in net/http until the body closes — breaking token-by-token
// streaming UX.
func (rw *responseWriter) Flush() {
if f, ok := rw.ResponseWriter.(http.Flusher); ok {
f.Flush()
}
}
2025-08-03 22:30:13 -04:00
// ANSI color codes
const (
colorReset = "\033[0m"
colorGreen = "\033[32m"
colorYellow = "\033[33m"
colorRed = "\033[31m"
colorBlue = "\033[34m"
colorCyan = "\033[36m"
Local fork: hardening + ops improvements (timeout knob, demotion, /livez, drain) This commit captures both the prior accumulated work-in-progress (framework migration web/→svelte/, postgres storage, conversation viewer, dashboard auth, OpenAPI spec, integration tests) AND today's operational improvements layered on top. History wasn't checkpointed incrementally; happy to split it via interactive rebase if a reviewer wants smaller commits. Today's changes (in addition to the older WIP): 1. Configurable upstream response-header timeout - ANTHROPIC_RESPONSE_HEADER_TIMEOUT env (default 300s) - Replaces hardcoded 300s in provider/anthropic.go that was firing on opus + 1M-context + extended thinking non-streaming requests - Files: internal/config/config.go, internal/provider/anthropic.go 2. Structured forward-error diagnostic logging - When a forward to Anthropic fails, log a single key=value line with request_id, model, stream, body_bytes, has_thinking, anthropic_beta, query, elapsed, ctx_err — alongside the existing human-readable error line for back-compat - Files: internal/handler/handlers.go (logForwardFailure) 3. Full SSE protocol passthrough + Flusher fix - handler/handlers.go: forward all SSE lines verbatim (event:, id:, retry:, : comments, blank-line terminators), not only data:. Previous code produced malformed SSE for strict parsers. - middleware/logging.go: explicit Flush() method on responseWriter. Embedding http.ResponseWriter (interface) does not auto-promote Flush(), so every w.(http.Flusher) check in the streaming handler was returning ok=false and SSE writes buffered in net/http until the body closed. 4. Non-streaming → streaming demotion (feature-flagged) - ANTHROPIC_DEMOTE_NONSTREAMING env (default false) - When enabled and the routed provider is anthropic, force stream=true upstream for clients that asked for stream=false. Receive SSE, accumulate via accumulateSSEToMessage (handles text, tool_use with partial_json reassembly, thinking, signature, citations_delta, usage merge), and synthesize a single non-streaming JSON response. - Eliminates the ResponseHeaderTimeout class of failure entirely. - Body rewrite uses json.Decoder + UseNumber() to preserve integer precision in unknown nested fields (tool inputs from prior turns). - Files: internal/config/config.go, internal/handler/handlers.go, cmd/proxy/main.go, cmd/proxy/main_test.go 5. Live operational state: /livez gauge + graceful drain - New internal/runtime package: atomic in-flight counter + draining flag - New middleware/inflight.go: increments runtime gauge, applied to /v1/* subrouter so Messages, ChatCompletions, and ProxyPassthrough are all counted - /v1/* moved to a gorilla/mux subrouter so the InFlight middleware applies surgically; /health, /livez, /openapi.* remain on parent router (unauthenticated, uncounted) - Health handler returns 503 draining when runtime.IsDraining() is true, so Traefik stops routing to a slot before drain begins - New /livez handler returns {status, in_flight, draining, timestamp} - SIGTERM handler in main.go: SetDraining(true), poll for in_flight==0 with 32-min ceiling and 1s tick (logs every 10s), then srv.Shutdown - Auth bypass list extended with /livez - Files: internal/runtime/runtime.go (new), internal/middleware/inflight.go (new), internal/middleware/auth.go, internal/handler/handlers.go (Health, Livez, runtime import), cmd/proxy/main.go (subrouter, drain loop) 6. OpenAPI spec updates - Document Health 503 response and new DrainingResponse schema - Add /livez path with LivezResponse schema - Files: internal/handler/openapi.go Verified: go build ./... clean, go test ./... all pass, go vet clean. Three rounds of codex peer review across changes 1-5; all feedback addressed (citations_delta, json.Number precision, drain-loop logging via lastLog timestamp, PathPrefix tightened to "/v1/").
2026-05-02 15:15:58 -06:00
colorDim = "\033[2m"
2025-08-03 22:30:13 -04:00
)
func getStatusColor(status int) string {
switch {
case status >= 200 && status < 300:
return colorGreen
case status >= 300 && status < 400:
return colorBlue
case status >= 400 && status < 500:
return colorYellow
case status >= 500:
return colorRed
default:
return colorReset
}
}
func formatDuration(d time.Duration) string {
if d < time.Millisecond {
return fmt.Sprintf("%dµs", d.Microseconds())
} else if d < time.Second {
return fmt.Sprintf("%dms", d.Milliseconds())
}
return fmt.Sprintf("%.2fs", d.Seconds())
}