claude-code-proxy/proxy/internal/provider/anthropic.go

159 lines
4.2 KiB
Go
Raw Normal View History

2025-08-01 21:19:30 -04:00
package provider
import (
"compress/gzip"
"context"
"fmt"
"io"
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
"net"
2025-08-01 21:19:30 -04:00
"net/http"
"net/url"
"path"
"strings"
"time"
"github.com/seifghazi/claude-code-monitor/internal/config"
)
type AnthropicProvider struct {
client *http.Client
config *config.AnthropicProviderConfig
}
func NewAnthropicProvider(cfg *config.AnthropicProviderConfig) Provider {
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
respHeaderTimeout := cfg.ResponseHeaderTimeout
if respHeaderTimeout <= 0 {
respHeaderTimeout = 300 * time.Second
}
2025-08-01 21:19:30 -04:00
return &AnthropicProvider{
client: &http.Client{
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
// No Client.Timeout: a global timeout would cancel long streaming
// responses mid-flight. Per-phase timeouts on the Transport plus the
// 30-min context in handlers.Messages bound the request instead.
Transport: &http.Transport{
DialContext: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
}).DialContext,
TLSHandshakeTimeout: 30 * time.Second,
// Tunable via ANTHROPIC_RESPONSE_HEADER_TIMEOUT — opus + extended
// thinking on large contexts can take longer than the 300s default.
ResponseHeaderTimeout: respHeaderTimeout,
ExpectContinueTimeout: 1 * time.Second,
MaxIdleConns: 100,
MaxIdleConnsPerHost: 10,
IdleConnTimeout: 90 * time.Second,
},
2025-08-01 21:19:30 -04:00
},
config: cfg,
}
}
func (p *AnthropicProvider) Name() string {
return "anthropic"
}
func (p *AnthropicProvider) ForwardRequest(ctx context.Context, originalReq *http.Request) (*http.Response, error) {
// Clone the request to avoid modifying the original
proxyReq := originalReq.Clone(ctx)
// Parse the configured base URL
baseURL, err := url.Parse(p.config.BaseURL)
if err != nil {
return nil, fmt.Errorf("failed to parse base URL '%s': %w", p.config.BaseURL, err)
}
if baseURL.Scheme == "" || baseURL.Host == "" {
return nil, fmt.Errorf("invalid base URL, scheme and host are required: %s", p.config.BaseURL)
}
// Update the destination URL
proxyReq.URL.Scheme = baseURL.Scheme
proxyReq.URL.Host = baseURL.Host
proxyReq.URL.Path = path.Join(baseURL.Path, originalReq.URL.Path)
// Preserve query parameters
proxyReq.URL.RawQuery = originalReq.URL.RawQuery
// Update request headers
proxyReq.RequestURI = ""
proxyReq.Host = baseURL.Host
// Remove hop-by-hop headers
removeHopByHopHeaders(proxyReq.Header)
// Add required headers if not present
if proxyReq.Header.Get("anthropic-version") == "" {
proxyReq.Header.Set("anthropic-version", p.config.Version)
}
// Handle Accept-Encoding: We accept gzip from upstream for efficiency,
// but we decompress before forwarding to the client. This is transparent
// to the client - they receive uncompressed data regardless of what they requested.
// We preserve gzip if client already requested it, otherwise add it.
clientEncoding := proxyReq.Header.Get("Accept-Encoding")
if clientEncoding == "" || !strings.Contains(clientEncoding, "gzip") {
proxyReq.Header.Set("Accept-Encoding", "gzip")
}
2025-08-01 21:19:30 -04:00
// Forward the request
resp, err := p.client.Do(proxyReq)
if err != nil {
return nil, fmt.Errorf("failed to forward request: %w", err)
}
// Handle gzip-encoded responses
if resp.Header.Get("Content-Encoding") == "gzip" {
resp.Header.Del("Content-Encoding")
resp.Header.Del("Content-Length")
gzipReader, err := gzip.NewReader(resp.Body)
if err != nil {
resp.Body.Close()
return nil, fmt.Errorf("failed to create gzip reader: %w", err)
}
resp.Body = &gzipResponseBody{
Reader: gzipReader,
closer: resp.Body,
}
}
return resp, nil
}
type gzipResponseBody struct {
io.Reader
closer io.Closer
}
func (g *gzipResponseBody) Close() error {
if gzReader, ok := g.Reader.(*gzip.Reader); ok {
gzReader.Close()
}
return g.closer.Close()
}
func removeHopByHopHeaders(header http.Header) {
hopByHopHeaders := []string{
"Connection",
"Keep-Alive",
"Proxy-Authenticate",
"Proxy-Authorization",
"TE",
"Trailers",
"Transfer-Encoding",
"Upgrade",
}
for _, h := range hopByHopHeaders {
header.Del(h)
}
// Remove any headers specified in the Connection header
if connection := header.Get("Connection"); connection != "" {
for _, h := range strings.Split(connection, ",") {
header.Del(strings.TrimSpace(h))
}
header.Del("Connection")
}
}