2025-06-29 19:27:00 -04:00
|
|
|
package model
|
|
|
|
|
|
|
|
|
|
import (
|
|
|
|
|
"encoding/json"
|
|
|
|
|
"time"
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
type ContextKey string
|
|
|
|
|
|
|
|
|
|
const BodyBytesKey ContextKey = "bodyBytes"
|
|
|
|
|
|
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
|
|
|
// ProxySettings holds dynamic proxy configuration (persisted in DB)
|
|
|
|
|
type ProxySettings struct {
|
|
|
|
|
RequestHeaderRules []HeaderRule `json:"requestHeaderRules"`
|
|
|
|
|
ResponseHeaderRules []HeaderRule `json:"responseHeaderRules"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// HeaderRule defines an action to take on a specific header.
|
|
|
|
|
// Actions: "block" (remove), "set" (override value), "replace" (find & replace in value)
|
|
|
|
|
type HeaderRule struct {
|
|
|
|
|
Header string `json:"header"` // Header name (case-insensitive match)
|
|
|
|
|
Action string `json:"action"` // "block", "set", "replace"
|
|
|
|
|
Value string `json:"value,omitempty"` // For "set": the new value. For "replace": the replacement string.
|
|
|
|
|
Find string `json:"find,omitempty"` // For "replace": the string to find in the header value
|
|
|
|
|
Enabled bool `json:"enabled"` // Toggle without deleting
|
|
|
|
|
}
|
|
|
|
|
|
2025-06-29 19:27:00 -04:00
|
|
|
type PromptGrade struct {
|
|
|
|
|
Score int `json:"score"`
|
|
|
|
|
MaxScore int `json:"maxScore"`
|
|
|
|
|
Feedback string `json:"feedback"`
|
|
|
|
|
ImprovedPrompt string `json:"improvedPrompt"`
|
|
|
|
|
Criteria map[string]CriteriaScore `json:"criteria"`
|
|
|
|
|
GradingTimestamp string `json:"gradingTimestamp"`
|
|
|
|
|
IsProcessing bool `json:"isProcessing"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type CriteriaScore struct {
|
|
|
|
|
Score int `json:"score"`
|
|
|
|
|
Feedback string `json:"feedback"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type RequestLog struct {
|
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
|
|
|
RequestID string `json:"requestId"`
|
|
|
|
|
Timestamp string `json:"timestamp"`
|
|
|
|
|
Method string `json:"method"`
|
|
|
|
|
Endpoint string `json:"endpoint"`
|
|
|
|
|
Headers map[string][]string `json:"headers"`
|
|
|
|
|
Body interface{} `json:"body"`
|
|
|
|
|
Model string `json:"model,omitempty"`
|
|
|
|
|
OriginalModel string `json:"originalModel,omitempty"`
|
|
|
|
|
RoutedModel string `json:"routedModel,omitempty"`
|
|
|
|
|
UserAgent string `json:"userAgent"`
|
|
|
|
|
ContentType string `json:"contentType"`
|
|
|
|
|
PromptGrade *PromptGrade `json:"promptGrade,omitempty"`
|
|
|
|
|
Response *ResponseLog `json:"response,omitempty"`
|
|
|
|
|
ConversationHash string `json:"conversationHash,omitempty"`
|
|
|
|
|
MessageCount int `json:"messageCount,omitempty"`
|
|
|
|
|
OrganizationID string `json:"organizationId,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ResponseLog struct {
|
|
|
|
|
StatusCode int `json:"statusCode"`
|
|
|
|
|
Headers map[string][]string `json:"headers"`
|
2025-07-07 02:07:37 -04:00
|
|
|
Body json.RawMessage `json:"body,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
BodyText string `json:"bodyText,omitempty"`
|
2026-03-19 18:52:09 -06:00
|
|
|
StreamError string `json:"streamError,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
ResponseTime int64 `json:"responseTime"`
|
|
|
|
|
StreamingChunks []string `json:"streamingChunks,omitempty"`
|
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
|
|
|
ChunkTimings []ChunkTiming `json:"chunkTimings,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
IsStreaming bool `json:"isStreaming"`
|
|
|
|
|
CompletedAt string `json:"completedAt"`
|
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
|
|
|
RateLimit *RateLimitInfo `json:"rateLimit,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// ChunkTiming records when each SSE chunk arrived during streaming
|
|
|
|
|
type ChunkTiming struct {
|
|
|
|
|
Index int `json:"index"`
|
|
|
|
|
Timestamp string `json:"timestamp"`
|
|
|
|
|
ByteSize int `json:"byteSize"`
|
|
|
|
|
ElapsedMs int64 `json:"elapsedMs"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// RateLimitInfo captures rate limit / quota data from upstream response headers
|
|
|
|
|
type RateLimitInfo struct {
|
|
|
|
|
// Organization
|
|
|
|
|
OrganizationID string `json:"organizationId,omitempty"`
|
|
|
|
|
|
|
|
|
|
// Legacy per-resource rate limits
|
|
|
|
|
RequestsLimit int `json:"requestsLimit,omitempty"`
|
|
|
|
|
RequestsRemaining int `json:"requestsRemaining,omitempty"`
|
|
|
|
|
RequestsReset string `json:"requestsReset,omitempty"`
|
|
|
|
|
TokensLimit int `json:"tokensLimit,omitempty"`
|
|
|
|
|
TokensRemaining int `json:"tokensRemaining,omitempty"`
|
|
|
|
|
TokensReset string `json:"tokensReset,omitempty"`
|
|
|
|
|
|
|
|
|
|
// Unified quota system (Anthropic's current model)
|
|
|
|
|
UnifiedStatus string `json:"unifiedStatus,omitempty"`
|
|
|
|
|
UnifiedUtilization5h float64 `json:"unifiedUtilization5h,omitempty"`
|
|
|
|
|
UnifiedReset5h string `json:"unifiedReset5h,omitempty"`
|
|
|
|
|
UnifiedUtilization7d float64 `json:"unifiedUtilization7d,omitempty"`
|
|
|
|
|
UnifiedReset7d string `json:"unifiedReset7d,omitempty"`
|
|
|
|
|
UnifiedFallbackPercentage float64 `json:"unifiedFallbackPercentage,omitempty"`
|
|
|
|
|
UnifiedOverageStatus string `json:"unifiedOverageStatus,omitempty"`
|
|
|
|
|
UnifiedRepresentativeClaim string `json:"unifiedRepresentativeClaim,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ChatMessage struct {
|
|
|
|
|
Role string `json:"role"`
|
|
|
|
|
Content string `json:"content"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ChatCompletionRequest struct {
|
|
|
|
|
Model string `json:"model"`
|
|
|
|
|
Messages []ChatMessage `json:"messages"`
|
|
|
|
|
Stream bool `json:"stream,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
2025-07-07 02:07:37 -04:00
|
|
|
type AnthropicUsage struct {
|
|
|
|
|
InputTokens int `json:"input_tokens"`
|
|
|
|
|
OutputTokens int `json:"output_tokens"`
|
|
|
|
|
CacheCreationInputTokens int `json:"cache_creation_input_tokens,omitempty"`
|
|
|
|
|
CacheReadInputTokens int `json:"cache_read_input_tokens,omitempty"`
|
|
|
|
|
ServiceTier string `json:"service_tier,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
2025-07-07 02:07:37 -04:00
|
|
|
type AnthropicResponse struct {
|
|
|
|
|
Content []AnthropicContentBlock `json:"content"`
|
|
|
|
|
ID string `json:"id"`
|
|
|
|
|
Model string `json:"model"`
|
|
|
|
|
Role string `json:"role"`
|
|
|
|
|
StopReason string `json:"stop_reason"`
|
|
|
|
|
StopSequence *string `json:"stop_sequence"`
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
Usage AnthropicUsage `json:"usage"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type AnthropicContentBlock struct {
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
Text string `json:"text"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type AnthropicMessage struct {
|
|
|
|
|
Role string `json:"role"`
|
|
|
|
|
Content interface{} `json:"content"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (m *AnthropicMessage) GetContentBlocks() []AnthropicContentBlock {
|
|
|
|
|
switch v := m.Content.(type) {
|
|
|
|
|
case string:
|
|
|
|
|
return []AnthropicContentBlock{{Type: "text", Text: v}}
|
|
|
|
|
case []interface{}:
|
|
|
|
|
var blocks []AnthropicContentBlock
|
|
|
|
|
for _, item := range v {
|
|
|
|
|
if block, ok := item.(map[string]interface{}); ok {
|
|
|
|
|
if typ, hasType := block["type"].(string); hasType {
|
|
|
|
|
if text, hasText := block["text"].(string); hasText {
|
|
|
|
|
blocks = append(blocks, AnthropicContentBlock{Type: typ, Text: text})
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return blocks
|
|
|
|
|
case []AnthropicContentBlock:
|
|
|
|
|
return v
|
|
|
|
|
default:
|
|
|
|
|
return []AnthropicContentBlock{}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type AnthropicSystemMessage struct {
|
|
|
|
|
Text string `json:"text"`
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
CacheControl *CacheControl `json:"cache_control,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type CacheControl struct {
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type Tool struct {
|
|
|
|
|
Name string `json:"name"`
|
|
|
|
|
Description string `json:"description"`
|
|
|
|
|
InputSchema InputSchema `json:"input_schema"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type InputSchema struct {
|
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
|
|
|
Type interface{} `json:"type"`
|
2025-08-02 22:40:01 -04:00
|
|
|
Properties map[string]interface{} `json:"properties"`
|
|
|
|
|
Required []string `json:"required,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type AnthropicRequest struct {
|
|
|
|
|
Model string `json:"model"`
|
|
|
|
|
Messages []AnthropicMessage `json:"messages"`
|
|
|
|
|
MaxTokens int `json:"max_tokens"`
|
|
|
|
|
Temperature *float64 `json:"temperature,omitempty"`
|
|
|
|
|
System []AnthropicSystemMessage `json:"system,omitempty"`
|
|
|
|
|
Stream bool `json:"stream,omitempty"`
|
|
|
|
|
Tools []Tool `json:"tools,omitempty"`
|
2025-08-02 22:40:01 -04:00
|
|
|
ToolChoice interface{} `json:"tool_choice,omitempty"`
|
2025-06-29 19:27:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ModelsResponse struct {
|
|
|
|
|
Object string `json:"object"`
|
|
|
|
|
Data []ModelInfo `json:"data"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ModelInfo struct {
|
|
|
|
|
ID string `json:"id"`
|
|
|
|
|
Object string `json:"object"`
|
|
|
|
|
Created int64 `json:"created"`
|
|
|
|
|
OwnedBy string `json:"owned_by"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type GradeRequest struct {
|
|
|
|
|
Messages []AnthropicMessage `json:"messages"`
|
|
|
|
|
SystemMessages []AnthropicSystemMessage `json:"systemMessages"`
|
|
|
|
|
RequestID string `json:"requestId,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type HealthResponse struct {
|
|
|
|
|
Status string `json:"status"`
|
|
|
|
|
Timestamp time.Time `json:"timestamp"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ErrorResponse struct {
|
|
|
|
|
Error string `json:"error"`
|
|
|
|
|
Details string `json:"details,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
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
|
|
|
// UsageStats represents aggregated token usage statistics
|
|
|
|
|
type UsageStats struct {
|
|
|
|
|
TotalRequests int `json:"total_requests"`
|
|
|
|
|
TotalInputTokens int64 `json:"total_input_tokens"`
|
|
|
|
|
TotalOutputTokens int64 `json:"total_output_tokens"`
|
|
|
|
|
TotalCacheTokens int64 `json:"total_cache_tokens"`
|
|
|
|
|
RequestsByModel map[string]ModelStats `json:"requests_by_model"`
|
|
|
|
|
StartDate string `json:"start_date,omitempty"`
|
|
|
|
|
EndDate string `json:"end_date,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// ModelStats represents per-model usage statistics
|
|
|
|
|
type ModelStats struct {
|
|
|
|
|
RequestCount int `json:"request_count"`
|
|
|
|
|
InputTokens int64 `json:"input_tokens"`
|
|
|
|
|
OutputTokens int64 `json:"output_tokens"`
|
|
|
|
|
CacheTokens int64 `json:"cache_tokens"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// RequestSummary is a lightweight version of RequestLog for fast list views
|
|
|
|
|
type RequestSummary struct {
|
|
|
|
|
RequestID string `json:"requestId"`
|
|
|
|
|
Timestamp string `json:"timestamp"`
|
|
|
|
|
Method string `json:"method"`
|
|
|
|
|
Endpoint string `json:"endpoint"`
|
|
|
|
|
Model string `json:"model,omitempty"`
|
|
|
|
|
OriginalModel string `json:"originalModel,omitempty"`
|
|
|
|
|
RoutedModel string `json:"routedModel,omitempty"`
|
|
|
|
|
StatusCode int `json:"statusCode,omitempty"`
|
|
|
|
|
ResponseTime int64 `json:"responseTime,omitempty"`
|
|
|
|
|
Usage *AnthropicUsage `json:"usage,omitempty"`
|
|
|
|
|
ConversationHash string `json:"conversationHash,omitempty"`
|
|
|
|
|
MessageCount int `json:"messageCount,omitempty"`
|
|
|
|
|
StopReason string `json:"stopReason,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Dashboard stats structures
|
|
|
|
|
type DashboardStats struct {
|
|
|
|
|
DailyStats []DailyTokens `json:"dailyStats"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type HourlyStatsResponse struct {
|
|
|
|
|
HourlyStats []HourlyTokens `json:"hourlyStats"`
|
|
|
|
|
TodayTokens int64 `json:"todayTokens"`
|
|
|
|
|
TodayRequests int `json:"todayRequests"`
|
|
|
|
|
AvgResponseTime int64 `json:"avgResponseTime"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ModelStatsResponse struct {
|
|
|
|
|
ModelStats []ModelTokens `json:"modelStats"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type DailyTokens struct {
|
|
|
|
|
Date string `json:"date"`
|
|
|
|
|
Tokens int64 `json:"tokens"`
|
|
|
|
|
Requests int `json:"requests"`
|
|
|
|
|
Models map[string]DailyModelStat `json:"models,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type HourlyTokens struct {
|
|
|
|
|
Hour int `json:"hour"`
|
|
|
|
|
Label string `json:"label,omitempty"`
|
|
|
|
|
Tokens int64 `json:"tokens"`
|
|
|
|
|
Requests int `json:"requests"`
|
|
|
|
|
Models map[string]DailyModelStat `json:"models,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// DailyModelStat represents per-model stats for dashboard aggregation
|
|
|
|
|
type DailyModelStat struct {
|
|
|
|
|
Tokens int64 `json:"tokens"`
|
|
|
|
|
Requests int `json:"requests"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ModelTokens struct {
|
|
|
|
|
Model string `json:"model"`
|
|
|
|
|
Tokens int64 `json:"tokens"`
|
|
|
|
|
Requests int `json:"requests"`
|
|
|
|
|
}
|
|
|
|
|
|
2025-06-29 19:27:00 -04:00
|
|
|
type StreamingEvent struct {
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
Index *int `json:"index,omitempty"`
|
|
|
|
|
Delta *Delta `json:"delta,omitempty"`
|
|
|
|
|
ContentBlock *ContentBlock `json:"content_block,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type Delta struct {
|
|
|
|
|
Type string `json:"type,omitempty"`
|
|
|
|
|
Text string `json:"text,omitempty"`
|
|
|
|
|
Name string `json:"name,omitempty"`
|
|
|
|
|
Input json.RawMessage `json:"input,omitempty"`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ContentBlock struct {
|
|
|
|
|
Type string `json:"type"`
|
|
|
|
|
ID string `json:"id,omitempty"`
|
|
|
|
|
Name string `json:"name,omitempty"`
|
|
|
|
|
Input json.RawMessage `json:"input,omitempty"`
|
|
|
|
|
Text string `json:"text,omitempty"`
|
|
|
|
|
}
|