shishantbiswas/coder
1
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2026-06-02 20:48:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: remove 429 from aibridge circuit breaker failure conditions (#24701)

Browse Source 
## Description

Removes 429 (Too Many Requests) from the circuit breaker failure conditions. Rate limiting is now handled by automatic key failover instead of tripping the circuit breaker.

## Changes

`DefaultIsFailure` no longer treats 429 as a circuit breaker failure. The circuit breaker now only trips on server overload responses (503, 529).

Tests and integration tests updated to use 503 instead of 429 for tripping circuits. Description strings in deployment config updated to reflect the change.

Closes https://github.com/coder/internal/issues/1445

> [!NOTE]
> Initially generated by Coder Agents, modified and reviewed by @ssncferreira
This commit is contained in:
Susana Ferreira
2026-04-30 09:31:32 +01:00
committed by GitHub
parent 101a4082dd
commit dbb50ebaaf
14 changed files with 48 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files
aibridge/circuitbreaker/circuitbreaker.go
+7 -5
View File
@@ -20,13 +20,15 @@ import (
// and the request was rejected without calling the handler. // and the request was rejected without calling the handler.
var ErrCircuitOpen = xerrors.New("circuit breaker is open") var ErrCircuitOpen = xerrors.New("circuit breaker is open")
// DefaultIsFailure returns true for standard HTTP status codes that typically // DefaultIsFailure returns true for standard HTTP status codes that
// indicate upstream overload. // typically indicate upstream overload.
//
// Note: 429 (Too Many Requests) is intentionally excluded. Rate
// limits are key-specific and handled by automatic key failover.
func DefaultIsFailure(statusCode int) bool { func DefaultIsFailure(statusCode int) bool {
switch statusCode { switch statusCode {
case http.StatusTooManyRequests, // 429 case http.StatusServiceUnavailable, // 503
http.StatusServiceUnavailable, // 503 http.StatusGatewayTimeout: // 504
http.StatusGatewayTimeout: // 504
return true return true
default: default:
return false return false
aibridge/circuitbreaker/circuitbreaker_test.go
+6 -6
View File
@@ -32,11 +32,11 @@ func TestExecute_PerModelIsolation(t *testing.T) {
sonnetModel := "claude-sonnet-4-20250514" sonnetModel := "claude-sonnet-4-20250514"
haikuModel := "claude-3-5-haiku-20241022" haikuModel := "claude-3-5-haiku-20241022"
// Trip circuit on sonnet model (returns 429) // Trip circuit on sonnet model (returns 503)
w := httptest.NewRecorder() w := httptest.NewRecorder()
err := cbs.Execute(endpoint, sonnetModel, w, func(rw http.ResponseWriter) error { err := cbs.Execute(endpoint, sonnetModel, w, func(rw http.ResponseWriter) error {
sonnetCalls.Add(1) sonnetCalls.Add(1)
rw.WriteHeader(http.StatusTooManyRequests) rw.WriteHeader(http.StatusServiceUnavailable)
return nil return nil
}) })
assert.NoError(t, err) assert.NoError(t, err)
@@ -79,11 +79,11 @@ func TestExecute_PerEndpointIsolation(t *testing.T) {
model := "test-model" model := "test-model"
// Trip circuit on /v1/messages endpoint (returns 429) // Trip circuit on /v1/messages endpoint (returns 503)
w := httptest.NewRecorder() w := httptest.NewRecorder()
err := cbs.Execute("/v1/messages", model, w, func(rw http.ResponseWriter) error { err := cbs.Execute("/v1/messages", model, w, func(rw http.ResponseWriter) error {
messagesCalls.Add(1) messagesCalls.Add(1)
rw.WriteHeader(http.StatusTooManyRequests) rw.WriteHeader(http.StatusServiceUnavailable)
return nil return nil
}) })
assert.NoError(t, err) assert.NoError(t, err)
@@ -179,7 +179,7 @@ func TestExecute_OnStateChange(t *testing.T) {
// Trip circuit // Trip circuit
w := httptest.NewRecorder() w := httptest.NewRecorder()
err := cbs.Execute(endpoint, model, w, func(rw http.ResponseWriter) error { err := cbs.Execute(endpoint, model, w, func(rw http.ResponseWriter) error {
rw.WriteHeader(http.StatusTooManyRequests) rw.WriteHeader(http.StatusServiceUnavailable)
return nil return nil
}) })
assert.NoError(t, err) assert.NoError(t, err)
@@ -202,7 +202,7 @@ func TestDefaultIsFailure(t *testing.T) {
{http.StatusOK, false}, {http.StatusOK, false},
{http.StatusBadRequest, false}, {http.StatusBadRequest, false},
{http.StatusUnauthorized, false}, {http.StatusUnauthorized, false},
{http.StatusTooManyRequests, true}, // 429 {http.StatusTooManyRequests, false}, // 429: handled by key failover, not circuit breaker
{http.StatusInternalServerError, false}, {http.StatusInternalServerError, false},
{http.StatusBadGateway, false}, {http.StatusBadGateway, false},
{http.StatusServiceUnavailable, true}, // 503 {http.StatusServiceUnavailable, true}, // 503
aibridge/internal/integrationtest/circuit_breaker_test.go
+21 -21
View File
@@ -23,8 +23,8 @@ import (
// Common response bodies for circuit breaker tests. // Common response bodies for circuit breaker tests.
const ( const (
anthropicRateLimitError = `{"type":"error","error":{"type":"rate_limit_error","message":"rate limited"}}` anthropicOverloadedError = `{"type":"error","error":{"type":"api_error","message":"Internal server error"}}`
openAIRateLimitError = `{"error":{"type":"rate_limit_error","message":"rate limited","code":"rate_limit_exceeded"}}` openAIOverloadedError = `{"error":{"message":"Service Unavailable.","type":"cf_service_unavailable","code":503}}`
) )
func anthropicSuccessResponse(model string) string { func anthropicSuccessResponse(model string) string {
@@ -59,7 +59,7 @@ func TestCircuitBreaker_FullRecoveryCycle(t *testing.T) {
expectProvider: config.ProviderAnthropic, expectProvider: config.ProviderAnthropic,
expectEndpoint: "/v1/messages", expectEndpoint: "/v1/messages",
expectModel: "claude-sonnet-4-20250514", expectModel: "claude-sonnet-4-20250514",
errorBody: anthropicRateLimitError, errorBody: anthropicOverloadedError,
successBody: anthropicSuccessResponse("claude-sonnet-4-20250514"), successBody: anthropicSuccessResponse("claude-sonnet-4-20250514"),
requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{ headers: http.Header{
@@ -80,7 +80,7 @@ func TestCircuitBreaker_FullRecoveryCycle(t *testing.T) {
expectProvider: config.ProviderOpenAI, expectProvider: config.ProviderOpenAI,
expectEndpoint: "/v1/chat/completions", expectEndpoint: "/v1/chat/completions",
expectModel: "gpt-4o", expectModel: "gpt-4o",
errorBody: openAIRateLimitError, errorBody: openAIOverloadedError,
successBody: openAISuccessResponse("gpt-4o"), successBody: openAISuccessResponse("gpt-4o"),
requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{"Authorization": {"Bearer test-key"}}, headers: http.Header{"Authorization": {"Bearer test-key"}},
@@ -103,14 +103,14 @@ func TestCircuitBreaker_FullRecoveryCycle(t *testing.T) {
var shouldFail atomic.Bool var shouldFail atomic.Bool
shouldFail.Store(true) shouldFail.Store(true)
// Mock upstream that returns 429 or 200 based on shouldFail flag. // Mock upstream that returns 503 or 200 based on shouldFail flag.
// x-should-retry: false is required to disable SDK automatic retries (default MaxRetries=2). // x-should-retry: false is required to disable SDK automatic retries (default MaxRetries=2).
mockUpstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { mockUpstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
upstreamCalls.Add(1) upstreamCalls.Add(1)
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
w.Header().Set("x-should-retry", "false") w.Header().Set("x-should-retry", "false")
if shouldFail.Load() { if shouldFail.Load() {
w.WriteHeader(http.StatusTooManyRequests) w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(tc.errorBody)) _, _ = w.Write([]byte(tc.errorBody))
} else { } else {
w.WriteHeader(http.StatusOK) w.WriteHeader(http.StatusOK)
@@ -146,10 +146,10 @@ func TestCircuitBreaker_FullRecoveryCycle(t *testing.T) {
} }
// Phase 1: Trip the circuit breaker // Phase 1: Trip the circuit breaker
// First FailureThreshold requests hit upstream, get 429 // First FailureThreshold requests hit upstream, get 503
for i := uint32(0); i < cbConfig.FailureThreshold; i++ { for i := uint32(0); i < cbConfig.FailureThreshold; i++ {
status := doRequest() status := doRequest()
assert.Equal(t, http.StatusTooManyRequests, status) assert.Equal(t, http.StatusServiceUnavailable, status)
} }
//nolint:gosec // G115: test constant, no overflow risk //nolint:gosec // G115: test constant, no overflow risk
assert.Equal(t, int32(cbConfig.FailureThreshold), upstreamCalls.Load()) assert.Equal(t, int32(cbConfig.FailureThreshold), upstreamCalls.Load())
@@ -227,7 +227,7 @@ func TestCircuitBreaker_HalfOpenFailure(t *testing.T) {
expectProvider: config.ProviderAnthropic, expectProvider: config.ProviderAnthropic,
expectEndpoint: "/v1/messages", expectEndpoint: "/v1/messages",
expectModel: "claude-sonnet-4-20250514", expectModel: "claude-sonnet-4-20250514",
errorBody: anthropicRateLimitError, errorBody: anthropicOverloadedError,
requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{ headers: http.Header{
"x-api-key": {"test"}, "x-api-key": {"test"},
@@ -247,7 +247,7 @@ func TestCircuitBreaker_HalfOpenFailure(t *testing.T) {
expectProvider: config.ProviderOpenAI, expectProvider: config.ProviderOpenAI,
expectEndpoint: "/v1/chat/completions", expectEndpoint: "/v1/chat/completions",
expectModel: "gpt-4o", expectModel: "gpt-4o",
errorBody: openAIRateLimitError, errorBody: openAIOverloadedError,
requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{"Authorization": {"Bearer test-key"}}, headers: http.Header{"Authorization": {"Bearer test-key"}},
path: pathOpenAIChatCompletions, path: pathOpenAIChatCompletions,
@@ -267,12 +267,12 @@ func TestCircuitBreaker_HalfOpenFailure(t *testing.T) {
var upstreamCalls atomic.Int32 var upstreamCalls atomic.Int32
// Mock upstream that always returns 429. // Mock upstream that always returns 503.
mockUpstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { mockUpstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
upstreamCalls.Add(1) upstreamCalls.Add(1)
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
w.Header().Set("x-should-retry", "false") w.Header().Set("x-should-retry", "false")
w.WriteHeader(http.StatusTooManyRequests) w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(tc.errorBody)) _, _ = w.Write([]byte(tc.errorBody))
})) }))
defer mockUpstream.Close() defer mockUpstream.Close()
@@ -305,7 +305,7 @@ func TestCircuitBreaker_HalfOpenFailure(t *testing.T) {
// Phase 1: Trip the circuit // Phase 1: Trip the circuit
for i := uint32(0); i < cbConfig.FailureThreshold; i++ { for i := uint32(0); i < cbConfig.FailureThreshold; i++ {
status := doRequest() status := doRequest()
assert.Equal(t, http.StatusTooManyRequests, status) assert.Equal(t, http.StatusServiceUnavailable, status)
} }
// Verify circuit is open // Verify circuit is open
@@ -321,7 +321,7 @@ func TestCircuitBreaker_HalfOpenFailure(t *testing.T) {
// Phase 3: Request in half-open state fails, circuit should re-open // Phase 3: Request in half-open state fails, circuit should re-open
upstreamCallsBefore := upstreamCalls.Load() upstreamCallsBefore := upstreamCalls.Load()
status = doRequest() status = doRequest()
assert.Equal(t, http.StatusTooManyRequests, status, "Request should fail in half-open state") assert.Equal(t, http.StatusServiceUnavailable, status, "Request should fail in half-open state")
assert.Equal(t, upstreamCallsBefore+1, upstreamCalls.Load(), "Request should reach upstream in half-open state") assert.Equal(t, upstreamCallsBefore+1, upstreamCalls.Load(), "Request should reach upstream in half-open state")
// Circuit should be open again - next request should be rejected immediately // Circuit should be open again - next request should be rejected immediately
@@ -363,7 +363,7 @@ func TestCircuitBreaker_HalfOpenMaxRequests(t *testing.T) {
expectProvider: config.ProviderAnthropic, expectProvider: config.ProviderAnthropic,
expectEndpoint: "/v1/messages", expectEndpoint: "/v1/messages",
expectModel: "claude-sonnet-4-20250514", expectModel: "claude-sonnet-4-20250514",
errorBody: anthropicRateLimitError, errorBody: anthropicOverloadedError,
successBody: anthropicSuccessResponse("claude-sonnet-4-20250514"), successBody: anthropicSuccessResponse("claude-sonnet-4-20250514"),
requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{ headers: http.Header{
@@ -384,7 +384,7 @@ func TestCircuitBreaker_HalfOpenMaxRequests(t *testing.T) {
expectProvider: config.ProviderOpenAI, expectProvider: config.ProviderOpenAI,
expectEndpoint: "/v1/chat/completions", expectEndpoint: "/v1/chat/completions",
expectModel: "gpt-4o", expectModel: "gpt-4o",
errorBody: openAIRateLimitError, errorBody: openAIOverloadedError,
successBody: openAISuccessResponse("gpt-4o"), successBody: openAISuccessResponse("gpt-4o"),
requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`, requestBody: `{"model":"gpt-4o","messages":[{"role":"user","content":"hi"}]}`,
headers: http.Header{"Authorization": {"Bearer test-key"}}, headers: http.Header{"Authorization": {"Bearer test-key"}},
@@ -413,7 +413,7 @@ func TestCircuitBreaker_HalfOpenMaxRequests(t *testing.T) {
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
w.Header().Set("x-should-retry", "false") w.Header().Set("x-should-retry", "false")
if shouldFail.Load() { if shouldFail.Load() {
w.WriteHeader(http.StatusTooManyRequests) w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(tc.errorBody)) _, _ = w.Write([]byte(tc.errorBody))
} else { } else {
// Slow response to ensure requests overlap // Slow response to ensure requests overlap
@@ -453,7 +453,7 @@ func TestCircuitBreaker_HalfOpenMaxRequests(t *testing.T) {
// Phase 1: Trip the circuit // Phase 1: Trip the circuit
for i := uint32(0); i < cbConfig.FailureThreshold; i++ { for i := uint32(0); i < cbConfig.FailureThreshold; i++ {
status := doRequest() status := doRequest()
assert.Equal(t, http.StatusTooManyRequests, status) assert.Equal(t, http.StatusServiceUnavailable, status)
} }
// Verify circuit is open // Verify circuit is open
@@ -529,8 +529,8 @@ func TestCircuitBreaker_PerModelIsolation(t *testing.T) {
if strings.Contains(string(body), "claude-sonnet-4-20250514") { if strings.Contains(string(body), "claude-sonnet-4-20250514") {
sonnetCalls.Add(1) sonnetCalls.Add(1)
if sonnetShouldFail.Load() { if sonnetShouldFail.Load() {
w.WriteHeader(http.StatusTooManyRequests) w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(anthropicRateLimitError)) _, _ = w.Write([]byte(anthropicOverloadedError))
} else { } else {
w.WriteHeader(http.StatusOK) w.WriteHeader(http.StatusOK)
_, _ = w.Write([]byte(anthropicSuccessResponse("claude-sonnet-4-20250514"))) _, _ = w.Write([]byte(anthropicSuccessResponse("claude-sonnet-4-20250514")))
@@ -578,7 +578,7 @@ func TestCircuitBreaker_PerModelIsolation(t *testing.T) {
// Phase 1: Trip the circuit for sonnet model // Phase 1: Trip the circuit for sonnet model
for i := uint32(0); i < cbConfig.FailureThreshold; i++ { for i := uint32(0); i < cbConfig.FailureThreshold; i++ {
status := doRequest("claude-sonnet-4-20250514") status := doRequest("claude-sonnet-4-20250514")
assert.Equal(t, http.StatusTooManyRequests, status) assert.Equal(t, http.StatusServiceUnavailable, status)
} }
//nolint:gosec // G115: test constant, no overflow risk //nolint:gosec // G115: test constant, no overflow risk
assert.Equal(t, int32(cbConfig.FailureThreshold), sonnetCalls.Load()) assert.Equal(t, int32(cbConfig.FailureThreshold), sonnetCalls.Load())
aibridge/provider/anthropic_test.go
+1 -1
View File
@@ -360,7 +360,7 @@ func Test_anthropicIsFailure(t *testing.T) {
{http.StatusOK, false}, {http.StatusOK, false},
{http.StatusBadRequest, false}, {http.StatusBadRequest, false},
{http.StatusUnauthorized, false}, {http.StatusUnauthorized, false},
{http.StatusTooManyRequests, true}, // 429 {http.StatusTooManyRequests, false}, // 429: handled by key failover, not circuit breaker
{http.StatusInternalServerError, false}, {http.StatusInternalServerError, false},
{http.StatusBadGateway, false}, {http.StatusBadGateway, false},
{http.StatusServiceUnavailable, true}, // 503 {http.StatusServiceUnavailable, true}, // 503
cli/testdata/coder_server_--help.golden
Vendored
+1 -1
View File
@@ -138,7 +138,7 @@ AI BRIDGE OPTIONS:
--aibridge-circuit-breaker-enabled bool, $CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED (default: false) --aibridge-circuit-breaker-enabled bool, $CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED (default: false)
Enable the circuit breaker to protect against cascading failures from Enable the circuit breaker to protect against cascading failures from
upstream AI provider rate limits (429, 503, 529 overloaded). upstream AI provider overload (503, 529).
--aibridge-retention duration, $CODER_AIBRIDGE_RETENTION (default: 60d) --aibridge-retention duration, $CODER_AIBRIDGE_RETENTION (default: 60d)
Length of time to retain data such as interceptions and all related Length of time to retain data such as interceptions and all related
cli/testdata/server-config.yaml.golden
Vendored
+1 -1
View File
@@ -821,7 +821,7 @@ aibridge:
# (default: true, type: bool) # (default: true, type: bool)
allow_byok: true allow_byok: true
# Enable the circuit breaker to protect against cascading failures from upstream # Enable the circuit breaker to protect against cascading failures from upstream
# AI provider rate limits (429, 503, 529 overloaded). # AI provider overload (503, 529).
# (default: false, type: bool) # (default: false, type: bool)
circuit_breaker_enabled: false circuit_breaker_enabled: false
# Number of consecutive failures that triggers the circuit breaker to open. # Number of consecutive failures that triggers the circuit breaker to open.
coderd/apidoc/docs.go
Generated
+1 -1
View File
@@ -13138,7 +13138,7 @@ const docTemplate = `{
] ]
}, },
"circuit_breaker_enabled": { "circuit_breaker_enabled": {
"description": "Circuit breaker protects against cascading failures from upstream AI\nprovider rate limits (429, 503, 529 overloaded).", "description": "Circuit breaker protects against cascading failures from upstream AI\nprovider overload (503, 529).",
"type": "boolean" "type": "boolean"
}, },
"circuit_breaker_failure_threshold": { "circuit_breaker_failure_threshold": {
coderd/apidoc/swagger.json
Generated
+1 -1
View File
@@ -11686,7 +11686,7 @@
] ]
}, },
"circuit_breaker_enabled": { "circuit_breaker_enabled": {
"description": "Circuit breaker protects against cascading failures from upstream AI\nprovider rate limits (429, 503, 529 overloaded).", "description": "Circuit breaker protects against cascading failures from upstream AI\nprovider overload (503, 529).",
"type": "boolean" "type": "boolean"
}, },
"circuit_breaker_failure_threshold": { "circuit_breaker_failure_threshold": {
codersdk/deployment.go
+2 -2
View File
@@ -3823,7 +3823,7 @@ Write out the current server config as YAML to stdout.`,
}, },
{ {
Name: "AI Bridge Circuit Breaker Enabled", Name: "AI Bridge Circuit Breaker Enabled",
Description: "Enable the circuit breaker to protect against cascading failures from upstream AI provider rate limits (429, 503, 529 overloaded).", Description: "Enable the circuit breaker to protect against cascading failures from upstream AI provider overload (503, 529).",
Flag: "aibridge-circuit-breaker-enabled", Flag: "aibridge-circuit-breaker-enabled",
Env: "CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED", Env: "CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED",
Value: &c.AI.BridgeConfig.CircuitBreakerEnabled, Value: &c.AI.BridgeConfig.CircuitBreakerEnabled,
@@ -4074,7 +4074,7 @@ type AIBridgeConfig struct {
SendActorHeaders serpent.Bool `json:"send_actor_headers" typescript:",notnull"` SendActorHeaders serpent.Bool `json:"send_actor_headers" typescript:",notnull"`
AllowBYOK serpent.Bool `json:"allow_byok" typescript:",notnull"` AllowBYOK serpent.Bool `json:"allow_byok" typescript:",notnull"`
// Circuit breaker protects against cascading failures from upstream AI // Circuit breaker protects against cascading failures from upstream AI
// provider rate limits (429, 503, 529 overloaded). // provider overload (503, 529).
CircuitBreakerEnabled serpent.Bool `json:"circuit_breaker_enabled" typescript:",notnull"` CircuitBreakerEnabled serpent.Bool `json:"circuit_breaker_enabled" typescript:",notnull"`
CircuitBreakerFailureThreshold serpent.Int64 `json:"circuit_breaker_failure_threshold" typescript:",notnull"` CircuitBreakerFailureThreshold serpent.Int64 `json:"circuit_breaker_failure_threshold" typescript:",notnull"`
CircuitBreakerInterval serpent.Duration `json:"circuit_breaker_interval" typescript:",notnull"` CircuitBreakerInterval serpent.Duration `json:"circuit_breaker_interval" typescript:",notnull"`
docs/reference/api/schemas.md
Generated
+1 -1
View File
@@ -487,7 +487,7 @@
| `allow_byok` | boolean | false | | | | `allow_byok` | boolean | false | | |
| `anthropic` | [codersdk.AIBridgeAnthropicConfig](#codersdkaibridgeanthropicconfig) | false | | Deprecated: Use Providers with indexed CODER_AIBRIDGE_PROVIDER_<N>_* env vars instead. | | `anthropic` | [codersdk.AIBridgeAnthropicConfig](#codersdkaibridgeanthropicconfig) | false | | Deprecated: Use Providers with indexed CODER_AIBRIDGE_PROVIDER_<N>_* env vars instead. |
| `bedrock` | [codersdk.AIBridgeBedrockConfig](#codersdkaibridgebedrockconfig) | false | | Deprecated: Use Providers with indexed CODER_AIBRIDGE_PROVIDER_<N>_* env vars instead. | | `bedrock` | [codersdk.AIBridgeBedrockConfig](#codersdkaibridgebedrockconfig) | false | | Deprecated: Use Providers with indexed CODER_AIBRIDGE_PROVIDER_<N>_* env vars instead. |
| `circuit_breaker_enabled` | boolean | false | | Circuit breaker protects against cascading failures from upstream AI provider rate limits (429, 503, 529 overloaded). | | `circuit_breaker_enabled` | boolean | false | | Circuit breaker protects against cascading failures from upstream AI provider overload (503, 529). |
| `circuit_breaker_failure_threshold` | integer | false | | | | `circuit_breaker_failure_threshold` | integer | false | | |
| `circuit_breaker_interval` | integer | false | | | | `circuit_breaker_interval` | integer | false | | |
| `circuit_breaker_max_requests` | integer | false | | | | `circuit_breaker_max_requests` | integer | false | | |
docs/reference/cli/server.md
Generated
+1 -1
View File
@@ -1899,7 +1899,7 @@ Allow users to provide their own LLM API keys or subscriptions. When disabled, o
| YAML | <code>aibridge.circuit_breaker_enabled</code> | | YAML | <code>aibridge.circuit_breaker_enabled</code> |
| Default | <code>false</code> | | Default | <code>false</code> |
Enable the circuit breaker to protect against cascading failures from upstream AI provider rate limits (429, 503, 529 overloaded). Enable the circuit breaker to protect against cascading failures from upstream AI provider overload (503, 529).
### --aibridge-proxy-enabled ### --aibridge-proxy-enabled
enterprise/aibridged/aibridged_integration_test.go
+3 -3
View File
@@ -438,13 +438,13 @@ func TestIntegrationCircuitBreaker(t *testing.T) {
registry := prometheus.NewRegistry() registry := prometheus.NewRegistry()
metrics := aibridge.NewMetrics(registry) metrics := aibridge.NewMetrics(registry)
// Set up mock OpenAI server that always returns 429 Too Many Requests. // Set up mock OpenAI server that always returns 503 Service Unavailable.
mockOpenAI := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { mockOpenAI := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
// Disable SDK retries. // Disable SDK retries.
w.Header().Set("x-should-retry", "false") w.Header().Set("x-should-retry", "false")
w.WriteHeader(http.StatusTooManyRequests) w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(`{"error":{"type":"rate_limit_error","message":"rate limited","code":"rate_limit_exceeded"}}`)) _, _ = w.Write([]byte(`{"error":{"message":"Service Unavailable.","type":"cf_service_unavailable","code":503}}`))
})) }))
t.Cleanup(mockOpenAI.Close) t.Cleanup(mockOpenAI.Close)
enterprise/cli/testdata/coder_server_--help.golden
Vendored
+1 -1
View File
@@ -139,7 +139,7 @@ AI BRIDGE OPTIONS:
--aibridge-circuit-breaker-enabled bool, $CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED (default: false) --aibridge-circuit-breaker-enabled bool, $CODER_AIBRIDGE_CIRCUIT_BREAKER_ENABLED (default: false)
Enable the circuit breaker to protect against cascading failures from Enable the circuit breaker to protect against cascading failures from
upstream AI provider rate limits (429, 503, 529 overloaded). upstream AI provider overload (503, 529).
--aibridge-retention duration, $CODER_AIBRIDGE_RETENTION (default: 60d) --aibridge-retention duration, $CODER_AIBRIDGE_RETENTION (default: 60d)
Length of time to retain data such as interceptions and all related Length of time to retain data such as interceptions and all related
site/src/api/typesGenerated.ts
Generated
+1 -1
View File
@@ -70,7 +70,7 @@ export interface AIBridgeConfig {
readonly allow_byok: boolean; readonly allow_byok: boolean;
/** /**
* Circuit breaker protects against cascading failures from upstream AI * Circuit breaker protects against cascading failures from upstream AI
* provider rate limits (429, 503, 529 overloaded). * provider overload (503, 529).
*/ */
readonly circuit_breaker_enabled: boolean; readonly circuit_breaker_enabled: boolean;
readonly circuit_breaker_failure_threshold: number; readonly circuit_breaker_failure_threshold: number;