doc: fix some typos in comment

.claude/skills/docs-writer/SKILL.md

@@ -176,7 +176,7 @@ grep -n 'func.*create\|func.*update\|func.*delete\|func.*get' transports/bifrost
 | `plugins.go` | `/api/plugins` | CRUD plugins |
 | `config.go` | `/api/config` | GET/PUT config |
 | `config.go` | `/api/proxy-config` | GET/PUT proxy config |
-| `cache.go` | `/api/cache/clear/{requestId}` | DELETE cache |
+| `cache.go` | `/api/cache/clear/{cacheId}` | DELETE cache |
 | `session.go` | `/api/session/*` | Login/logout/auth check |
 | `oauth2.go` | `/api/oauth/*` | OAuth callback/status |
 

.github/workflows/release-pipeline.yml

@@ -1771,7 +1771,7 @@ jobs:
             production.cloudflare.docker.com:443
             proxy.golang.org:443
             registry-1.docker.io:443
-            registry.access.redhat.com:443
+            *.redhat.com:443
             *.quay.io:443
             registry.npmjs.org:443
             storage.googleapis.com:443
@@ -1862,7 +1862,7 @@ jobs:
             production.cloudflare.docker.com:443
             proxy.golang.org:443
             registry-1.docker.io:443
-            registry.access.redhat.com:443
+            *.redhat.com:443
             *.quay.io:443
             registry.npmjs.org:443
             storage.googleapis.com:443

.github/workflows/scripts/validate-helm-config-fields.sh

@@ -637,7 +637,6 @@ bifrost:
         cache_by_model: true
         cache_by_provider: false
         exclude_system_prompt: true
-        cleanup_on_shutdown: true
         vector_store_namespace: "bifrost-cache"
     otel:
       enabled: true
@@ -710,7 +709,6 @@ assert_field_value 'plugins: semantic_cache conversation_history_threshold' '.pl
 assert_field_value 'plugins: semantic_cache cache_by_model' '.plugins.[4].config.cache_by_model' 'true'
 assert_field_value 'plugins: semantic_cache cache_by_provider' '.plugins.[4].config.cache_by_provider' 'false'
 assert_field_value 'plugins: semantic_cache exclude_system_prompt' '.plugins.[4].config.exclude_system_prompt' 'true'
-assert_field_value 'plugins: semantic_cache cleanup_on_shutdown' '.plugins.[4].config.cleanup_on_shutdown' 'true'
 assert_field_value 'plugins: semantic_cache vector_store_namespace' '.plugins.[4].config.vector_store_namespace' '"bifrost-cache"'
 
 # OTEL plugin

.gitignore

@@ -1,6 +1,7 @@
 .env
 .vscode
 .DS_Store
+.tool-versions
 *_creds*
 **/venv/
 **/__pycache__/**
@@ -45,6 +46,7 @@ transports/schema/config.schema.json
 *.db
 *.db-shm
 *.db-wal
+transports/bifrost-http/v1.5.x
 
 # Test reports
 test-reports
@@ -173,3 +175,7 @@ ui/app/routeTree.gen.ts
 .next
 
 .infisical
+
+# e2e test artifacts
+examples/mcps/auth-demo-server/auth-demo-server
+examples/mcps/oauth-demo-server/oauth-demo-server

core/bifrost.go

@@ -4051,6 +4051,31 @@ func (bifrost *Bifrost) SelectKeyForProviderRequestType(ctx *schemas.BifrostCont
 	return bifrost.keySelector(ctx, supportedKeys, providerKey, model)
 }
 
+// ComputeRawStorageForProvider determines whether raw request/response payloads should be
+// captured and stored in log records for the given provider. This is the same computation
+// performed inside executeRequest (lines 5675-5713), exported for callers that bypass
+// the normal inference path (e.g. realtime WebSocket/WebRTC sessions).
+func (bifrost *Bifrost) ComputeRawStorageForProvider(ctx *schemas.BifrostContext, providerKey schemas.ModelProvider) bool {
+	if ctx == nil {
+		ctx = bifrost.ctx
+	}
+	if ctx == nil {
+		return false
+	}
+	config, err := bifrost.account.GetConfigForProvider(providerKey)
+	if err != nil || config == nil {
+		return false
+	}
+	effectiveStore := config.StoreRawRequestResponse
+	allowStorageOverride, _ := ctx.Value(schemas.BifrostContextKeyAllowPerRequestStorageOverride).(bool)
+	if allowStorageOverride {
+		if override, ok := ctx.Value(schemas.BifrostContextKeyStoreRawRequestResponse).(bool); ok {
+			effectiveStore = override
+		}
+	}
+	return effectiveStore
+}
+
 // WSStreamHooks holds the post-hook runner and cleanup function returned by RunStreamPreHooks.
 // Call PostHookRunner for each streaming chunk, setting StreamEndIndicator on the final chunk.
 // Call Cleanup when done to release the pipeline back to the pool.

core/changelog.md

@@ -0,0 +1,2 @@
+[fix]: openai provider - add usage to completed event in responses to chat completions fallback [@kevinpdev](https://github.com/kevinpdev)
+[feat]: use chat completions for openai custom providers that disable responses [@kevinpdev](https://github.com/kevinpdev)

core/internal/llmtests/realtime.go

@@ -49,7 +49,10 @@ func RunRealtimeTest(t *testing.T, client *bifrost.Bifrost, ctx context.Context,
 		}
 
 		wsURL := rtProvider.RealtimeWebSocketURL(key, testConfig.RealtimeModel)
-		hdrs := rtProvider.RealtimeHeaders(key)
+		hdrs, headerErr := rtProvider.RealtimeHeaders(bfCtx, key)
+		if headerErr != nil {
+			t.Fatalf("failed to build realtime headers for provider %s: %v", testConfig.Provider, headerErr)
+		}
 
 		httpHeaders := http.Header{}
 		for k, v := range hdrs {

core/providers/anthropic/anthropic.go

@@ -564,8 +564,6 @@ func (provider *AnthropicProvider) ChatCompletionStream(ctx *schemas.BifrostCont
 		headers["x-api-key"] = key.Value.GetValue()
 	}
 
-	providerUtils.SetStreamIdleTimeoutIfEmpty(ctx, provider.networkConfig.StreamIdleTimeoutInSeconds)
-
 	// Use shared Anthropic streaming logic
 	return HandleAnthropicChatCompletionStreaming(
 		ctx,
@@ -574,6 +572,7 @@ func (provider *AnthropicProvider) ChatCompletionStream(ctx *schemas.BifrostCont
 		jsonData,
 		headers,
 		provider.networkConfig.ExtraHeaders,
+		provider.networkConfig.StreamIdleTimeoutInSeconds,
 		provider.networkConfig.BetaHeaderOverrides,
 		providerUtils.ShouldSendBackRawRequest(ctx, provider.sendBackRawRequest),
 		providerUtils.ShouldSendBackRawResponse(ctx, provider.sendBackRawResponse),
@@ -594,6 +593,7 @@ func HandleAnthropicChatCompletionStreaming(
 	jsonBody []byte,
 	headers map[string]string,
 	extraHeaders map[string]string,
+	streamIdleTimeoutInSeconds int,
 	betaHeaderOverrides map[string]bool,
 	sendBackRawRequest bool,
 	sendBackRawResponse bool,
@@ -603,6 +603,7 @@ func HandleAnthropicChatCompletionStreaming(
 	logger schemas.Logger,
 	postHookSpanFinalizer func(context.Context),
 ) (chan *schemas.BifrostStreamChunk, *schemas.BifrostError) {
+	providerUtils.SetStreamIdleTimeoutIfEmpty(ctx, streamIdleTimeoutInSeconds)
 	req := fasthttp.AcquireRequest()
 	resp := fasthttp.AcquireResponse()
 	resp.StreamBody = true // Initialize for streaming
@@ -636,7 +637,7 @@ func HandleAnthropicChatCompletionStreaming(
 		providerUtils.DrainLargePayloadRemainder(ctx)
 	}
 	if err != nil {
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		if errors.Is(err, context.Canceled) {
 			return nil, providerUtils.EnrichError(ctx, &schemas.BifrostError{
 				IsBifrostError: false,
@@ -658,7 +659,7 @@ func HandleAnthropicChatCompletionStreaming(
 
 	// Check for HTTP errors
 	if resp.StatusCode() != fasthttp.StatusOK {
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		return nil, providerUtils.EnrichError(ctx, parseAnthropicError(resp), jsonBody, nil, sendBackRawRequest, sendBackRawResponse)
 	}
 
@@ -683,7 +684,7 @@ func HandleAnthropicChatCompletionStreaming(
 			}
 			close(responseChan)
 		}()
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 
 		if resp.BodyStream() == nil {
 			bifrostErr := providerUtils.NewBifrostOperationError(
@@ -739,6 +740,10 @@ func HandleAnthropicChatCompletionStreaming(
 			}
 			eventType, eventDataBytes, readErr := sseReader.ReadEvent()
 			if readErr != nil {
+				// Recheck context cancellation
+				if ctx.Err() != nil {
+					return
+				}
 				if readErr != io.EOF {
 					ctx.SetValue(schemas.BifrostContextKeyStreamEndIndicator, true)
 					logger.Warn("Error reading %s stream: %v", providerName, readErr)
@@ -1030,15 +1035,14 @@ func (provider *AnthropicProvider) ResponsesStream(ctx *schemas.BifrostContext,
 		headers["x-api-key"] = key.Value.GetValue()
 	}
 
-	providerUtils.SetStreamIdleTimeoutIfEmpty(ctx, provider.networkConfig.StreamIdleTimeoutInSeconds)
-
 	return HandleAnthropicResponsesStream(
 		ctx,
 		provider.streamingClient,
 		provider.buildRequestURL(ctx, "/v1/messages", schemas.ResponsesStreamRequest),
 		jsonBody,
 		headers,
 		provider.networkConfig.ExtraHeaders,
+		provider.networkConfig.StreamIdleTimeoutInSeconds,
 		provider.networkConfig.BetaHeaderOverrides,
 		providerUtils.ShouldSendBackRawRequest(ctx, provider.sendBackRawRequest),
 		providerUtils.ShouldSendBackRawResponse(ctx, provider.sendBackRawResponse),
@@ -1059,6 +1063,7 @@ func HandleAnthropicResponsesStream(
 	jsonBody []byte,
 	headers map[string]string,
 	extraHeaders map[string]string,
+	streamIdleTimeoutInSeconds int,
 	betaHeaderOverrides map[string]bool,
 	sendBackRawRequest bool,
 	sendBackRawResponse bool,
@@ -1068,6 +1073,7 @@ func HandleAnthropicResponsesStream(
 	logger schemas.Logger,
 	postHookSpanFinalizer func(context.Context),
 ) (chan *schemas.BifrostStreamChunk, *schemas.BifrostError) {
+	providerUtils.SetStreamIdleTimeoutIfEmpty(ctx, streamIdleTimeoutInSeconds)
 	req := fasthttp.AcquireRequest()
 	resp := fasthttp.AcquireResponse()
 	resp.StreamBody = true
@@ -1103,7 +1109,7 @@ func HandleAnthropicResponsesStream(
 		providerUtils.DrainLargePayloadRemainder(ctx)
 	}
 	if err != nil {
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		if errors.Is(err, context.Canceled) {
 			return nil, providerUtils.EnrichError(ctx, &schemas.BifrostError{
 				IsBifrostError: false,
@@ -1125,7 +1131,7 @@ func HandleAnthropicResponsesStream(
 
 	// Check for HTTP errors
 	if resp.StatusCode() != fasthttp.StatusOK {
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		return nil, providerUtils.EnrichError(ctx, parseAnthropicError(resp), jsonBody, nil, sendBackRawRequest, sendBackRawResponse)
 	}
 
@@ -1150,7 +1156,7 @@ func HandleAnthropicResponsesStream(
 			}
 			close(responseChan)
 		}()
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		// If body stream is nil, return an error
 		if resp.BodyStream() == nil {
 			bifrostErr := providerUtils.NewBifrostOperationError(
@@ -1204,6 +1210,10 @@ func HandleAnthropicResponsesStream(
 			}
 			eventType, eventDataBytes, readErr := sseReader.ReadEvent()
 			if readErr != nil {
+				// Recheck context cancellation
+				if ctx.Err() != nil {
+					return
+				}
 				if readErr != io.EOF {
 					ctx.SetValue(schemas.BifrostContextKeyStreamEndIndicator, true)
 					logger.Warn("Error reading %s stream: %v", providerName, readErr)
@@ -2647,7 +2657,7 @@ func (provider *AnthropicProvider) PassthroughStream(
 
 	activeClient := providerUtils.PrepareResponseStreaming(ctx, provider.streamingClient, resp)
 	if err := activeClient.Do(fasthttpReq, resp); err != nil {
-		providerUtils.ReleaseStreamingResponse(resp)
+		providerUtils.ReleaseStreamingResponse(ctx, resp)
 		if errors.Is(err, context.Canceled) {
 			return nil, &schemas.BifrostError{
 				IsBifrostError: false,
@@ -2669,7 +2679,7 @@ func (provider *AnthropicProvider) PassthroughStream(
 
 	bodyStream := resp.BodyStream()
 	if bodyStream == nil {
-		providerUtils.ReleaseStreamingResponse(resp)
+		providerUtils.ReleaseStreamingResponse(ctx, resp)
 		return nil, providerUtils.NewBifrostOperationError(
 			"provider returned an empty stream body",
 			fmt.Errorf("provider returned an empty stream body"),
@@ -2700,7 +2710,7 @@ func (provider *AnthropicProvider) PassthroughStream(
 			}
 			close(ch)
 		}()
-		defer providerUtils.ReleaseStreamingResponse(resp)
+		defer providerUtils.ReleaseStreamingResponse(ctx, resp)
 		defer stopIdleTimeout()
 		defer stopCancellation()
 

core/providers/anthropic/chat.go

@@ -745,6 +745,20 @@ func ToAnthropicChatRequest(ctx *schemas.BifrostContext, bifrostReq *schemas.Bif
 	anthropicReq.Messages = anthropicMessages
 	anthropicReq.System = systemContent
 
+	// Trim trailing whitespace from the last assistant message text blocks
+	// ContentStr is converted to a single text ContentBlock during message conversion
+	// so we trim the text of that block instead.
+	lastMsgIndex := len(anthropicReq.Messages) - 1
+	if lastMsgIndex >= 0 && anthropicReq.Messages[lastMsgIndex].Role == AnthropicMessageRoleAssistant {
+		blocks := anthropicReq.Messages[lastMsgIndex].Content.ContentBlocks
+		for j := len(blocks) - 1; j >= 0; j-- {
+			if blocks[j].Type == AnthropicContentBlockTypeText && blocks[j].Text != nil {
+				anthropicReq.Messages[lastMsgIndex].Content.ContentBlocks[j].Text = schemas.Ptr(strings.TrimRight(*blocks[j].Text, " \n\r\t"))
+				break
+			}
+		}
+	}
+
 	// Strip request- and tool-level fields the target Anthropic-family
 	// provider does not support. Fail-closed tool validation stays in
 	// ValidateToolsForProvider; this is strip-silently for additive fields.

core/providers/anthropic/responses.go

@@ -2738,9 +2738,8 @@ func (response *AnthropicMessageResponse) ToBifrostResponsesResponse(ctx *schema
 
 	bifrostResp.Model = response.Model
 
-	// Preserve stop reason from Anthropic response
 	if response.StopReason != "" {
-		bifrostResp.StopReason = schemas.Ptr(string(response.StopReason))
+		bifrostResp.StopReason = schemas.Ptr(ConvertAnthropicFinishReasonToBifrost(response.StopReason))
 	}
 
 	return bifrostResp
@@ -3344,6 +3343,20 @@ func ConvertBifrostMessagesToAnthropicMessages(ctx *schemas.BifrostContext, bifr
 	// Flush any remaining pending tool calls (with tracking)
 	flushPendingToolCallsWithTracking()
 
+	// Trim trailing whitespace from the last assistant message
+	// ContentStr is converted to a single text ContentBlock during message conversion
+	// so we trim the text of that block instead.
+	lastMsgIndex := len(anthropicMessages) - 1
+	if isRequestMessage && lastMsgIndex >= 0 && anthropicMessages[lastMsgIndex].Role == AnthropicMessageRoleAssistant {
+		blocks := anthropicMessages[lastMsgIndex].Content.ContentBlocks
+		for j := len(blocks) - 1; j >= 0; j-- {
+			if blocks[j].Type == AnthropicContentBlockTypeText && blocks[j].Text != nil {
+				anthropicMessages[lastMsgIndex].Content.ContentBlocks[j].Text = schemas.Ptr(strings.TrimRight(*blocks[j].Text, " \n\r\t"))
+				break
+			}
+		}
+	}
+
 	return anthropicMessages, systemContent
 }
 
@@ -5957,4 +5970,4 @@ func generateSyntheticInputJSONDeltas(argumentsJSON string, contentIndex *int) [
 	}
 
 	return events
-}
+}

core/providers/anthropic/types.go

@@ -183,7 +183,9 @@ var ProviderFeatures = map[schemas.ModelProvider]ProviderFeatureSupport{
 	// WebSearch, CodeExecution, FastMode, TaskBudgets, AdvisorTool,
 	// InferenceGeo, RedactThinking, AdvancedToolUse (full), PromptCachingScope.
 	schemas.Bedrock: {
-		ComputerUse: true, Bash: true, Memory: true, TextEditor: true, ToolSearch: true,
+		WebSearch:     true,
+		CodeExecution: true,
+		ComputerUse:   true, Bash: true, Memory: true, TextEditor: true, ToolSearch: true,
 		ContainerBasic: true,
 		// StructuredOutputs: kept true to match pre-existing behavior and the
 		// provider_feature_support_test.go assertion, but NEITHER B-header
@@ -1200,18 +1202,18 @@ const (
 type AnthropicToolName string
 
 const (
-	AnthropicToolNameComputer        AnthropicToolName = "computer"
-	AnthropicToolNameWebSearch       AnthropicToolName = "web_search"
-	AnthropicToolNameWebFetch        AnthropicToolName = "web_fetch"
-	AnthropicToolNameBash            AnthropicToolName = "bash"
-	AnthropicToolNameTextEditor      AnthropicToolName = "str_replace_based_edit_tool"
+	AnthropicToolNameComputer   AnthropicToolName = "computer"
+	AnthropicToolNameWebSearch  AnthropicToolName = "web_search"
+	AnthropicToolNameWebFetch   AnthropicToolName = "web_fetch"
+	AnthropicToolNameBash       AnthropicToolName = "bash"
+	AnthropicToolNameTextEditor AnthropicToolName = "str_replace_based_edit_tool"
 	// AnthropicToolNameTextEditorLegacy is the name required for text_editor_20250124
 	// and text_editor_20250429. Newer text_editor_20250728+ use AnthropicToolNameTextEditor.
 	AnthropicToolNameTextEditorLegacy AnthropicToolName = "str_replace_editor"
-	AnthropicToolNameCodeExecution   AnthropicToolName = "code_execution"
-	AnthropicToolNameMemory          AnthropicToolName = "memory"
-	AnthropicToolNameToolSearchBM25  AnthropicToolName = "tool_search_tool_bm25"
-	AnthropicToolNameToolSearchRegex AnthropicToolName = "tool_search_tool_regex"
+	AnthropicToolNameCodeExecution    AnthropicToolName = "code_execution"
+	AnthropicToolNameMemory           AnthropicToolName = "memory"
+	AnthropicToolNameToolSearchBM25   AnthropicToolName = "tool_search_tool_bm25"
+	AnthropicToolNameToolSearchRegex  AnthropicToolName = "tool_search_tool_regex"
 )
 
 type AnthropicToolComputerUse struct {

core/providers/anthropic/utils.go

@@ -2654,6 +2654,14 @@ func convertResponsesTextConfigToAnthropicOutputFormat(textConfig *schemas.Respo
 			schema["required"] = format.JSONSchema.Required
 		}
 
+		if format.JSONSchema.Defs != nil {
+			schema["$defs"] = *format.JSONSchema.Defs
+		}
+
+		if format.JSONSchema.Definitions != nil {
+			schema["definitions"] = *format.JSONSchema.Definitions
+		}
+
 		if format.JSONSchema.Type != nil && *format.JSONSchema.Type == "object" {
 			schema["additionalProperties"] = false
 		} else if format.JSONSchema.AdditionalProperties != nil {