fix: litellm compat

.github/workflows/configs/withpostgresmcpclientsinconfig/config.json

@@ -7,7 +7,6 @@
     ],
     "disable_content_logging": false,
     "drop_excess_requests": false,
-    "enable_litellm_fallbacks": false,
     "enable_logging": true,
     "enforce_auth_on_inference": true,
     "initial_pool_size": 300,

.github/workflows/scripts/run-migration-tests.sh

@@ -542,8 +542,8 @@ VALUES ('migration-test-lock', 'holder-migration-test-001', $future, $now)
 ON CONFLICT DO NOTHING;
 
 -- config_client (global client configuration)
-INSERT INTO config_client (id, drop_excess_requests, prometheus_labels_json, allowed_origins_json, allowed_headers_json, header_filter_config_json, initial_pool_size, enable_logging, disable_content_logging, disable_db_pings_in_health, log_retention_days, enforce_governance_header, allow_direct_keys, max_request_body_size_mb, mcp_agent_depth, mcp_tool_execution_timeout, mcp_code_mode_binding_level, mcp_tool_sync_interval, enable_litellm_fallbacks, config_hash, created_at, updated_at)
-VALUES (1, false, '["provider", "model"]', '["*"]', '["Authorization"]', '{}', 300, true, false, false, 365, true, false, true, 100, 10, 30, 'server', 10, false, 'client-config-hash-001', $now, $now)
+INSERT INTO config_client (id, drop_excess_requests, prometheus_labels_json, allowed_origins_json, allowed_headers_json, header_filter_config_json, initial_pool_size, enable_logging, disable_content_logging, disable_db_pings_in_health, log_retention_days, enforce_governance_header, allow_direct_keys, max_request_body_size_mb, mcp_agent_depth, mcp_tool_execution_timeout, mcp_code_mode_binding_level, mcp_tool_sync_interval, compat_convert_text_to_chat, compat_convert_chat_to_responses, compat_should_drop_params, compat_should_convert_params, config_hash, created_at, updated_at)
+VALUES (1, false, '["provider", "model"]', '["*"]', '["Authorization"]', '{}', 300, true, false, false, 365, true, false, 100, 10, 30, 'server', 10, false, false, false, true, 'client-config-hash-001', $now, $now)
 ON CONFLICT DO NOTHING;
 
 -- governance_config (key-value config table)
@@ -3509,4 +3509,4 @@ main() {
   exit $exit_code
 }
 
-main "$@"
+main "$@"

.github/workflows/scripts/test-docker-image.sh

@@ -212,8 +212,7 @@ cat > "$CONFIG_FILE" << 'CONFIGEOF'
     "enable_logging": true,
     "enforce_governance_header": false,
     "allow_direct_keys": false,
-    "max_request_body_size_mb": 100,
-    "enable_litellm_fallbacks": false
+    "max_request_body_size_mb": 100
   },
   "encryption_key": ""
 }

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

@@ -164,7 +164,11 @@ bifrost:
     enforceGovernanceHeader: true
     allowDirectKeys: true
     maxRequestBodySizeMb: 50
-    enableLitellmFallbacks: true
+    compat:
+      convertTextToChat: true
+      convertChatToResponses: true
+      shouldDropParams: true
+      shouldConvertParams: true
     prometheusLabels:
       - "team"
       - "env"
@@ -200,7 +204,10 @@ assert_field_value 'client.log_retention_days' '.client.log_retention_days' '30'
 assert_field_value 'client.enforce_governance_header' '.client.enforce_governance_header' 'true'
 assert_field_value 'client.allow_direct_keys' '.client.allow_direct_keys' 'true'
 assert_field_value 'client.max_request_body_size_mb' '.client.max_request_body_size_mb' '50'
-assert_field_value 'client.enable_litellm_fallbacks' '.client.enable_litellm_fallbacks' 'true'
+assert_field_value 'client.compat.convert_text_to_chat' '.client.compat.convert_text_to_chat' 'true'
+assert_field_value 'client.compat.convert_chat_to_responses' '.client.compat.convert_chat_to_responses' 'true'
+assert_field_value 'client.compat.should_drop_params' '.client.compat.should_drop_params' 'true'
+assert_field_value 'client.compat.should_convert_params' '.client.compat.should_convert_params' 'true'
 assert_field 'client.prometheus_labels' '.client.prometheus_labels'
 assert_field 'client.header_filter_config.allowlist' '.client.header_filter_config.allowlist'
 assert_field 'client.header_filter_config.denylist' '.client.header_filter_config.denylist'
@@ -1194,4 +1201,4 @@ if [ "$TESTS_FAILED" -gt 0 ]; then
 else
   echo -e "${GREEN}✅ All config.json field validations passed!${NC}"
   exit 0
-fi
+fi

AGENTS.md

@@ -103,7 +103,7 @@ bifrost/
 │   ├── mocker/                    # Mock responses for testing
 │   ├── jsonparser/                # JSON extraction utilities
 │   ├── maxim/                     # Maxim observability
-│   └── litellmcompat/             # LiteLLM SDK compatibility (HTTP transport)
+│   └── compat/                    # LiteLLM SDK compatibility (HTTP transport)
 │
 ├── ui/                            # Next.js web interface
 │   ├── app/workspace/             # Feature pages (20+ workspace sections)
@@ -647,4 +647,4 @@ Systematically address unresolved PR review comments. Uses GraphQL to get unreso
 - **Provider types**: Prefixed with provider name in PascalCase (`AnthropicChatRequest`, `GeminiEmbeddingResponse`).
 - **Converter functions**: Pure — no side effects, no logging, no HTTP.
 - **Pool names**: Descriptive string passed to `pool.New()` (e.g., `"channel-message"`, `"response-stream"`).
-- **Context keys**: Use `BifrostContextKey` type. Custom plugins should define their own key types to avoid collisions.
+- **Context keys**: Use `BifrostContextKey` type. Custom plugins should define their own key types to avoid collisions.

core/bifrost.go

@@ -612,7 +612,7 @@ func (bifrost *Bifrost) TextCompletionRequest(ctx *schemas.BifrostContext, req *
 	if err != nil {
 		return nil, err
 	}
-	//TODO: Release the response
+	// TODO: Release the response
 	return response.TextCompletionResponse, nil
 }
 
@@ -934,7 +934,7 @@ func (bifrost *Bifrost) EmbeddingRequest(ctx *schemas.BifrostContext, req *schem
 	if err != nil {
 		return nil, err
 	}
-	//TODO: Release the response
+	// TODO: Release the response
 	return response.EmbeddingResponse, nil
 }
 
@@ -1042,7 +1042,7 @@ func (bifrost *Bifrost) SpeechRequest(ctx *schemas.BifrostContext, req *schemas.
 	if err != nil {
 		return nil, err
 	}
-	//TODO: Release the response
+	// TODO: Release the response
 	return response.SpeechResponse, nil
 }
 
@@ -1117,7 +1117,7 @@ func (bifrost *Bifrost) TranscriptionRequest(ctx *schemas.BifrostContext, req *s
 	if err != nil {
 		return nil, err
 	}
-	//TODO: Release the response
+	// TODO: Release the response
 	return response.TranscriptionResponse, nil
 }
 
@@ -1158,7 +1158,8 @@ func (bifrost *Bifrost) TranscriptionStreamRequest(ctx *schemas.BifrostContext,
 
 // ImageGenerationRequest sends an image generation request to the specified provider.
 func (bifrost *Bifrost) ImageGenerationRequest(ctx *schemas.BifrostContext,
-	req *schemas.BifrostImageGenerationRequest) (*schemas.BifrostImageGenerationResponse, *schemas.BifrostError) {
+	req *schemas.BifrostImageGenerationRequest,
+) (*schemas.BifrostImageGenerationResponse, *schemas.BifrostError) {
 	if req == nil {
 		return nil, &schemas.BifrostError{
 			IsBifrostError: false,
@@ -1213,7 +1214,8 @@ func (bifrost *Bifrost) ImageGenerationRequest(ctx *schemas.BifrostContext,
 
 // ImageGenerationStreamRequest sends an image generation stream request to the specified provider.
 func (bifrost *Bifrost) ImageGenerationStreamRequest(ctx *schemas.BifrostContext,
-	req *schemas.BifrostImageGenerationRequest) (chan *schemas.BifrostStreamChunk, *schemas.BifrostError) {
+	req *schemas.BifrostImageGenerationRequest,
+) (chan *schemas.BifrostStreamChunk, *schemas.BifrostError) {
 	if req == nil {
 		return nil, &schemas.BifrostError{
 			IsBifrostError: false,
@@ -1434,7 +1436,8 @@ func (bifrost *Bifrost) ImageVariationRequest(ctx *schemas.BifrostContext, req *
 
 // VideoGenerationRequest sends a video generation request to the specified provider.
 func (bifrost *Bifrost) VideoGenerationRequest(ctx *schemas.BifrostContext,
-	req *schemas.BifrostVideoGenerationRequest) (*schemas.BifrostVideoGenerationResponse, *schemas.BifrostError) {
+	req *schemas.BifrostVideoGenerationRequest,
+) (*schemas.BifrostVideoGenerationResponse, *schemas.BifrostError) {
 	if req == nil {
 		return nil, &schemas.BifrostError{
 			IsBifrostError: false,
@@ -4692,7 +4695,7 @@ func (bifrost *Bifrost) tryStreamRequest(ctx *schemas.BifrostContext, req *schem
 					// Send the processed message to the output stream
 					outputStream <- streamResponse
 
-					//TODO: Release the processed response immediately after use
+					// TODO: Release the processed response immediately after use
 				}
 			}()
 
@@ -5264,12 +5267,34 @@ func (bifrost *Bifrost) handleProviderRequest(provider schemas.Provider, req *Ch
 		}
 		response.ListModelsResponse = listModelsResponse
 	case schemas.TextCompletionRequest:
+		if changeType, ok := req.Context.Value(schemas.BifrostContextKeyChangeRequestType).(schemas.RequestType); ok && changeType == schemas.ChatCompletionRequest {
+			chatRequest := req.BifrostRequest.TextCompletionRequest.ToBifrostChatRequest()
+			if chatRequest != nil {
+				chatCompletionResponse, bifrostError := provider.ChatCompletion(req.Context, key, chatRequest)
+				if bifrostError != nil {
+					return nil, bifrostError
+				}
+				response.TextCompletionResponse = chatCompletionResponse.ToBifrostTextCompletionResponse()
+				break
+			}
+		}
 		textCompletionResponse, bifrostError := provider.TextCompletion(req.Context, key, req.BifrostRequest.TextCompletionRequest)
 		if bifrostError != nil {
 			return nil, bifrostError
 		}
 		response.TextCompletionResponse = textCompletionResponse
 	case schemas.ChatCompletionRequest:
+		if changeType, ok := req.Context.Value(schemas.BifrostContextKeyChangeRequestType).(schemas.RequestType); ok && changeType == schemas.ResponsesRequest {
+			responsesRequest := req.BifrostRequest.ChatRequest.ToResponsesRequest()
+			if responsesRequest != nil {
+				responsesResponse, bifrostError := provider.Responses(req.Context, key, responsesRequest)
+				if bifrostError != nil {
+					return nil, bifrostError
+				}
+				response.ChatResponse = responsesResponse.ToBifrostChatResponse()
+				break
+			}
+		}
 		chatCompletionResponse, bifrostError := provider.ChatCompletion(req.Context, key, req.BifrostRequest.ChatRequest)
 		if bifrostError != nil {
 			return nil, bifrostError
@@ -5519,8 +5544,20 @@ func (bifrost *Bifrost) handleProviderRequest(provider schemas.Provider, req *Ch
 func (bifrost *Bifrost) handleProviderStreamRequest(provider schemas.Provider, req *ChannelMessage, key schemas.Key, postHookRunner schemas.PostHookRunner) (chan *schemas.BifrostStreamChunk, *schemas.BifrostError) {
 	switch req.RequestType {
 	case schemas.TextCompletionStreamRequest:
+		if changeType, ok := req.Context.Value(schemas.BifrostContextKeyChangeRequestType).(schemas.RequestType); ok && changeType == schemas.ChatCompletionRequest {
+			chatRequest := req.BifrostRequest.TextCompletionRequest.ToBifrostChatRequest()
+			if chatRequest != nil {
+				return provider.ChatCompletionStream(req.Context, wrapConvertedStreamPostHookRunner(postHookRunner, schemas.ChatCompletionRequest), key, chatRequest)
+			}
+		}
 		return provider.TextCompletionStream(req.Context, postHookRunner, key, req.BifrostRequest.TextCompletionRequest)
 	case schemas.ChatCompletionStreamRequest:
+		if changeType, ok := req.Context.Value(schemas.BifrostContextKeyChangeRequestType).(schemas.RequestType); ok && changeType == schemas.ResponsesRequest {
+			responsesRequest := req.BifrostRequest.ChatRequest.ToResponsesRequest()
+			if responsesRequest != nil {
+				return provider.ResponsesStream(req.Context, wrapConvertedStreamPostHookRunner(postHookRunner, schemas.ResponsesRequest), key, responsesRequest)
+			}
+		}
 		return provider.ChatCompletionStream(req.Context, postHookRunner, key, req.BifrostRequest.ChatRequest)
 	case schemas.ResponsesStreamRequest:
 		return provider.ResponsesStream(req.Context, postHookRunner, key, req.BifrostRequest.ResponsesRequest)

core/providers/anthropic/types.go

@@ -1316,4 +1316,4 @@ func parseAnthropicFileTimestamp(timestamp string) int64 {
 // AnthropicCountTokensResponse models the payload returned by Anthropic's count tokens endpoint.
 type AnthropicCountTokensResponse struct {
 	InputTokens int `json:"input_tokens"`
-}
+}

core/providers/bedrock/images.go

@@ -153,7 +153,6 @@ func ToBedrockImageGenerationRequest(request *schemas.BifrostImageGenerationRequ
 	}
 
 	return bedrockReq, nil
-
 }
 
 // ToStabilityAIImageGenerationResponse converts a BifrostImageGenerationResponse back to

core/providers/bedrock/models.go

@@ -81,7 +81,6 @@ type BedrockRerankResponseDocument struct {
 	TextDocument *BedrockRerankTextValue `json:"textDocument,omitempty"`
 }
 
-
 func (response *BedrockListModelsResponse) ToBifrostListModelsResponse(providerKey schemas.ModelProvider, allowedModels schemas.WhiteList, blacklistedModels schemas.BlackList, aliases map[string]string, unfiltered bool) *schemas.BifrostListModelsResponse {
 	if response == nil {
 		return nil
@@ -128,4 +127,4 @@ func (response *BedrockListModelsResponse) ToBifrostListModelsResponse(providerK
 		pipeline.BackfillModels(included)...)
 
 	return bifrostResponse
-}
+}

core/providers/cohere/types.go

@@ -9,8 +9,11 @@ import (
 	"github.com/maximhq/bifrost/core/schemas"
 )
 
-const MinimumReasoningMaxTokens = 1
-const DefaultCompletionMaxTokens = 4096 // Only used for relative reasoning max token calculation - not passed in body by default
+const (
+	MinimumReasoningMaxTokens  = 1
+	DefaultCompletionMaxTokens = 4096 // Only used for relative reasoning max token calculation - not passed in body by default
+)
+
 // Limits for tokenize input api call https://docs.cohere.com/reference/tokenize#request
 const (
 	cohereTokenizeMinTextLength = 1

core/providers/gemini/types.go

@@ -17,11 +17,13 @@ import (
 	"github.com/maximhq/bifrost/core/schemas"
 )
 
-const MinReasoningMaxTokens = 1         // Minimum max tokens for reasoning - used for estimation of effort level
-const DefaultCompletionMaxTokens = 8192 // Default max output tokens for Gemini - used for relative reasoning max token calculation
-const DefaultReasoningMinBudget = 1024  // Default minimum reasoning budget for Gemini
-const DynamicReasoningBudget = -1       // Special value for dynamic reasoning budget in Gemini
-const skipThoughtSignatureValidator = "skip_thought_signature_validator"
+const (
+	MinReasoningMaxTokens         = 1    // Minimum max tokens for reasoning - used for estimation of effort level
+	DefaultCompletionMaxTokens    = 8192 // Default max output tokens for Gemini - used for relative reasoning max token calculation
+	DefaultReasoningMinBudget     = 1024 // Default minimum reasoning budget for Gemini
+	DynamicReasoningBudget        = -1   // Special value for dynamic reasoning budget in Gemini
+	skipThoughtSignatureValidator = "skip_thought_signature_validator"
+)
 
 type thinkingBudgetRange struct {
 	Min int
@@ -509,8 +511,7 @@ type GoogleMaps struct {
 }
 
 // URLContext is a tool to support URL context retrieval.
-type URLContext struct {
-}
+type URLContext struct{}
 
 // ToolComputerUse is a tool to support computer use.
 type ToolComputerUse struct {
@@ -555,8 +556,7 @@ type ExternalAPIElasticSearchParams struct {
 }
 
 // ExternalAPISimpleSearchParams represents the search parameters to use for SIMPLE_SEARCH spec.
-type ExternalAPISimpleSearchParams struct {
-}
+type ExternalAPISimpleSearchParams struct{}
 
 // ExternalAPI retrieves from data source powered by external API for grounding. The external API
 // is not owned by Google, but needs to follow the pre-defined API spec.
@@ -714,8 +714,7 @@ type Retrieval struct {
 // ToolCodeExecution is a tool that executes code generated by the model, and automatically returns the result
 // to the model. See also [ExecutableCode]and [CodeExecutionResult] which are input
 // and output to this tool.
-type ToolCodeExecution struct {
-}
+type ToolCodeExecution struct{}
 
 // Tool details of a tool that the model may use to generate a response.
 type Tool struct {

core/providers/openai/types.go

@@ -6,8 +6,8 @@ import (
 	"fmt"
 
 	"github.com/bytedance/sonic"
-	"github.com/maximhq/bifrost/core/schemas"
 	providerUtils "github.com/maximhq/bifrost/core/providers/utils"
+	"github.com/maximhq/bifrost/core/schemas"
 )
 
 const MinMaxCompletionTokens = 16
@@ -82,7 +82,7 @@ type OpenAIChatRequest struct {
 	// PromptCacheIsolationKey is the Fireworks chat-completions field for cache isolation.
 	PromptCacheIsolationKey *string `json:"prompt_cache_isolation_key,omitempty"`
 
-	//NOTE: MaxCompletionTokens is a new replacement for max_tokens but some providers still use max_tokens.
+	// NOTE: MaxCompletionTokens is a new replacement for max_tokens but some providers still use max_tokens.
 	// This Field is populated only for such providers and is NOT to be used externally.
 	MaxTokens *int `json:"max_tokens,omitempty"`