Skip to content

Open source AI coding agent. Designed for large projects and real world tasks.

Merge branch 'main' into litellm

Dane Schneiderf3c81fc5f442
diff --cc app/server/handlers/models.go
index 7aa56460,5fc7c0e1..9a234393
--- a/app/server/handlers/models.go
+++ b/app/server/handlers/models.go
@@@ -40,176 -32,71 +40,176 @@@ func CreateCustomModelHandler(w http.Re
return
}
- baseModelConfig := model.BaseModelConfig
+ for _, model := range modelsInput.CustomModels {
+ if model.ModelId == "" {
+ msg := "Model id is required"
+ log.Println(msg)
+ http.Error(w, msg, http.StatusBadRequest)
+ return
+ }
- dbModel := &db.AvailableModel{
- Id: model.Id,
- OrgId: auth.OrgId,
- Provider: baseModelConfig.Provider,
- CustomProvider: baseModelConfig.CustomProvider,
- BaseUrl: baseModelConfig.BaseUrl,
- ModelId: baseModelConfig.ModelId,
- ModelName: baseModelConfig.ModelName,
- Description: model.Description,
- MaxTokens: baseModelConfig.MaxTokens,
- ApiKeyEnvVar: baseModelConfig.ApiKeyEnvVar,
- HasImageSupport: baseModelConfig.ModelCompatibility.HasImageSupport,
- DefaultMaxConvoTokens: model.DefaultMaxConvoTokens,
- MaxOutputTokens: baseModelConfig.MaxOutputTokens,
- ReservedOutputTokens: baseModelConfig.ReservedOutputTokens,
- PreferredOutputFormat: baseModelConfig.PreferredModelOutputFormat,
+ if shared.BuiltInBaseModelsById[model.ModelId] != nil {
+ msg := fmt.Sprintf("%s is a built-in base model id, so it can't be used for a custom model", model.ModelId)
+ log.Println(msg)
+ http.Error(w, msg, http.StatusUnprocessableEntity)
+ return
+ }
}
- if err := db.CreateCustomModel(dbModel); err != nil {
- log.Printf("Error creating custom model: %v\n", err)
- http.Error(w, "Failed to create custom model: "+err.Error(), http.StatusInternalServerError)
- return
+ for _, provider := range modelsInput.CustomProviders {
+ if provider.Name == "" {
+ msg := "Provider name is required"
+ log.Println(msg)
+ http.Error(w, msg, http.StatusBadRequest)
+ return
+ }
}
- w.WriteHeader(http.StatusCreated)
+ for _, modelPack := range modelsInput.CustomModelPacks {
+ if modelPack.Name == "" {
+ msg := "Model pack name is required"
+ log.Println(msg)
+ http.Error(w, msg, http.StatusBadRequest)
+ return
+ }
+ }
- log.Println("Successfully created custom model")
-}
+ var customModelIds = make(map[string]bool)
+ var customProviderNames = make(map[string]bool)
-func UpdateCustomModelHandler(w http.ResponseWriter, r *http.Request) {
- log.Println("Received request for UpdateCustomModelHandler")
+ if len(modelsInput.CustomModels) > 0 {
- auth := Authenticate(w, r, true)
- if auth == nil {
- return
- }
+ var modelIds []string
+ for _, model := range modelsInput.CustomModels {
+ modelIds = append(modelIds, string(model.ModelId))
+ }
- modelId := mux.Vars(r)["modelId"]
+ customModels, err := db.ListCustomModelsForModelIds(auth.OrgId, modelIds)
+ if err != nil {
+ log.Printf("Error fetching custom models: %v\n", err)
+ http.Error(w, "Failed to create custom model: "+err.Error(), http.StatusInternalServerError)
+ return
+ }
+
+ for _, model := range customModels {
+ customModelIds[model.Id] = true
+ }
- var model shared.AvailableModel
- if err := json.NewDecoder(r.Body).Decode(&model); err != nil {
- log.Printf("Error decoding request body: %v\n", err)
- http.Error(w, "Invalid request body", http.StatusBadRequest)
- return
}
- models, err := db.ListCustomModels(auth.OrgId)
- if err != nil {
- log.Printf("Error fetching custom models: %v\n", err)
- http.Error(w, "Failed to fetch custom models: "+err.Error(), http.StatusInternalServerError)
- return
+ if len(modelsInput.CustomProviders) > 0 {
+ var names []string
+ for _, provider := range modelsInput.CustomProviders {
+ names = append(names, provider.Name)
+ }
+
+ customProviders, err := db.ListCustomProvidersForNames(auth.OrgId, names)
+ if err != nil {
+ log.Printf("Error fetching custom providers: %v\n", err)
+ http.Error(w, "Failed to create custom model: "+err.Error(), http.StatusInternalServerError)
+ return
+ }
+
+ for _, provider := range customProviders {
+ customProviderNames[provider.Name] = true
+ }
}
- found := false
- for _, m := range models {
- if m.Id == modelId {
- found = true
- break
+ var toUpsertCustomModels []*db.CustomModel
+ var toUpsertCustomProviders []*db.CustomProvider
+ var toUpsertModelPacks []*db.ModelPack
+
+ willCreateModelIds := make(map[string]bool)
+ willCreateProviderNames := make(map[string]bool)
+
+ for _, provider := range modelsInput.CustomProviders {
- dbProvider := db.CustomProviderFromApi(&provider)
++ dbProvider := db.CustomProviderFromApi(provider)
+ dbProvider.Id = provider.Id
+ dbProvider.OrgId = auth.OrgId
+
+ toUpsertCustomProviders = append(toUpsertCustomProviders, dbProvider)
+ willCreateProviderNames[provider.Name] = true
+ }
+
+ for _, model := range modelsInput.CustomModels {
+ // ensure that providers are either built-in, being created, or already exist
+ for _, provider := range model.Providers {
+ if provider.Provider == shared.ModelProviderCustom {
+ _, exists := customProviderNames[*provider.CustomProvider]
+ _, creating := willCreateProviderNames[*provider.CustomProvider]
+ if !exists && !creating {
+ msg := fmt.Sprintf("'%s' is not a custom model provider that exists or is being created", *provider.CustomProvider)
+ log.Println(msg)
+ http.Error(w, msg, http.StatusUnprocessableEntity)
+ return
+ }
+ } else {
+ _, builtIn := shared.BuiltInModelProviderConfigs[provider.Provider]
+ if !builtIn {
+ msg := fmt.Sprintf("'%s' is not a built-in model provider", provider.Provider)
+ log.Println(msg)
+ http.Error(w, msg, http.StatusUnprocessableEntity)
+ return
+ }
+ }
+ }
+
- dbModel := db.CustomModelFromApi(&model)
++ dbModel := db.CustomModelFromApi(model)
+ dbModel.Id = model.Id
+ dbModel.OrgId = auth.OrgId
+
+ toUpsertCustomModels = append(toUpsertCustomModels, dbModel)
+ willCreateModelIds[string(model.ModelId)] = true
+ }
+
+ for _, modelPack := range modelsInput.CustomModelPacks {
+ // ensure that all models are either built-in, being created, or already exist
+ allModelIds := modelPack.AllModelIds()
+ for _, modelId := range allModelIds {
+ _, exists := customModelIds[string(modelId)]
+ _, creating := willCreateModelIds[string(modelId)]
+ _, builtIn := shared.BuiltInBaseModelsById[modelId]
+ if !exists && !creating && !builtIn {
+ msg := fmt.Sprintf("'%s' is not built-in, not being created, and not an existing custom model", modelId)
+ log.Println(msg)
+ http.Error(w, msg, http.StatusUnprocessableEntity)
+ return
+ }
}
+
+ mp := modelPack.ToModelPack()
+ dbMp := db.ModelPackFromApi(&mp)
+ dbMp.OrgId = auth.OrgId
+ dbMp.Id = mp.Id
+
+ toUpsertModelPacks = append(toUpsertModelPacks, dbMp)
}
- if !found {
- http.Error(w, "Custom model not found", http.StatusNotFound)
+ err := db.WithTx(r.Context(), "create custom models/providers/model packs", func(tx *sqlx.Tx) error {
+ for _, model := range toUpsertCustomModels {
+ if err := db.UpsertCustomModel(tx, model); err != nil {
+ return fmt.Errorf("error creating custom model: %w", err)
+ }
+ }
+
+ for _, provider := range toUpsertCustomProviders {
+ if err := db.UpsertCustomProvider(tx, provider); err != nil {
+ return fmt.Errorf("error creating custom provider: %w", err)
+ }
+ }
+
+ for _, modelPack := range toUpsertModelPacks {
+ if err := db.UpsertModelPack(tx, modelPack); err != nil {
+ return fmt.Errorf("error creating model pack: %w", err)
+ }
+ }
+
+ return nil
+ })
+
+ if err != nil {
+ log.Printf("Error: %v\n", err)
+ http.Error(w, "Failed to create custom model: "+err.Error(), http.StatusInternalServerError)
return
}
diff --cc app/server/model/client.go
index 8ffb43fe,2f57c563..4a50e8d1
--- a/app/server/model/client.go
+++ b/app/server/model/client.go
@@@ -126,12 -122,24 +126,24 @@@ func CreateChatCompletionStream
req.Model += ":nitro"
}
- if baseModelConfig.IncludeReasoning {
- req.IncludeReasoning = true
- if modelConfig.BaseModelConfig.ReasoningBudgetEnabled {
++ if baseModelConfig.ReasoningBudgetEnabled {
+ req.ReasoningConfig = &types.ReasoningConfig{
- MaxTokens: modelConfig.BaseModelConfig.ReasoningBudget,
- Exclude: !modelConfig.BaseModelConfig.IncludeReasoning,
++ MaxTokens: baseModelConfig.ReasoningBudget,
++ Exclude: !baseModelConfig.IncludeReasoning,
+ }
- } else if modelConfig.BaseModelConfig.ReasoningEffortEnabled {
++ } else if baseModelConfig.ReasoningEffortEnabled {
+ req.ReasoningConfig = &types.ReasoningConfig{
- Effort: shared.ReasoningEffort(modelConfig.BaseModelConfig.ReasoningEffort),
- Exclude: !modelConfig.BaseModelConfig.IncludeReasoning,
++ Effort: shared.ReasoningEffort(baseModelConfig.ReasoningEffort),
++ Exclude: !baseModelConfig.IncludeReasoning,
+ }
- } else if modelConfig.BaseModelConfig.IncludeReasoning {
++ } else if baseModelConfig.IncludeReasoning {
+ req.ReasoningConfig = &types.ReasoningConfig{
+ Exclude: false,
+ }
}
- return withStreamingRetries(ctx, func(numTotalRetry int, modelErr *shared.ModelError, stripCacheControl bool) (*ExtendedChatCompletionStream, shared.FallbackResult, error) {
- fallbackRes := modelConfig.GetFallbackForModelError(numTotalRetry, modelErr)
+ return withStreamingRetries(ctx, func(numTotalRetry int, didProviderFallback bool, modelErr *shared.ModelError) (*ExtendedChatCompletionStream, shared.FallbackResult, error) {
+ fallbackRes := modelConfig.GetFallbackForModelError(numTotalRetry, didProviderFallback, modelErr, authVars)
resolvedModelConfig := fallbackRes.ModelRoleConfig
if resolvedModelConfig == nil {
diff --cc app/shared/ai_models_available.go
index cbb6fecc,ab29ca2d..62d8d0cd
--- a/app/shared/ai_models_available.go
+++ b/app/shared/ai_models_available.go
@@@ -46,894 -44,773 +46,417 @@@ When checking for sufficient credits o
'ApiKeyEnvVar' is the environment variable that contains the API key for the model.
*/
-var AvailableModels = []*AvailableModel{
- // Direct OpenAI models
+var BuiltInModels = []*BaseModelConfigSchema{
{
- Description: "OpenAI o3-high",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o3",
- ModelId: "openai/o3-high",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
+ ModelTag: "openai/o3",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI o3",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- StopDisabled: true,
- },
-
- RequiresVariantOverrides: []string{
- "ReasoningEffort",
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 100000,
++ ReservedOutputTokens: 40000, ModelCompatibility: FullCompatibility,
++ PreferredOutputFormat: ModelOutputFormatXml, SystemPromptDisabled: true,
++ RoleParamsDisabled: true, ReasoningEffortEnabled: true, StopDisabled: true,
+ },
-
++ RequiresVariantOverrides: []string{"ReasoningEffort"},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "high",
- Description: "high",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortHigh,
- },
- },
- {
- VariantTag: "medium",
- Description: "medium",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortMedium,
- },
- },
- {
- VariantTag: "low",
- Description: "low",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortLow,
- },
- },
++ {VariantTag: "high", Description: "high", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortHigh}},
++ {VariantTag: "medium", Description: "medium", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortMedium}},
++ {VariantTag: "low", Description: "low", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortLow}},
},
- },
--
- {
- Description: "OpenAI o3-medium",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o3",
- ModelId: "openai/o3-medium",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortMedium,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "o3",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/o3",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o3",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "o3"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/o3"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/o3"},
+ },
+ },
+ {
+ ModelTag: "openai/o4-mini",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI o4-mini",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 10000,
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatToolCallJson,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
- },
-
- RequiresVariantOverrides: []string{
- "ReasoningEffort",
++ DefaultMaxConvoTokens: 10000, MaxTokens: 200000, MaxOutputTokens: 100000,
++ ReservedOutputTokens: 40000, ModelCompatibility: FullCompatibility,
++ PreferredOutputFormat: ModelOutputFormatToolCallJson, SystemPromptDisabled: true,
++ RoleParamsDisabled: true, ReasoningEffortEnabled: true, ReasoningEffort: ReasoningEffortHigh,
++ StopDisabled: true,
+ },
-
++ RequiresVariantOverrides: []string{"ReasoningEffort"},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "high",
- Description: "high",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortHigh,
- },
- },
- {
- VariantTag: "medium",
- Description: "medium",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortMedium,
- ReservedOutputTokens: 30000, // 15k for reasoning, 15k for output
- },
- },
- {
- VariantTag: "low",
- Description: "low",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortLow,
- ReservedOutputTokens: 20000, // 5-10k for reasoning, 5-10k for output
- },
- },
++ {VariantTag: "high", Description: "high", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortHigh}},
++ {VariantTag: "medium", Description: "medium", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortMedium, ReservedOutputTokens: 30000}},
++ {VariantTag: "low", Description: "low", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortLow, ReservedOutputTokens: 20000}},
+ },
-
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "o4-mini",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/o4-mini",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o4-mini",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "o4-mini"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/o4-mini"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/o4-mini"},
+ },
+ },
-
+ {
+ ModelTag: "openai/gpt-4.1",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI gpt-4.1",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatToolCallJson,
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1047576,
++ MaxOutputTokens: 32768, ReservedOutputTokens: 32768,
++ ModelCompatibility: FullCompatibility, PreferredOutputFormat: ModelOutputFormatToolCallJson,
+ },
-
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/gpt-4.1",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "gpt-4.1"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/gpt-4.1"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/gpt-4.1"},
+ },
+ },
-
+ {
+ ModelTag: "openai/gpt-4.1-mini",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI gpt-4.1-mini",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatToolCallJson,
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1047576,
++ MaxOutputTokens: 32768, ReservedOutputTokens: 32768,
++ ModelCompatibility: FullCompatibility, PreferredOutputFormat: ModelOutputFormatToolCallJson,
+ },
-
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1-mini",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/gpt-4.1-mini",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1-mini",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "gpt-4.1-mini"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/gpt-4.1-mini"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/gpt-4.1-mini"},
+ },
+ },
-
+ {
+ ModelTag: "openai/gpt-4.1-nano",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI gpt-4.1-nano",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatToolCallJson,
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1047576,
++ MaxOutputTokens: 32768, ReservedOutputTokens: 32768,
++ ModelCompatibility: FullCompatibility, PreferredOutputFormat: ModelOutputFormatToolCallJson,
+ },
-
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1-nano",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/gpt-4.1-nano",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1-nano",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "gpt-4.1-nano"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/gpt-4.1-nano"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/gpt-4.1-nano"},
+ },
+ },
-
+ {
+ ModelTag: "openai/o3-mini",
+ Publisher: ModelPublisherOpenAI,
+ Description: "OpenAI o3-mini",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 10000,
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ModelCompatibility: FullCompatibility,
- PreferredOutputFormat: ModelOutputFormatToolCallJson,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
- },
-
- RequiresVariantOverrides: []string{
- "ReasoningEffort",
++ DefaultMaxConvoTokens: 10000, MaxTokens: 200000, MaxOutputTokens: 100000,
++ ReservedOutputTokens: 40000, ModelCompatibility: FullCompatibility,
++ PreferredOutputFormat: ModelOutputFormatToolCallJson, SystemPromptDisabled: true,
++ RoleParamsDisabled: true, ReasoningEffortEnabled: true, ReasoningEffort: ReasoningEffortHigh,
++ StopDisabled: true,
+ },
-
++ RequiresVariantOverrides: []string{"ReasoningEffort"},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "high",
- Description: "high",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortHigh,
- },
- },
- {
- VariantTag: "medium",
- Description: "medium",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortMedium,
- ReservedOutputTokens: 30000, // 15k for reasoning, 15k for output
- },
- },
- {
- VariantTag: "low",
- Description: "low",
- Overrides: BaseModelShared{
- ReasoningEffort: ReasoningEffortLow,
- ReservedOutputTokens: 20000, // 5-10k for reasoning, 5-10k for output
- },
- },
++ {VariantTag: "high", Description: "high", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortHigh}},
++ {VariantTag: "medium", Description: "medium", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortMedium, ReservedOutputTokens: 30000}},
++ {VariantTag: "low", Description: "low", Overrides: BaseModelShared{ReasoningEffort: ReasoningEffortLow, ReservedOutputTokens: 20000}},
+ },
-
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderOpenAI,
- ModelName: "o3-mini",
- },
- {
- Provider: ModelProviderAzureOpenAI,
- ModelName: "azure/o3-mini",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o3-mini",
- },
++ {Provider: ModelProviderOpenAI, ModelName: "o3-mini"},
++ {Provider: ModelProviderAzureOpenAI, ModelName: "azure/o3-mini"},
++ {Provider: ModelProviderOpenRouter, ModelName: "openai/o3-mini"},
+ },
+ },
-
+ {
+ ModelTag: "anthropic/claude-opus-4",
+ Publisher: ModelPublisherAnthropic,
+ Description: "Anthropic Claude Opus 4",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- PreferredOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 128000,
++ ReservedOutputTokens: 20000, SupportsCacheControl: true,
++ PreferredOutputFormat: ModelOutputFormatXml, SingleMessageNoSystemPrompt: true,
++ TokenEstimatePaddingPct: 0.10,
+ },
-
+ Variants: []BaseModelConfigVariant{
++ {IsBaseVariant: true},
+ {
- IsBaseVariant: true,
- },
- {
- VariantTag: "thinking",
- Description: "thinking",
- Overrides: BaseModelShared{
- ReasoningBudget: AnthropicMaxReasoningBudget,
- },
-
++ VariantTag: "thinking", Description: "thinking",
++ Overrides: BaseModelShared{ReasoningBudget: AnthropicMaxReasoningBudget},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "visible",
- Description: "visible",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "hidden",
- Description: "hidden",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "visible", Description: "visible", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "hidden", Description: "hidden", Overrides: BaseModelShared{IncludeReasoning: false}},
+ },
+ },
},
- },
--
- {
- Description: "OpenAI o3-low",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o3",
- ModelId: "openai/o3-low",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortLow,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderAnthropic,
- ModelName: "anthropic/claude-opus-4-0",
- },
- {
- Provider: ModelProviderAmazonBedrock,
- ModelName: "bedrock/anthropic.claude-opus-4-20250514-v1:0",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/claude-opus-4@20250514",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-opus-4",
- },
++ {Provider: ModelProviderAnthropic, ModelName: "anthropic/claude-opus-4-0"},
++ {Provider: ModelProviderAmazonBedrock, ModelName: "bedrock/anthropic.claude-opus-4-20250514-v1:0"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/claude-opus-4@20250514"},
++ {Provider: ModelProviderOpenRouter, ModelName: "anthropic/claude-opus-4"},
+ },
+ },
-
+ {
+ ModelTag: "anthropic/claude-sonnet-4",
+ Publisher: ModelPublisherAnthropic,
+ Description: "Anthropic Claude Sonnet 4",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- PreferredOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 128000,
++ ReservedOutputTokens: 40000, SupportsCacheControl: true,
++ PreferredOutputFormat: ModelOutputFormatXml, SingleMessageNoSystemPrompt: true,
++ TokenEstimatePaddingPct: 0.10,
+ },
-
+ Variants: []BaseModelConfigVariant{
++ {IsBaseVariant: true},
+ {
- IsBaseVariant: true,
- },
- {
- VariantTag: "thinking",
- Description: "thinking",
- Overrides: BaseModelShared{
- ReasoningBudget: AnthropicMaxReasoningBudget,
- },
-
++ VariantTag: "thinking", Description: "thinking",
++ Overrides: BaseModelShared{ReasoningBudget: AnthropicMaxReasoningBudget},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "visible",
- Description: "visible",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "hidden",
- Description: "hidden",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "visible", Description: "visible", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "hidden", Description: "hidden", Overrides: BaseModelShared{IncludeReasoning: false}},
+ },
+ },
},
- },
--
- {
- Description: "OpenAI o4-mini-high",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o4-mini",
- ModelId: "openai/o4-mini-high",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 30000,
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderAnthropic,
- ModelName: "anthropic/claude-sonnet-4-0", // “-0” alias
- },
- {
- Provider: ModelProviderAmazonBedrock,
- ModelName: "bedrock/anthropic.claude-sonnet-4-20250514-v1:0",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/claude-sonnet-4@20250514",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-sonnet-4",
- },
++ {Provider: ModelProviderAnthropic, ModelName: "anthropic/claude-sonnet-4-0"},
++ {Provider: ModelProviderAmazonBedrock, ModelName: "bedrock/anthropic.claude-sonnet-4-20250514-v1:0"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/claude-sonnet-4@20250514"},
++ {Provider: ModelProviderOpenRouter, ModelName: "anthropic/claude-sonnet-4"},
+ },
+ },
-
+ {
+ ModelTag: "anthropic/claude-3.7-sonnet",
+ Publisher: ModelPublisherAnthropic,
+ Description: "Anthropic Claude 3.7 Sonnet",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- PreferredOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 128000,
++ ReservedOutputTokens: 20000, SupportsCacheControl: true,
++ PreferredOutputFormat: ModelOutputFormatXml, SingleMessageNoSystemPrompt: true,
++ TokenEstimatePaddingPct: 0.10,
+ },
-
+ Variants: []BaseModelConfigVariant{
++ {IsBaseVariant: true},
+ {
- IsBaseVariant: true,
- },
- {
- VariantTag: "thinking",
- Description: "thinking",
- Overrides: BaseModelShared{
- ReasoningBudget: AnthropicMaxReasoningBudget,
- },
-
++ VariantTag: "thinking", Description: "thinking",
++ Overrides: BaseModelShared{ReasoningBudget: AnthropicMaxReasoningBudget},
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "visible",
- Description: "visible",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "hidden",
- Description: "hidden",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "visible", Description: "visible", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "hidden", Description: "hidden", Overrides: BaseModelShared{IncludeReasoning: false}},
+ },
+ },
},
-
- },
- {
- Description: "OpenAI o4-mini-medium",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o4-mini",
- ModelId: "openai/o4-mini-medium",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortMedium,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderAnthropic,
- ModelName: "anthropic/claude-3-7-sonnet-latest",
- },
- {
- Provider: ModelProviderAmazonBedrock,
- ModelName: "bedrock/anthropic.claude-3-7-sonnet-20250219-v1:0",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/claude-3-7-sonnet@20250219",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.7-sonnet",
- },
++ {Provider: ModelProviderAnthropic, ModelName: "anthropic/claude-3-7-sonnet-latest"},
++ {Provider: ModelProviderAmazonBedrock, ModelName: "bedrock/anthropic.claude-3-7-sonnet-20250219-v1:0"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/claude-3-7-sonnet@20250219"},
++ {Provider: ModelProviderOpenRouter, ModelName: "anthropic/claude-3.7-sonnet"},
},
},
-
{
- Description: "OpenAI o4-mini-low",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "o4-mini",
- ModelId: "openai/o4-mini-low",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortLow,
- StopDisabled: true,
+ ModelTag: "anthropic/claude-3.5-sonnet",
+ Publisher: ModelPublisherAnthropic,
+ Description: "Anthropic Claude 3.5 Sonnet",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- PreferredOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 128000,
++ ReservedOutputTokens: 20000, SupportsCacheControl: true,
++ PreferredOutputFormat: ModelOutputFormatXml, SingleMessageNoSystemPrompt: true,
++ TokenEstimatePaddingPct: 0.10,
},
- },
--
- {
- Description: "OpenAI gpt-4.1",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1",
- ModelId: "openai/gpt-4.1",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderAnthropic,
- ModelName: "anthropic/claude-3-5-sonnet-latest",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/claude-3-5-sonnet@20240620",
- },
- {
- Provider: ModelProviderAmazonBedrock,
- ModelName: "bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.5-sonnet",
- },
++ {Provider: ModelProviderAnthropic, ModelName: "anthropic/claude-3-5-sonnet-latest"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/claude-3-5-sonnet@20240620"},
++ {Provider: ModelProviderAmazonBedrock, ModelName: "bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"},
++ {Provider: ModelProviderOpenRouter, ModelName: "anthropic/claude-3.5-sonnet"},
},
},
--
{
- Description: "OpenAI gpt-4.1-mini",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1-mini",
- ModelId: "openai/gpt-4.1-mini",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
+ ModelTag: "anthropic/claude-3.5-haiku",
+ Publisher: ModelPublisherAnthropic,
+ Description: "Anthropic Claude 3.5 Haiku",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 200000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- SupportsCacheControl: true,
- PreferredOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 200000, MaxOutputTokens: 8192,
++ ReservedOutputTokens: 8192, SupportsCacheControl: true,
++ PreferredOutputFormat: ModelOutputFormatXml, SingleMessageNoSystemPrompt: true,
++ TokenEstimatePaddingPct: 0.10,
},
- },
--
- {
- Description: "OpenAI gpt-4.1-nano",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenAI,
- ModelName: "gpt-4.1-nano",
- ModelId: "openai/gpt-4.1-nano",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenAIEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: OpenAIV1BaseUrl,
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderAnthropic,
- ModelName: "anthropic/claude-3-5-haiku-latest",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/claude-3-5-haiku@20241022",
- },
- {
- Provider: ModelProviderAmazonBedrock,
- ModelName: "bedrock/anthropic.claude-3-5-haiku-20241022-v1:0",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.5-haiku",
- },
++ {Provider: ModelProviderAnthropic, ModelName: "anthropic/claude-3-5-haiku-latest"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/claude-3-5-haiku@20241022"},
++ {Provider: ModelProviderAmazonBedrock, ModelName: "bedrock/anthropic.claude-3-5-haiku-20241022-v1:0"},
++ {Provider: ModelProviderOpenRouter, ModelName: "anthropic/claude-3.5-haiku"},
},
},
--
- // OpenRouter models
{
- Description: "Anthropic Claude 4 Sonnet via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-sonnet-4",
- ModelId: "anthropic/claude-sonnet-4",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ ModelTag: "google/gemini-pro-1.5",
+ Publisher: ModelPublisherGoogle,
+ Description: "Google Gemini 1.5 Pro",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 2000000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- // SupportsCacheControl: true, // gemini now uses implicit caching
++ DefaultMaxConvoTokens: 75000, MaxTokens: 2000000,
++ MaxOutputTokens: 8192, ReservedOutputTokens: 8192,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "Anthropic Claude 4 Opus via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-opus-4",
- ModelId: "anthropic/claude-opus-4",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 40000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderGoogleAIStudio,
- ModelName: "gemini/gemini-1.5-pro",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/gemini-1.5-pro-latest",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-pro-1.5",
- },
++ {Provider: ModelProviderGoogleAIStudio, ModelName: "gemini/gemini-1.5-pro"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/gemini-1.5-pro-latest"},
++ {Provider: ModelProviderOpenRouter, ModelName: "google/gemini-pro-1.5"},
},
},
{
- Description: "Anthropic Claude 4 Sonnet (thinking—includes reasoning) via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-sonnet-4",
- ModelId: "anthropic/claude-sonnet-4:thinking",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 40000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
- ReasoningBudgetEnabled: true,
- ReasoningBudget: AnthropicMaxReasoningBudget,
- IncludeReasoning: true,
+ ModelTag: "google/gemini-2.5-pro-preview",
+ Publisher: ModelPublisherGoogle,
+ Description: "Google Gemini 2.5 Pro (Preview)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- // SupportsCacheControl: true, // gemini now uses implicit caching
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1048576,
++ MaxOutputTokens: 65535, ReservedOutputTokens: 65535,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "Anthropic Claude 4 Sonnet (thinking-reasoning hidden) via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-sonnet-4",
- ModelId: "anthropic/claude-sonnet-4:thinking-hidden",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
- ReasoningBudgetEnabled: true,
- ReasoningBudget: AnthropicMaxReasoningBudget,
- IncludeReasoning: false,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderGoogleAIStudio,
- ModelName: "gemini/gemini-2.5-pro-preview-05-06",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/gemini-2.5-pro-preview-05-06",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-pro-preview",
- },
++ {Provider: ModelProviderGoogleAIStudio, ModelName: "gemini/gemini-2.5-pro-preview-05-06"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/gemini-2.5-pro-preview-05-06"},
++ {Provider: ModelProviderOpenRouter, ModelName: "google/gemini-2.5-pro-preview"},
},
},
{
- Description: "Anthropic Claude 3.7 Sonnet via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.7-sonnet",
- ModelId: "anthropic/claude-3.7-sonnet",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ ModelTag: "google/gemini-2.5-pro-exp",
+ Publisher: ModelPublisherGoogle,
+ Description: "Google Gemini 2.5 Pro (Experimental)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1048576,
++ MaxOutputTokens: 65535, ReservedOutputTokens: 65535,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "Anthropic Claude 3.7 Sonnet (thinking—includes reasoning) via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.7-sonnet:thinking",
- ModelId: "anthropic/claude-3.7-sonnet:thinking",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 40000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- IncludeReasoning: true,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderGoogleAIStudio,
- ModelName: "gemini/gemini-2.5-pro-exp-03-25",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/gemini-2.5-pro-exp-03-25",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-pro-exp-03-25",
- },
++ {Provider: ModelProviderGoogleAIStudio, ModelName: "gemini/gemini-2.5-pro-exp-03-25"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/gemini-2.5-pro-exp-03-25"},
++ {Provider: ModelProviderOpenRouter, ModelName: "google/gemini-2.5-pro-exp-03-25"},
},
},
{
- Description: "Anthropic Claude 3.7 Sonnet (thinking—reasoning hidden) via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.7-sonnet:thinking",
- ModelId: "anthropic/claude-3.7-sonnet:thinking-hidden",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 40000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- IncludeReasoning: false,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ ModelTag: "google/gemini-2.5-flash-preview",
+ Publisher: ModelPublisherGoogle,
+ Description: "Google Gemini 2.5 Flash (Preview)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 75000,
-
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- // SupportsCacheControl: true, // gemini now uses implicit caching
++ DefaultMaxConvoTokens: 75000, MaxTokens: 1048576,
++ MaxOutputTokens: 65535, ReservedOutputTokens: 65535,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "Anthropic Claude 3.5 Sonnet via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.5-sonnet",
- ModelId: "anthropic/claude-3.5-sonnet",
- MaxTokens: 200000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 20000,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ Variants: []BaseModelConfigVariant{
- {
- IsBaseVariant: true,
- },
- {
- VariantTag: "thinking",
- Description: "thinking",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
++ {IsBaseVariant: true},
++ {VariantTag: "thinking", Description: "thinking", Overrides: BaseModelShared{IncludeReasoning: true}},
},
-
- },
- {
- Description: "Anthropic Claude 3.5 Haiku via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "anthropic/claude-3.5-haiku",
- ModelId: "anthropic/claude-3.5-haiku",
- MaxTokens: 200000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- SupportsCacheControl: true,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SingleMessageNoSystemPrompt: true,
- TokenEstimatePaddingPct: 0.10,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderGoogleAIStudio,
- ModelName: "gemini/gemini-2.5-flash-preview-05-20",
- },
- {
- Provider: ModelProviderGoogleVertex,
- ModelName: "vertex_ai/gemini-2.5-flash-preview-05-20",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-flash-preview-05-20",
- },
++ {Provider: ModelProviderGoogleAIStudio, ModelName: "gemini/gemini-2.5-flash-preview-05-20"},
++ {Provider: ModelProviderGoogleVertex, ModelName: "vertex_ai/gemini-2.5-flash-preview-05-20"},
++ {Provider: ModelProviderOpenRouter, ModelName: "google/gemini-2.5-flash-preview-05-20"},
},
},
-
{
- Description: "Google Gemini Pro 1.5 via OpenRouter",
- DefaultMaxConvoTokens: 100000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-pro-1.5",
- ModelId: "google/gemini-pro-1.5",
- MaxTokens: 2000000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SupportsCacheControl: true,
+ ModelTag: "deepseek/v3-0324",
+ Publisher: ModelPublisherDeepSeek,
+ Description: "DeepSeek V3 (0324)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 7500,
-
- MaxTokens: 64000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
++ DefaultMaxConvoTokens: 7500, MaxTokens: 64000,
++ MaxOutputTokens: 8192, ReservedOutputTokens: 8192,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
- },
--
- {
- Description: "Google Gemini Pro 2.5 Preview via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-pro-preview",
- ModelId: "google/gemini-2.5-pro-preview",
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SupportsCacheControl: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderDeepSeek,
- ModelName: "deepseek/deepseek-chat",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "deepseek/deepseek-chat-v3-0324",
- },
++ {Provider: ModelProviderDeepSeek, ModelName: "deepseek/deepseek-chat"},
++ {Provider: ModelProviderOpenRouter, ModelName: "deepseek/deepseek-chat-v3-0324"},
},
},
-
{
- Description: "Google Gemini Pro 2.5 Experimental via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-pro-exp-03-25",
- ModelId: "google/gemini-2.5-pro-exp-03-25",
- MaxTokens: 1000000,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
+ ModelTag: "deepseek/r1",
+ Publisher: ModelPublisherDeepSeek,
+ Description: "DeepSeek R1",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 7500,
-
- MaxTokens: 64000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
++ DefaultMaxConvoTokens: 7500, MaxTokens: 64000,
++ MaxOutputTokens: 8192, ReservedOutputTokens: 8192,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "Google Gemini Flash 2.5 Preview via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-flash-preview",
- ModelId: "google/gemini-2.5-flash-preview",
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SupportsCacheControl: true,
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "reasoning-visible",
- Description: "(reasoning visible)",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "reasoning-hidden",
- Description: "(reasoning hidden)",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "reasoning-visible", Description: "(reasoning visible)", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "reasoning-hidden", Description: "(reasoning hidden)", Overrides: BaseModelShared{IncludeReasoning: false}},
},
- },
--
- {
- Description: "Google Gemini Flash 2.5 Preview (thinking-includes reasoning) via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-flash-preview-05-20:thinking",
- ModelId: "google/gemini-2.5-flash-preview-05-20:thinking",
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SupportsCacheControl: true,
- ReasoningBudgetEnabled: true,
- ReasoningBudget: GoogleMaxReasoningBudget,
- IncludeReasoning: true,
+ Providers: []BaseModelUsesProvider{
- {
- Provider: ModelProviderDeepSeek,
- ModelName: "deepseek/deepseek-reasoner",
- },
- {
- Provider: ModelProviderOpenRouter,
- ModelName: "deepseek/deepseek-r1",
- },
++ {Provider: ModelProviderDeepSeek, ModelName: "deepseek/deepseek-reasoner"},
++ {Provider: ModelProviderOpenRouter, ModelName: "deepseek/deepseek-r1"},
},
},
--
{
- Description: "Google Gemini Flash 2.5 Preview (thinking-reasoning hidden) via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "google/gemini-2.5-flash-preview-05-20:thinking",
- ModelId: "google/gemini-2.5-flash-preview-05-20:thinking-hidden",
- MaxTokens: 1048576,
- MaxOutputTokens: 65535,
- ReservedOutputTokens: 65535,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SupportsCacheControl: true,
- ReasoningBudgetEnabled: true,
- ReasoningBudget: GoogleMaxReasoningBudget,
- IncludeReasoning: false,
+ ModelTag: "perplexity/r1-1776",
+ Publisher: ModelPublisherPerplexity,
+ Description: "Perplexity R1-1776",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 7500,
-
- MaxTokens: 128000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 30000,
++ DefaultMaxConvoTokens: 7500, MaxTokens: 128000,
++ MaxOutputTokens: 128000, ReservedOutputTokens: 30000,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
- },
--
- {
- Description: "DeepSeek V3 0324 via OpenRouter",
- DefaultMaxConvoTokens: 7500,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "deepseek/deepseek-chat-v3-0324",
- ModelId: "deepseek/deepseek-chat-v3-0324",
- MaxTokens: 64000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: ModelCompatibility{
- HasImageSupport: false,
- },
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "reasoning-visible",
- Description: "(reasoning visible)",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "reasoning-hidden",
- Description: "(reasoning hidden)",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "reasoning-visible", Description: "(reasoning visible)", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "reasoning-hidden", Description: "(reasoning hidden)", Overrides: BaseModelShared{IncludeReasoning: false}},
},
- },
--
- {
- Description: "DeepSeek R1 via OpenRouter (includes reasoning)",
- DefaultMaxConvoTokens: 7500,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "deepseek/deepseek-r1",
- ModelId: "deepseek/deepseek-r1-reasoning",
- MaxTokens: 64000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: ModelCompatibility{
- HasImageSupport: false,
- },
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- IncludeReasoning: true,
+ Providers: []BaseModelUsesProvider{
+ {Provider: ModelProviderPerplexity, ModelName: "r1-1776-online"},
+ {Provider: ModelProviderOpenRouter, ModelName: "perplexity/r1-1776"},
},
},
-
{
- Description: "DeepSeek R1 via OpenRouter (reasoning hidden)",
- DefaultMaxConvoTokens: 7500,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "deepseek/deepseek-r1",
- ModelId: "deepseek/deepseek-r1-no-reasoning",
- MaxTokens: 64000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: ModelCompatibility{
- HasImageSupport: false,
- },
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
+ ModelTag: "perplexity/sonar-reasoning",
- Description: "Perplexity Sonar Reasoning",
+ Publisher: ModelPublisherPerplexity,
-
++ Description: "Perplexity Sonar Reasoning",
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 7500,
-
- MaxTokens: 127000,
- MaxOutputTokens: 127000,
- ReservedOutputTokens: 30000,
++ DefaultMaxConvoTokens: 7500, MaxTokens: 127000,
++ MaxOutputTokens: 127000, ReservedOutputTokens: 30000,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
- },
--
- {
- Description: "Perplexity R1 1776 via OpenRouter (includes reasoning)",
- DefaultMaxConvoTokens: 7500,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "perplexity/r1-1776",
- ModelId: "perplexity/r1-1776",
- MaxTokens: 128000,
- MaxOutputTokens: 128000,
- ReservedOutputTokens: 30000,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: ModelCompatibility{
- HasImageSupport: false,
- },
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- IncludeReasoning: true,
+ Variants: []BaseModelConfigVariant{
- {
- VariantTag: "visible",
- Description: "(reasoning visible)",
- Overrides: BaseModelShared{
- IncludeReasoning: true,
- },
- },
- {
- VariantTag: "hidden",
- Description: "(reasoning hidden)",
- Overrides: BaseModelShared{
- IncludeReasoning: false,
- },
- },
++ {VariantTag: "visible", Description: "(reasoning visible)", Overrides: BaseModelShared{IncludeReasoning: true}},
++ {VariantTag: "hidden", Description: "(reasoning hidden)", Overrides: BaseModelShared{IncludeReasoning: false}},
},
- },
--
- {
- Description: "Perplexity Sonar Reasoning via OpenRouter (includes reasoning)",
- DefaultMaxConvoTokens: 7500,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "perplexity/sonar-reasoning",
- ModelId: "perplexity/sonar-reasoning",
- MaxTokens: 127000,
- MaxOutputTokens: 127000,
- ReservedOutputTokens: 30000,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: ModelCompatibility{
- HasImageSupport: false,
- },
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- IncludeReasoning: true,
+ Providers: []BaseModelUsesProvider{
+ {Provider: ModelProviderPerplexity, ModelName: "sonar-reasoning-online"},
+ {Provider: ModelProviderOpenRouter, ModelName: "perplexity/sonar-reasoning"},
},
},
--
{
- Description: "Qwen 2.5 Coder 32B via OpenRouter",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "qwen/qwen-2.5-coder-32b-instruct",
- ModelId: "qwen/qwen-2.5-coder-32b-instruct",
- MaxTokens: 128000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
- ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
+ ModelTag: "qwen/qwen-2.5-coder-32b-instruct",
+ Publisher: ModelPublisherQwen,
+ Description: "Qwen 2.5 Coder 32B (Instruct)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 10000,
-
- MaxTokens: 128000,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
++ DefaultMaxConvoTokens: 10000, MaxTokens: 128000,
++ MaxOutputTokens: 8192, ReservedOutputTokens: 8192,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
- },
-
- // OpenAI models via OpenRouter
-
- {
- Description: "OpenAI o3-high via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o3",
- ModelId: "openai/o3-high",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
- },
- },
--
- {
- Description: "OpenAI o3-medium via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o3",
- ModelId: "openai/o3-medium",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortMedium,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
+ {Provider: ModelProviderOpenRouter, ModelName: "qwen/qwen-2.5-coder-32b-instruct"},
},
},
--
{
- Description: "OpenAI o3-low via OpenRouter",
- DefaultMaxConvoTokens: 15000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o3",
- ModelId: "openai/o3-low",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatXml,
- SystemPromptDisabled: true,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortLow,
- StopDisabled: true,
+ ModelTag: "qwen/qwen3-32b",
+ Publisher: ModelPublisherQwen,
+ Description: "Qwen 3-32B (Experimental)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 15000,
- MaxTokens: 128000,
- MaxOutputTokens: 16384,
- ReservedOutputTokens: 16384,
++ DefaultMaxConvoTokens: 15000, MaxTokens: 128000,
++ MaxOutputTokens: 16384, ReservedOutputTokens: 16384,
+ PreferredOutputFormat: ModelOutputFormatXml,
- // leave IncludeReasoning unset – Qwen3 has its own “thinking” mode but OR
- // exposes it via a special prompt key, not a flag
},
- },
--
- {
- Description: "OpenAI o4-mini-high via OpenRouter",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o4-mini",
- ModelId: "openai/o4-mini-high",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 30000,
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortHigh,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
- {Provider: ModelProviderOpenRouter, ModelName: "qwen/qwen3-32b-a3b"}, // OR cloud
- // optional local provider once you wire up Ollama or HF Inference
- // {Provider: ModelProviderOllama, ModelName: "qwen3-32b"},
++ {Provider: ModelProviderOpenRouter, ModelName: "qwen/qwen3-32b-a3b"},
},
},
{
- Description: "OpenAI o4-mini-medium via OpenRouter",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o4-mini",
- ModelId: "openai/o4-mini-medium",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortMedium,
- StopDisabled: true,
+ ModelTag: "qwen/qwen3-8b",
+ Publisher: ModelPublisherQwen,
+ Description: "Qwen 3-8B (Experimental)",
-
+ BaseModelShared: BaseModelShared{
- DefaultMaxConvoTokens: 7500,
- MaxTokens: 32768,
- MaxOutputTokens: 8192,
- ReservedOutputTokens: 8192,
++ DefaultMaxConvoTokens: 7500, MaxTokens: 32768,
++ MaxOutputTokens: 8192, ReservedOutputTokens: 8192,
+ PreferredOutputFormat: ModelOutputFormatXml,
},
-
- },
- {
- Description: "OpenAI o4-mini-low via OpenRouter",
- DefaultMaxConvoTokens: 10000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/o4-mini",
- ModelId: "openai/o4-mini-low",
- MaxTokens: 200000,
- MaxOutputTokens: 100000,
- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- RoleParamsDisabled: true,
- ReasoningEffortEnabled: true,
- ReasoningEffort: ReasoningEffortLow,
- StopDisabled: true,
+ Providers: []BaseModelUsesProvider{
+ {Provider: ModelProviderOpenRouter, ModelName: "qwen/qwen3-8b-04-28"},
},
},
+}
- {
- Description: "OpenAI gpt-4.1 via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1",
- ModelId: "openai/gpt-4.1",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- },
- },
+var BuiltInBaseModelsById = map[ModelId]*BaseModelConfigSchema{}
- {
- Description: "OpenAI gpt-4.1-mini via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1-mini",
- ModelId: "openai/gpt-4.1-mini",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- },
- },
+var BuiltInModelProvidersByModelId = map[ModelId][]BaseModelUsesProvider{}
+var BuiltInBaseModels = []*BaseModelConfigSchema{}
- {
- Description: "OpenAI gpt-4.1-nano via OpenRouter",
- DefaultMaxConvoTokens: 75000,
- BaseModelConfig: BaseModelConfig{
- Provider: ModelProviderOpenRouter,
- ModelName: "openai/gpt-4.1-nano",
- ModelId: "openai/gpt-4.1-nano",
- MaxTokens: 1047576,
- MaxOutputTokens: 32768,
- ReservedOutputTokens: 32768,
- ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
- ModelCompatibility: fullCompatibility,
- BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],
- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,
- },
- },
-}
+var AvailableModels = []*AvailableModel{}
var AvailableModelsByComposite = map[string]*AvailableModel{}
diff --cc app/shared/ai_models_data_models.go
index 29caf578,a95a3d15..17af1ff8
--- a/app/shared/ai_models_data_models.go
+++ b/app/shared/ai_models_data_models.go
@@@ -19,212 -18,47 +19,217 @@@ const
ModelOutputFormatXml ModelOutputFormat = "xml"
)
-const AnthropicMaxReasoningBudget = 32000
-const GoogleMaxReasoningBudget = 32000
-
-// to help avoid confusion between the model name and the model id
+// to help avoid confusion between model tag, price id, model name, and the model id
type ModelName string
type ModelId string
+type ModelTag string
+type VariantTag string
+
+type BaseModelShared struct {
- DefaultMaxConvoTokens int `json:"defaultMaxConvoTokens"`
- MaxTokens int `json:"maxTokens"`
- MaxOutputTokens int `json:"maxOutputTokens"`
- ReservedOutputTokens int `json:"reservedOutputTokens"`
- PreferredOutputFormat ModelOutputFormat `json:"preferredOutputFormat"`
- SystemPromptDisabled bool `json:"systemPromptDisabled,omitempty"`
- RoleParamsDisabled bool `json:"roleParamsDisabled,omitempty"`
- StopDisabled bool `json:"stopDisabled,omitempty"`
- PredictedOutputEnabled bool `json:"predictedOutputEnabled,omitempty"`
- ReasoningEffortEnabled bool `json:"reasoningEffortEnabled,omitempty"`
- ReasoningEffort ReasoningEffort `json:"reasoningEffort,omitempty"`
- IncludeReasoning bool `json:"includeReasoning,omitempty"`
- ReasoningBudget int `json:"reasoningBudget,omitempty"`
- SupportsCacheControl bool `json:"supportsCacheControl,omitempty"`
- // for anthropic, single message system prompt needs to be flipped to 'user'
- SingleMessageNoSystemPrompt bool `json:"singleMessageNoSystemPrompt,omitempty"`
-
- // for anthropic, token estimate padding percentage
- TokenEstimatePaddingPct float64 `json:"tokenEstimatePaddingPct,omitempty"`
-
++ DefaultMaxConvoTokens int `json:"defaultMaxConvoTokens"`
++ MaxTokens int `json:"maxTokens"`
++ MaxOutputTokens int `json:"maxOutputTokens"`
++ ReservedOutputTokens int `json:"reservedOutputTokens"`
++ PreferredOutputFormat ModelOutputFormat `json:"preferredOutputFormat"`
++ SystemPromptDisabled bool `json:"systemPromptDisabled,omitempty"`
++ RoleParamsDisabled bool `json:"roleParamsDisabled,omitempty"`
++ StopDisabled bool `json:"stopDisabled,omitempty"`
++ PredictedOutputEnabled bool `json:"predictedOutputEnabled,omitempty"`
++ ReasoningEffortEnabled bool `json:"reasoningEffortEnabled,omitempty"`
++ ReasoningEffort ReasoningEffort `json:"reasoningEffort,omitempty"`
++ IncludeReasoning bool `json:"includeReasoning,omitempty"`
++ ReasoningBudget int `json:"reasoningBudget,omitempty"`
++ SupportsCacheControl bool `json:"supportsCacheControl,omitempty"`
++ SingleMessageNoSystemPrompt bool `json:"singleMessageNoSystemPrompt,omitempty"`
++ TokenEstimatePaddingPct float64 `json:"tokenEstimatePaddingPct,omitempty"`
+ ModelCompatibility
+}
+
+type BaseModelProviderConfig struct {
+ ModelProviderConfigSchema
+ ModelName ModelName `json:"modelName"`
+}
type BaseModelConfig struct {
- ModelId ModelId `json:"modelId"`
- Provider ModelProvider `json:"provider"`
- CustomProvider *string `json:"customProvider,omitempty"`
- BaseUrl string `json:"baseUrl"`
- ModelName ModelName `json:"modelName"`
- ModelId ModelId `json:"modelId"`
- MaxTokens int `json:"maxTokens"`
- MaxOutputTokens int `json:"maxOutputTokens"`
- ReservedOutputTokens int `json:"reservedOutputTokens"`
- ApiKeyEnvVar string `json:"apiKeyEnvVar"`
- PreferredModelOutputFormat ModelOutputFormat `json:"preferredModelOutputFormat"`
- SystemPromptDisabled bool `json:"systemPromptDisabled"`
- RoleParamsDisabled bool `json:"roleParamsDisabled"`
- StopDisabled bool `json:"stopDisabled"`
- PredictedOutputEnabled bool `json:"predictedOutputEnabled"`
- ReasoningEffortEnabled bool `json:"reasoningEffortEnabled"`
- ReasoningEffort ReasoningEffort `json:"reasoningEffort"`
- IncludeReasoning bool `json:"includeReasoning"`
-
- ReasoningBudgetEnabled bool `json:"reasoningBudgetEnabled"`
- ReasoningBudget int `json:"reasoningBudget"`
-
- SupportsCacheControl bool `json:"supportsCacheControl"`
-
- // for openai responses API, not fully implemented yet
- UsesOpenAIResponsesAPI bool `json:"usesOpenAIResponsesAPI"`
-
- // for anthropic, single message system prompt needs to be flipped to 'user'
- SingleMessageNoSystemPrompt bool `json:"singleMessageNoSystemPrompt"`
-
- // for openai, token estimate padding percentage
- TokenEstimatePaddingPct float64 `json:"tokenEstimatePaddingPct"`
+ ModelTag ModelTag `json:"modelTag"`
- Publisher ModelPublisher `json:"publisher"`
++ ModelId ModelId `json:"modelId"`
++ Publisher ModelPublisher `json:"publisher,omitempty"`
+ BaseModelShared
+ BaseModelProviderConfig
++ ApiKeyEnvVar string `json:"apiKeyEnvVar,omitempty"`
++ PreferredModelOutputFormat ModelOutputFormat `json:"preferredModelOutputFormat,omitempty"`
++ ReasoningBudgetEnabled bool `json:"reasoningBudgetEnabled,omitempty"`
++ ReasoningBudget int `json:"reasoningBudget,omitempty"`
++ SupportsCacheControl bool `json:"supportsCacheControl,omitempty"`
++ UsesOpenAIResponsesAPI bool `json:"usesOpenAIResponsesAPI,omitempty"`
+}
- ModelCompatibility
+type BaseModelUsesProvider struct {
+ Provider ModelProvider `json:"provider"`
+ CustomProvider *string `json:"customProvider,omitempty"`
+ ModelName ModelName `json:"modelName"`
+}
+
+func (b BaseModelUsesProvider) ToComposite() string {
+ if b.CustomProvider != nil {
+ return fmt.Sprintf("%s|%s", b.Provider, *b.CustomProvider)
+ }
+ return string(b.Provider)
+}
+
+type BaseModelConfigSchema struct {
+ ModelTag ModelTag `json:"modelTag"`
+ ModelId ModelId `json:"modelId"`
+ Publisher ModelPublisher `json:"publisher"`
+ Description string `json:"description"`
+
+ BaseModelShared
+
+ RequiresVariantOverrides []string `json:"requiresVariantOverrides"`
+
+ Variants []BaseModelConfigVariant `json:"variants"`
+ Providers []BaseModelUsesProvider `json:"providers"`
+}
+
+type BaseModelConfigVariant struct {
+ IsBaseVariant bool `json:"isBaseVariant"`
+ VariantTag VariantTag `json:"variantTag"`
+ Description string `json:"description"`
+ Overrides BaseModelShared `json:"overrides"`
+ Variants []BaseModelConfigVariant `json:"variants"`
+ RequiresVariantOverrides []string `json:"requiresVariantOverrides"`
+}
+
+func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {
+ avail := []*AvailableModel{}
+ for _, provider := range b.Providers {
+
+ providerConfig, ok := BuiltInModelProviderConfigs[provider.Provider]
+ if !ok {
+ panic(fmt.Sprintf("provider %s not found", provider.Provider))
+ }
+
+ addBase := func() {
+ avail = append(avail, &AvailableModel{
+ Description: b.Description,
+ DefaultMaxConvoTokens: b.DefaultMaxConvoTokens,
+ BaseModelConfig: BaseModelConfig{
+ ModelTag: b.ModelTag,
+ ModelId: ModelId(string(b.ModelTag)),
+ BaseModelShared: b.BaseModelShared,
+ BaseModelProviderConfig: BaseModelProviderConfig{
+ ModelProviderConfigSchema: providerConfig,
+ ModelName: provider.ModelName,
+ },
++ ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,
+ },
+ })
+ }
+
+ type variantParams struct {
+ BaseVariant *BaseModelConfigVariant
+ BaseId ModelId
+ BaseDescription string
+ Overrides BaseModelShared
+ RequiresVariantOverrides []string
+ }
+
+ addBaseVariant := func(params variantParams) {
+ if params.BaseVariant == nil {
+ addBase()
+ return
+ }
+ baseDescription := params.BaseVariant.Description
+ baseId := params.BaseId
+ mergedOverrides := params.Overrides
+
+ avail = append(avail, &AvailableModel{
+ Description: baseDescription,
+ DefaultMaxConvoTokens: b.DefaultMaxConvoTokens,
+ BaseModelConfig: BaseModelConfig{
+ ModelTag: b.ModelTag,
+ ModelId: baseId,
+ Publisher: b.Publisher,
+ BaseModelShared: mergedOverrides,
+ BaseModelProviderConfig: BaseModelProviderConfig{
+ ModelProviderConfigSchema: providerConfig,
+ ModelName: provider.ModelName,
+ },
++ ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,
+ },
+ })
+ }
+
+ if len(b.Variants) == 0 {
+ addBase()
+ } else {
+
+ var addVariants func(variants []BaseModelConfigVariant, baseParams variantParams)
+ addVariants = func(variants []BaseModelConfigVariant, baseParams variantParams) {
+ for _, variant := range variants {
+ if variant.IsBaseVariant {
+ addBaseVariant(baseParams)
+ continue
+ }
+
+ if len(baseParams.RequiresVariantOverrides) > 0 {
+ ok, missing := FieldsDefined(variant.Overrides, baseParams.RequiresVariantOverrides)
+ if !ok {
+ panic(fmt.Sprintf("variant %s is missing required field %s", variant.VariantTag, missing))
+ }
+ }
+
+ var baseId ModelId
+ var baseDescription string
+ if baseParams.BaseId != "" {
+ baseId = baseParams.BaseId
+ baseDescription = baseParams.BaseDescription
+ } else {
+ baseId = ModelId(string(b.ModelTag))
+ baseDescription = b.Description
+ }
+
+ modelId := ModelId(strings.Join([]string{string(baseId), string(variant.VariantTag)}, "-"))
+
+ description := strings.Join([]string{baseDescription, variant.Description}, " ")
+
+ merged := Merge(b.BaseModelShared, variant.Overrides)
+
+ if len(variant.Variants) > 0 {
+ addVariants(variant.Variants, variantParams{
+ BaseVariant: &variant,
+ BaseId: modelId,
+ BaseDescription: description,
+ Overrides: merged,
+ RequiresVariantOverrides: variant.RequiresVariantOverrides,
+ })
+ }
+ avail = append(avail, &AvailableModel{
+ Description: description,
+ DefaultMaxConvoTokens: b.DefaultMaxConvoTokens,
+ BaseModelConfig: BaseModelConfig{
+ ModelTag: b.ModelTag,
+ ModelId: modelId,
+ BaseModelShared: merged,
+ BaseModelProviderConfig: BaseModelProviderConfig{
+ ModelProviderConfigSchema: providerConfig,
+ ModelName: provider.ModelName,
+ },
++ ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,
+ },
+ })
+ }
+ }
+
+ addVariants(b.Variants, variantParams{
+ BaseId: ModelId(string(b.ModelTag)),
+ BaseDescription: b.Description,
+ Overrides: b.BaseModelShared,
+ RequiresVariantOverrides: b.RequiresVariantOverrides,
+ })
+ }
+ }
+ return avail
}
type AvailableModel struct {
diff --cc app/shared/ai_models_packs.go
index e50ad855,85fd82a9..74f9a425
--- a/app/shared/ai_models_packs.go
+++ b/app/shared/ai_models_packs.go
@@@ -23,279 -23,466 +23,282 @@@ var BuiltInModelPacks = []*ModelPack
&StrongModelPack,
&CheapModelPack,
&OSSModelPack,
-
- &Sonnet4ModelPack,
- &Opus4PlannerModelPack,
--
+ &OpusPlannerModelPack,
+ &StrongModelOpus,
-
&AnthropicModelPack,
&OpenAIModelPack,
--
&GeminiPreviewModelPack,
&GeminiExperimentalModelPack,
--
&R1PlannerModelPack,
&PerplexityPlannerModelPack,
}
var DefaultModelPack *ModelPack = &DailyDriverModelPack
+func getModelRoleConfig(role ModelRole, modelId ModelId, fns ...func(*ModelRoleConfigSchema)) ModelRoleConfigSchema {
+ c := ModelRoleConfigSchema{
+ ModelId: modelId,
+ Temperature: DefaultConfigByRole[role].Temperature,
+ TopP: DefaultConfigByRole[role].TopP,
+ }
+ for _, f := range fns {
+ f(&c)
+ }
+ return c
+}
+
+func getLargeContextFallback(role ModelRole, modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {
+ return func(c *ModelRoleConfigSchema) {
+ n := getModelRoleConfig(role, modelId)
+ for _, f := range fns {
+ f(&n)
+ }
+ c.LargeContextFallback = &n
+ }
+}
+
+func getErrorFallback(role ModelRole, modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {
+ return func(c *ModelRoleConfigSchema) {
+ n := getModelRoleConfig(role, modelId)
+ for _, f := range fns {
+ f(&n)
+ }
+ c.ErrorFallback = &n
+ }
+}
+
+func getStrongModelFallback(role ModelRole, modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {
+ return func(c *ModelRoleConfigSchema) {
+ n := getModelRoleConfig(role, modelId)
+ for _, f := range fns {
+ f(&n)
+ }
+ c.StrongModel = &n
+ }
+}
+
+var (
+ DailyDriverPackSchema ModelPackSchema
+ ReasoningPackSchema ModelPackSchema
+ StrongPackSchema ModelPackSchema
+ OSSModelPackSchema ModelPackSchema
+ CheapModelPackSchema ModelPackSchema
+ AnthropicPackSchema ModelPackSchema
+ OpenAIPackSchema ModelPackSchema
+ GeminiPreviewPackSchema ModelPackSchema
+ GeminiExperimentalPackSchema ModelPackSchema
+ R1PlannerPackSchema ModelPackSchema
+ PerplexityPlannerPackSchema ModelPackSchema
+)
+
- // An ordered list you can iterate over if you need to display / validate.
+var BuiltInModelPackSchemas = []*ModelPackSchema{
+ &DailyDriverPackSchema,
+ &ReasoningPackSchema,
+ &StrongPackSchema,
+ &CheapModelPackSchema,
+ &OSSModelPackSchema,
+ &AnthropicPackSchema,
+ &OpenAIPackSchema,
+ &GeminiPreviewPackSchema,
+ &GeminiExperimentalPackSchema,
+ &R1PlannerPackSchema,
+ &PerplexityPlannerPackSchema,
+}
+
func init() {
- DailyDriverModelPack = ModelPack{
+ defaultBuilder := getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium",
+ getStrongModelFallback(ModelRoleBuilder, "openai/o4-mini-high"),
+ )
+
+ DailyDriverPackSchema = ModelPackSchema{
Name: "daily-driver",
- Description: "A mix of models from Anthropic, OpenAI, and Google that balances speed, quality, and cost. Supports up to 2M context. Plandex prompts are especially tested and optimized for this pack.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *claudeSonnet4With37Fallback(ModelRolePlanner, &modelConfig{
- largeContextFallback: geminipro25preview(ModelRolePlanner, &modelConfig{
- largeContextFallback: gemini15pro(ModelRolePlanner, nil),
- }),
- }),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-sonnet-4"),
- },
- Coder: claudeSonnet4With37Fallback(ModelRoleCoder, &modelConfig{
- largeContextFallback: openaigpt41(ModelRoleCoder, nil),
- }),
- Architect: claudeSonnet4With37Fallback(ModelRoleArchitect, &modelConfig{
- largeContextFallback: geminipro25preview(ModelRolePlanner, &modelConfig{
- largeContextFallback: gemini15pro(ModelRolePlanner, nil),
- }),
- }),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
+ Description: "A mix of models from Anthropic, OpenAI, and Google that balances speed, quality, and cost. Supports up to 2M context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-3.7-sonnet",
++ Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-sonnet-4",
+ getLargeContextFallback(ModelRolePlanner, "google/gemini-2.5-pro-preview",
+ getLargeContextFallback(ModelRolePlanner, "google/gemini-pro-1.5"),
+ ),
+ ),
- Architect: Pointer(getModelRoleConfig(ModelRoleArchitect, "anthropic/claude-3.7-sonnet",
++ Architect: Pointer(getModelRoleConfig(ModelRoleArchitect, "anthropic/claude-sonnet-4",
+ getLargeContextFallback(ModelRoleArchitect, "google/gemini-2.5-pro-preview",
+ getLargeContextFallback(ModelRoleArchitect, "google/gemini-pro-1.5"),
+ ),
+ )),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet",
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4",
+ getLargeContextFallback(ModelRoleCoder, "openai/gpt-4.1"),
+ )),
+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
+ Builder: defaultBuilder,
+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium")),
+ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
+ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
+ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- ReasoningModelPack = ModelPack{
+ ReasoningPackSchema = ModelPackSchema{
- Name: "reasoning",
- Description: "Like the daily driver, but uses 3.7-sonnet:thinking with reasoning enabled for planning and coding. Supports up to 160k input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-3.7-sonnet-thinking-hidden"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet-thinking-hidden")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: defaultBuilder,
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
+ Name: "reasoning",
- Description: "Like the daily driver, but uses sonnet-4 with reasoning enabled for planning and coding. Supports up to 160k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *claudeSonnet4ThinkingHidden(ModelRolePlanner, &modelConfig{
- errorFallback: claude37SonnetThinkingHidden(ModelRolePlanner, nil),
- }),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-sonnet-4:thinking-hidden"),
- },
- Coder: claudeSonnet4ThinkingHidden(ModelRoleCoder, &modelConfig{
- errorFallback: claude37SonnetThinkingHidden(ModelRoleCoder, nil),
- }),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
-
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
++ Description: "Like the daily driver, but uses sonnet-4-thinking with reasoning enabled for planning and coding. Supports up to 160k input context.",
++ Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-sonnet-4-thinking-hidden"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4-thinking-hidden")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: defaultBuilder,
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-medium")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- StrongModelPack = ModelPack{
+ StrongPackSchema = ModelPackSchema{
- Name: "strong",
- Description: "For difficult tasks where slower responses and builds are ok. Uses o3-high for architecture and planning, claude-3.7-sonnet thinking for implementation, prioritizes reliability over speed for builds. Supports up to 160k input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "openai/o3-high"),
- Architect: Pointer(getModelRoleConfig(ModelRoleArchitect, "openai/o3-high")),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet-thinking-hidden")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-high"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-high")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
+ Name: "strong",
- Description: "For difficult tasks where slower responses and builds are ok. Uses o3-high for architecture and planning, claude-sonnet-4 thinking for implementation, prioritizes reliability over speed for builds. Supports up to 160k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *openaio3high(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenAI, "openai/o3-high"),
- },
- Architect: openaio3high(ModelRoleArchitect, nil),
- Coder: claudeSonnet4ThinkingHidden(ModelRoleCoder, &modelConfig{
- errorFallback: claude37SonnetThinkingHidden(ModelRoleCoder, nil),
- }),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniHigh(ModelRoleBuilder, nil),
- WholeFileBuilder: openaio4miniHigh(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniMedium(ModelRoleExecStatus, nil),
- }
-
- StrongModelOpus = ModelPack{
- Name: "strong-opus",
- Description: "Like the strong pack, but uses Claude Opus 4 thinking for planning and coding. Supports up to 160k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *claudeOpus4(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-opus-4"),
- },
- Architect: claudeOpus4(ModelRoleArchitect, nil),
- Coder: claudeOpus4(ModelRoleCoder, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniHigh(ModelRoleBuilder, nil),
- WholeFileBuilder: openaio4miniHigh(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniMedium(ModelRoleExecStatus, nil),
++ Description: "For difficult tasks where slower responses and builds are ok. Uses o3-high for architecture and planning, claude-sonnet-4 thinking for implementation. Supports up to 160k input context.",
++ Planner: getModelRoleConfig(ModelRolePlanner, "openai/o3-high"),
++ Architect: Pointer(getModelRoleConfig(ModelRoleArchitect, "openai/o3-high")),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4-thinking-hidden")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-high"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-high")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
}
- CheapModelPack = ModelPack{
+ CheapModelPackSchema = ModelPackSchema{
- Name: "cheap",
- Description: "Cost-effective models that can still get the job done for easier tasks. Supports up to 160k context. Uses OpenAI's o4-mini model for planning, GPT-4.1 for coding, and GPT-4.1 Mini for lighter tasks.",
- Planner: getModelRoleConfig(ModelRolePlanner, "openai/o4-mini-medium"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "openai/gpt-4.1")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/gpt-4.1-mini"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-low"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
+ Name: "cheap",
+ Description: "Cost-effective models that can still get the job done for easier tasks. Supports up to 160k context. Uses OpenAI's o4-mini model for planning, GPT-4.1 for coding, and GPT-4.1 Mini for lighter tasks.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *openaio4miniMedium(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenAI, "openai/o4-mini-medium"),
- },
- Coder: openaigpt41(ModelRoleCoder, nil),
- PlanSummary: *openaigpt41mini(ModelRolePlanSummary, nil),
- Builder: *openaio4miniLow(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniMedium(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniLow(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "openai/o4-mini-medium"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "openai/gpt-4.1")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/gpt-4.1-mini"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-low"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-low")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- OSSModelPack = ModelPack{
+ OSSModelPackSchema = ModelPackSchema{
- Name: "oss",
- Description: "An experimental mix of the best open source models for coding. Supports up to 56k context, 8k per file. Works best with smaller projects and files. Includes reasoning.",
- Planner: getModelRoleConfig(ModelRolePlanner, "deepseek/r1-reasoning-visible"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "deepseek/v3-0324")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "deepseek/r1-reasoning-hidden"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "deepseek/r1-reasoning-hidden"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "deepseek/r1-reasoning-hidden")),
- Namer: getModelRoleConfig(ModelRoleName, "qwen/qwen-2.5-coder-32b-instruct"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "qwen/qwen-2.5-coder-32b-instruct"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "deepseek/r1-reasoning-hidden"),
+ Name: "oss",
+ Description: "An experimental mix of the best open source models for coding. Supports up to 56k context, 8k per file. Works best with smaller projects and files. Includes reasoning.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *deepseekr1Reasoning(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "deepseek/deepseek-r1-reasoning"),
- },
- Coder: deepseekv3(ModelRoleCoder, nil),
- PlanSummary: *deepseekr1NoReasoning(ModelRolePlanSummary, nil),
- Builder: *deepseekr1NoReasoning(ModelRoleBuilder, nil),
- WholeFileBuilder: deepseekr1NoReasoning(ModelRoleWholeFileBuilder, nil),
- Namer: *qwen25coder32b(ModelRoleName, nil),
- CommitMsg: *qwen25coder32b(ModelRoleCommitMsg, nil),
- ExecStatus: *deepseekr1NoReasoning(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "deepseek/r1-reasoning-visible"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "deepseek/v3-0324")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "deepseek/r1-reasoning-hidden"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "deepseek/r1-reasoning-hidden"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "deepseek/r1-reasoning-hidden")),
++ Namer: getModelRoleConfig(ModelRoleName, "qwen/qwen-2.5-coder-32b-instruct"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "qwen/qwen-2.5-coder-32b-instruct"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "deepseek/r1-reasoning-hidden"),
}
- OpusPlannerModelPack = ModelPack{
- Name: "opus-planner",
- Description: "Like daily driver, but uses Claude Opus 4 for planning. Supports up to 160k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *claudeOpus4(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-opus-4"),
- },
- Coder: claudeSonnet4With37Fallback(ModelRoleCoder, nil),
- Architect: claudeOpus4(ModelRoleArchitect, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
- }
-
- OpenAIModelPack = ModelPack{
+ OpenAIPackSchema = ModelPackSchema{
- Name: "openai",
- Description: "OpenAI blend. Supports up to 1M context. Uses OpenAI's GPT-4.1 model for heavy lifting, GPT-4.1 Mini for lighter tasks.",
- Planner: getModelRoleConfig(ModelRolePlanner, "openai/gpt-4.1"),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: defaultBuilder,
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
+ Name: "openai",
+ Description: "OpenAI blend. Supports up to 1M context. Uses OpenAI's GPT-4.1 model for heavy lifting, GPT-4.1 Mini for lighter tasks.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *openaigpt41(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenAI, "openai/gpt-4.1"),
- },
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "openai/gpt-4.1"),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: defaultBuilder,
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-medium")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- AnthropicModelPack = ModelPack{
+ AnthropicPackSchema = ModelPackSchema{
- Name: "anthropic",
- Description: "Anthropic blend. Supports up to 180k context. Uses Claude 3.5 Sonnet for heavy lifting, Claude 3 Haiku for lighter tasks.",
- Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-3.7-sonnet"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "anthropic/claude-3.5-haiku"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "anthropic/claude-3.7-sonnet"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "anthropic/claude-3.7-sonnet")),
- Namer: getModelRoleConfig(ModelRoleName, "anthropic/claude-3.5-haiku"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "anthropic/claude-3.5-haiku"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "anthropic/claude-3.7-sonnet"),
+ Name: "anthropic",
- Description: "Anthropic blend. Supports up to 180k context. Uses Claude Sonnet 4 for heavy lifting, Claude 3.5 Haiku for lighter tasks.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *claudeSonnet4With37Fallback(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-sonnet-4"),
- },
- PlanSummary: *claude35haiku(ModelRolePlanSummary, nil),
- Builder: *claudeSonnet4With37Fallback(ModelRoleBuilder, nil),
- WholeFileBuilder: claudeSonnet4With37Fallback(ModelRoleWholeFileBuilder, nil),
- Namer: *claude35haiku(ModelRoleName, nil),
- CommitMsg: *claude35haiku(ModelRoleCommitMsg, nil),
- ExecStatus: *claudeSonnet4With37Fallback(ModelRoleExecStatus, nil),
++ Description: "Anthropic blend. Supports up to 180k context. Uses Claude Sonnet 4 for heavy lifting, Claude 3 Haiku for lighter tasks.",
++ Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-sonnet-4"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "anthropic/claude-3.5-haiku"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "anthropic/claude-sonnet-4"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "anthropic/claude-sonnet-4")),
++ Namer: getModelRoleConfig(ModelRoleName, "anthropic/claude-3.5-haiku"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "anthropic/claude-3.5-haiku"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "anthropic/claude-sonnet-4"),
}
- GeminiPreviewModelPack = ModelPack{
+ GeminiPreviewPackSchema = ModelPackSchema{
- Name: "gemini-preview",
- Description: "Uses Gemini 2.5 Pro Preview for planning and coding, default models for other roles. Supports up to 1M input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "google/gemini-2.5-pro-preview"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "google/gemini-2.5-pro-preview")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: defaultBuilder,
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
+ Name: "gemini-preview",
+ Description: "Uses Gemini 2.5 Pro Preview for planning and coding, default models for other roles. Supports up to 1M input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *geminipro25preview(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "google/gemini-2.5-pro-preview"),
- },
- Coder: geminipro25preview(ModelRoleCoder, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "google/gemini-2.5-pro-preview"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "google/gemini-2.5-pro-preview")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: defaultBuilder,
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-medium")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- GeminiExperimentalModelPack = ModelPack{
+ GeminiExperimentalPackSchema = ModelPackSchema{
- Name: "gemini-exp",
- Description: "Uses Gemini 2.5 Pro Experimental (free) for planning and coding, default models for other roles. Supports up to 1M input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "google/gemini-2.5-pro-exp"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "google/gemini-2.5-pro-exp")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: defaultBuilder,
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
+ Name: "gemini-exp",
+ Description: "Uses Gemini 2.5 Pro Experimental (free) for planning and coding, default models for other roles. Supports up to 1M input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *geminipro25exp(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "google/gemini-2.5-pro-exp-03-25"),
- },
- Coder: geminipro25exp(ModelRoleCoder, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, &modelConfig{
- strongModel: openaio4miniHigh(ModelRoleBuilder, nil),
- }),
- WholeFileBuilder: openaio4miniMedium(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniLow(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "google/gemini-2.5-pro-exp"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "google/gemini-2.5-pro-exp")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: defaultBuilder,
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-medium")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-low"),
}
- R1PlannerModelPack = ModelPack{
+ R1PlannerPackSchema = ModelPackSchema{
- Name: "r1-planner",
- Description: "Uses DeepSeek R1 for planning, Qwen for light tasks, and default models for implementation. Supports up to 56k input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "deepseek/r1-reasoning-visible"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
+ Name: "r1-planner",
+ Description: "Uses DeepSeek R1 for planning, Qwen for light tasks, and default models for implementation. Supports up to 56k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *deepseekr1Reasoning(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "deepseek/deepseek-r1-reasoning"),
- },
- Coder: claude37Sonnet(ModelRoleCoder, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, nil),
- WholeFileBuilder: openaio4miniLow(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniMedium(ModelRoleExecStatus, nil),
++ Planner: getModelRoleConfig(ModelRolePlanner, "deepseek/r1-reasoning-visible"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-low")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
}
- PerplexityPlannerModelPack = ModelPack{
+ PerplexityPlannerPackSchema = ModelPackSchema{
- Name: "perplexity-planner",
- Description: "Uses Perplexity Sonar for planning, Qwen for light tasks, and default models for implementation. Supports up to 97k input context.",
- Planner: getModelRoleConfig(ModelRolePlanner, "perplexity/sonar-reasoning-visible"),
- Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet")),
- PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
- Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium"),
- WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low")),
- Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
- CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
- ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
+ Name: "perplexity-planner",
+ Description: "Uses Perplexity Sonar for planning, Qwen for light tasks, and default models for implementation. Supports up to 97k input context.",
- Planner: PlannerRoleConfig{
- ModelRoleConfig: *perplexitySonarReasoning(ModelRolePlanner, nil),
- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "perplexity/sonar-reasoning"),
- },
- Coder: claude37Sonnet(ModelRoleCoder, nil),
- PlanSummary: *openaio4miniLow(ModelRolePlanSummary, nil),
- Builder: *openaio4miniMedium(ModelRoleBuilder, nil),
- WholeFileBuilder: openaio4miniLow(ModelRoleWholeFileBuilder, nil),
- Namer: *openaigpt41mini(ModelRoleName, nil),
- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),
- ExecStatus: *openaio4miniMedium(ModelRoleExecStatus, nil),
- }
-
-}
-
-type modelConfig struct {
- largeContextFallback *ModelRoleConfig
- largeOutputFallback *ModelRoleConfig
- errorFallback *ModelRoleConfig
- strongModel *ModelRoleConfig
- missingKeyFallback *ModelRoleConfig
-}
-
-func getModelConfig(role ModelRole, provider ModelProvider, modelId ModelId, fallbacks *modelConfig) *ModelRoleConfig {
- if fallbacks == nil {
- fallbacks = &modelConfig{}
++ Planner: getModelRoleConfig(ModelRolePlanner, "perplexity/sonar-reasoning-visible"),
++ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-sonnet-4")),
++ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low"),
++ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium"),
++ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder,
++ "openai/o4-mini-low")),
++ Namer: getModelRoleConfig(ModelRoleName, "openai/gpt-4.1-mini"),
++ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "openai/gpt-4.1-mini"),
++ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "openai/o4-mini-medium"),
}
- return &ModelRoleConfig{
- Role: role,
- BaseModelConfig: GetAvailableModel(provider, modelId).BaseModelConfig,
- Temperature: DefaultConfigByRole[role].Temperature,
- TopP: DefaultConfigByRole[role].TopP,
-
- LargeContextFallback: fallbacks.largeContextFallback,
- LargeOutputFallback: fallbacks.largeOutputFallback,
- ErrorFallback: fallbacks.errorFallback,
- StrongModel: fallbacks.strongModel,
- MissingKeyFallback: fallbacks.missingKeyFallback,
- }
-}
-
-func claude37Sonnet(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet", fallbacks)
-}
-
-func claudeSonnet4(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-sonnet-4", fallbacks)
-}
-
-func claudeSonnet4Thinking(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-sonnet-4:thinking", fallbacks)
-}
-
-func claudeSonnet4ThinkingHidden(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-sonnet-4:thinking-hidden", fallbacks)
-}
-
-func claudeOpus4(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-opus-4", fallbacks)
-}
-
-func claudeSonnet4With37Fallback(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- claude37Fallback := getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
+ DailyDriverModelPack = DailyDriverPackSchema.ToModelPack()
+ ReasoningModelPack = ReasoningPackSchema.ToModelPack()
+ StrongModelPack = StrongPackSchema.ToModelPack()
+ CheapModelPack = CheapModelPackSchema.ToModelPack()
+ OSSModelPack = OSSModelPackSchema.ToModelPack()
+ AnthropicModelPack = AnthropicPackSchema.ToModelPack()
+ OpenAIModelPack = OpenAIPackSchema.ToModelPack()
+ GeminiPreviewModelPack = GeminiPreviewPackSchema.ToModelPack()
+ GeminiExperimentalModelPack = GeminiExperimentalPackSchema.ToModelPack()
+ R1PlannerModelPack = R1PlannerPackSchema.ToModelPack()
+ PerplexityPlannerModelPack = PerplexityPlannerPackSchema.ToModelPack()
+
- // Preserve the old slices/vars
+ BuiltInModelPacks = []*ModelPack{
+ &DailyDriverModelPack,
+ &ReasoningModelPack,
+ &StrongModelPack,
+ &CheapModelPack,
+ &OSSModelPack,
+ &AnthropicModelPack,
+ &OpenAIModelPack,
+ &GeminiPreviewModelPack,
+ &GeminiExperimentalModelPack,
+ &R1PlannerModelPack,
+ &PerplexityPlannerModelPack,
}
- fallbacks.errorFallback = claude37Fallback
-
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-sonnet-4", fallbacks)
-}
-
-func claude37SonnetThinking(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet:thinking", fallbacks)
-}
-
-func claude37SonnetThinkingHidden(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet:thinking-hidden", fallbacks)
-}
-
-func claude35Sonnet(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.5-sonnet", fallbacks)
-}
-
-func claude35haiku(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.5-haiku", fallbacks)
-}
-
-func gemini15pro(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-pro-1.5", fallbacks)
-}
-
-func gemini25propreview(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-2.5-pro-preview", fallbacks)
-}
-
-func openaigpt41(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/gpt-4.1", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/gpt-4.1", fallbacks)
-}
-
-func openaigpt41mini(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/gpt-4.1-mini", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/gpt-4.1-mini", fallbacks)
-}
-
-func openaio3highWitho4miniFallback(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-high", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- fallbacks.errorFallback = openaio4miniHigh(role, nil)
- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-high", fallbacks)
-}
-
-func openaio3high(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-high", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-high", fallbacks)
-}
-
-func openaio3medium(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-medium", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-medium", fallbacks)
-}
-
-func openaio3low(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-low", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-low", fallbacks)
-}
-
-func openaio4miniHigh(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-high", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-high", fallbacks)
-}
-
-func openaio4miniMedium(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-medium", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- res := getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-medium", fallbacks)
- return res
-}
-
-func openaio4miniLow(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-low", nil)
- if fallbacks == nil {
- fallbacks = &modelConfig{}
- }
- fallbacks.missingKeyFallback = openrouterFallback
- return getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-low", fallbacks)
-}
-
-func qwen25coder32b(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "qwen/qwen-2.5-coder-32b-instruct", fallbacks)
-}
-
-func deepseekr1Reasoning(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "deepseek/deepseek-r1-reasoning", fallbacks)
-}
-
-func deepseekr1NoReasoning(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "deepseek/deepseek-r1-no-reasoning", fallbacks)
-}
-
-func deepseekv3(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "deepseek/deepseek-chat-v3-0324", fallbacks)
-}
-
-func geminiflash25preview(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-2.5-flash-preview", fallbacks)
-}
-
-func perplexityR11776(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "perplexity/r1-1776", fallbacks)
-}
-
-func perplexitySonarReasoning(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "perplexity/sonar-reasoning", fallbacks)
-}
-
-// func r1distillqwen32b(role ModelRole, fallbacks *modelConfigFallbacks) *ModelRoleConfig {
-// return getModelConfig(role, ModelProviderOpenRouter, "deepseek/deepseek-r1-distill-qwen-32b", fallbacks)
-// }
-
-func geminipro25exp(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-2.5-pro-exp-03-25", fallbacks)
-}
-func geminipro25preview(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {
- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-2.5-pro-preview", fallbacks)
+ DefaultModelPack = &DailyDriverModelPack
-
}
diff --cc app/shared/ai_models_providers.go
index 91a7b8df,8fad6cb2..79f4de58
--- a/app/shared/ai_models_providers.go
+++ b/app/shared/ai_models_providers.go
@@@ -33,242 -10,22 +33,243 @@@ type ModelProvider strin
const (
ModelProviderOpenRouter ModelProvider = "openrouter"
ModelProviderOpenAI ModelProvider = "openai"
- ModelProviderCustom ModelProvider = "custom"
+
+ ModelProviderAnthropic ModelProvider = "anthropic"
+ ModelProviderGoogleAIStudio ModelProvider = "google-ai-studio"
+ ModelProviderGoogleVertex ModelProvider = "google-vertex"
+ ModelProviderAzureOpenAI ModelProvider = "azure-openai"
+ ModelProviderDeepSeek ModelProvider = "deepseek"
+ ModelProviderPerplexity ModelProvider = "perplexity"
+
+ ModelProviderAmazonBedrock ModelProvider = "aws-bedrock"
+
+ ModelProviderOllama ModelProvider = "ollama"
+
+ ModelProviderCustom ModelProvider = "custom"
)
-var AllModelProviders = []string{
- string(ModelProviderOpenAI),
- string(ModelProviderOpenRouter),
- // string(ModelProviderTogether),
- string(ModelProviderCustom),
+var ModelProviderToLiteLLMId = map[ModelProvider]string{
+ ModelProviderGoogleAIStudio: "gemini",
+ ModelProviderGoogleVertex: "vertex_ai",
+ ModelProviderAzureOpenAI: "azure",
+ ModelProviderAmazonBedrock: "bedrock",
+}
+
+var AllModelProviders = []ModelProvider{
+ ModelProviderOpenAI,
+ ModelProviderOpenRouter,
+ ModelProviderAnthropic,
+ ModelProviderGoogleAIStudio,
+ ModelProviderGoogleVertex,
+ ModelProviderAzureOpenAI,
+ ModelProviderDeepSeek,
+ ModelProviderPerplexity,
+ ModelProviderAmazonBedrock,
+ ModelProviderOllama,
+ ModelProviderCustom,
}
-var BaseUrlByProvider = map[ModelProvider]string{
- ModelProviderOpenAI: OpenAIV1BaseUrl,
- ModelProviderOpenRouter: OpenRouterBaseUrl,
+type ModelProviderExtraAuthVars struct {
+ Var string `json:"var"`
+ MaybeJSONFilePath bool `json:"maybeJSONFilePath,omitempty"`
+ Required bool `json:"required,omitempty"`
+ Default string `json:"default,omitempty"`
+}
+
+type ModelProviderConfigSchema struct {
+ Provider ModelProvider `json:"provider"`
+ CustomProvider *string `json:"customProvider,omitempty"`
+ BaseUrl string `json:"baseUrl"`
+
+ // for AWS Bedrock models
+ HasAWSAuth bool `json:"hasAWSAuth,omitempty"`
+
+ // for local models that don't require auth (ollama, etc.)
+ SkipAuth bool `json:"skipAuth,omitempty"`
+ LocalOnly bool `json:"localOnly,omitempty"`
+
+ ApiKeyEnvVar string `json:"apiKeyEnvVar,omitempty"`
+ ExtraAuthVars []ModelProviderExtraAuthVars `json:"extraAuthVars,omitempty"`
+}
+
+func (m *ModelProviderConfigSchema) ToComposite() string {
+ if m.CustomProvider != nil {
+ return fmt.Sprintf("%s|%s", m.Provider, *m.CustomProvider)
+ }
+ return string(m.Provider)
+}
+
+const DefaultAzureApiVersion = "2024-10-21"
+const AnthropicMaxReasoningBudget = 32000
++const GoogleMaxReasoningBudget = 32000
+
+var BuiltInModelProviderConfigs = map[ModelProvider]ModelProviderConfigSchema{
+ ModelProviderOpenAI: {
+ Provider: ModelProviderOpenAI,
+ BaseUrl: OpenAIV1BaseUrl,
+ ApiKeyEnvVar: OpenAIEnvVar,
+ ExtraAuthVars: []ModelProviderExtraAuthVars{
+ {
+ Var: "OPENAI_ORG_ID",
+ Required: false,
+ },
+ },
+ },
+ ModelProviderOpenRouter: {
+ Provider: ModelProviderOpenRouter,
+ BaseUrl: OpenRouterBaseUrl,
+ ApiKeyEnvVar: OpenRouterApiKeyEnvVar,
+ },
+ ModelProviderAnthropic: {
+ Provider: ModelProviderAnthropic,
+ BaseUrl: LiteLLMBaseUrl,
+ ApiKeyEnvVar: AnthropicApiKeyEnvVar,
+ },
+ ModelProviderGoogleAIStudio: {
+ Provider: ModelProviderGoogleAIStudio,
+ BaseUrl: LiteLLMBaseUrl,
+ ApiKeyEnvVar: GoogleAIStudioApiKeyEnvVar,
+ },
+ ModelProviderGoogleVertex: {
+ Provider: ModelProviderGoogleVertex,
+ BaseUrl: LiteLLMBaseUrl,
+ ExtraAuthVars: []ModelProviderExtraAuthVars{
+ {
+ // this is a file path, but client-side it will be read and then passed along as an auth var just as if it were an env var
+ Var: "GOOGLE_APPLICATION_CREDENTIALS",
+ MaybeJSONFilePath: true,
+ Required: true,
+ },
+ {
+ Var: "VERTEXAI_PROJECT",
+ Required: true,
+ },
+ {
+ Var: "VERTEXAI_LOCATION",
+ Required: true,
+ },
+ },
+ },
+ ModelProviderAzureOpenAI: {
+ Provider: ModelProviderAzureOpenAI,
+ BaseUrl: LiteLLMBaseUrl,
+ ApiKeyEnvVar: AzureOpenAIEnvVar,
+ ExtraAuthVars: []ModelProviderExtraAuthVars{
+ {
+ Var: "AZURE_API_BASE",
+ Required: true,
+ },
+ {
+ Var: "AZURE_API_VERSION",
+ Required: false,
+ Default: DefaultAzureApiVersion,
+ },
+ },
+ },
+ ModelProviderDeepSeek: {
+ Provider: ModelProviderDeepSeek,
+ BaseUrl: LiteLLMBaseUrl,
+ ApiKeyEnvVar: DeepSeekApiKeyEnvVar,
+ },
+ ModelProviderPerplexity: {
+ Provider: ModelProviderPerplexity,
+ BaseUrl: LiteLLMBaseUrl,
+ ApiKeyEnvVar: PerplexityApiKeyEnvVar,
+ },
+ ModelProviderAmazonBedrock: {
+ Provider: ModelProviderAmazonBedrock,
+ BaseUrl: LiteLLMBaseUrl,
+ HasAWSAuth: true,
+
+ // these aren't required as env vars—but if found in the credentials file, they are passed along as auth vars just as if they were env vars
+ ExtraAuthVars: []ModelProviderExtraAuthVars{
+ {Var: "AWS_ACCESS_KEY_ID", Required: true},
+ {Var: "AWS_SECRET_ACCESS_KEY", Required: true},
+ {Var: "AWS_REGION", Required: true},
+ {Var: "AWS_SESSION_TOKEN", Required: false},
+ },
+ },
+ ModelProviderOllama: {
+ Provider: ModelProviderOllama,
+ BaseUrl: LiteLLMBaseUrl,
+ SkipAuth: true,
+ LocalOnly: true,
+ },
+}
+
+var BuiltInModelProviderConfigsByComposite = map[string]ModelProviderConfigSchema{}
+
+func init() {
+ for _, providerConfig := range BuiltInModelProviderConfigs {
+ BuiltInModelProviderConfigsByComposite[providerConfig.ToComposite()] = providerConfig
+ }
+}
+
+func GetProvidersForAuthVars(authVars map[string]string) []ModelProviderConfigSchema {
+ var foundProviders []ModelProviderConfigSchema
+ for _, providerConfig := range BuiltInModelProviderConfigs {
+ if providerConfig.SkipAuth && len(authVars) == 0 {
+ foundProviders = append(foundProviders, providerConfig)
+ break
+ }
+
+ var checkVars []string
+ if providerConfig.ApiKeyEnvVar != "" {
+ checkVars = append(checkVars, providerConfig.ApiKeyEnvVar)
+ }
+ for _, extraAuthVar := range providerConfig.ExtraAuthVars {
+ if extraAuthVar.Required {
+ checkVars = append(checkVars, extraAuthVar.Var)
+ }
+ }
+
+ missingAny := false
+ for _, checkVar := range checkVars {
+ if _, ok := authVars[checkVar]; !ok {
+ missingAny = true
+ break
+ }
+ }
+ if missingAny {
+ continue
+ }
+
+ foundProviders = append(foundProviders, providerConfig)
+ }
+
+ return foundProviders
+}
+
+func (m ModelRoleConfig) GetProvidersForAuthVars(authVars map[string]string) []ModelProviderConfigSchema {
+ usesProviders, ok := BuiltInModelProvidersByModelId[m.ModelId]
+ if !ok {
+ return []ModelProviderConfigSchema{}
+ }
+
+ providers := GetProvidersForAuthVars(authVars)
+ providersByComposite := map[string]ModelProviderConfigSchema{}
+ for _, provider := range providers {
+ providersByComposite[provider.ToComposite()] = provider
+ }
+
+ res := []ModelProviderConfigSchema{}
+ for _, usesProvider := range usesProviders {
+ composite := usesProvider.ToComposite()
+ provider, ok := providersByComposite[composite]
+ if !ok {
+ continue
+ }
+
+ res = append(res, provider)
+ }
+
+ return res
}
-var ApiKeyByProvider = map[ModelProvider]string{
- ModelProviderOpenAI: OpenAIEnvVar,
- ModelProviderOpenRouter: OpenRouterApiKeyEnvVar,
+func (m ModelRoleConfig) GetFirstProviderForAuthVars(authVars map[string]string) *ModelProviderConfigSchema {
+ providers := m.GetProvidersForAuthVars(authVars)
+ if len(providers) == 0 {
+ return nil
+ }
+ return &providers[0]
}