Open source AI coding agent. Designed for large projects and real world tasks.
Go aiai-agentsai-developer-toolsai-toolsclicommand-linedeveloper-toolsgitgolanggpt-4llmopenaipolyglot-programmingterminalterminal-basedterminal-ui
wip: model refactor
Dane Schneider750df527d39d
diff --git a/app/server/db/data_models.go b/app/server/db/data_models.goindex 048e73ae..72d95dae 100644--- a/app/server/db/data_models.go+++ b/app/server/db/data_models.go@@ -352,17 +352,26 @@ func (model *AvailableModel) ToApi() *shared.AvailableModel {return &shared.AvailableModel{Id: model.Id,BaseModelConfig: shared.BaseModelConfig{- Provider: model.Provider,- CustomProvider: model.CustomProvider,- BaseUrl: model.BaseUrl,- ModelId: model.ModelId,- ModelName: model.ModelName,- MaxTokens: model.MaxTokens,- ApiKeyEnvVar: model.ApiKeyEnvVar,- MaxOutputTokens: model.MaxOutputTokens,- ReservedOutputTokens: model.ReservedOutputTokens,- ModelCompatibility: shared.ModelCompatibility{- HasImageSupport: model.HasImageSupport,++ ModelId: model.ModelId,++ BaseModelShared: shared.BaseModelShared{+ MaxTokens: model.MaxTokens,+ MaxOutputTokens: model.MaxOutputTokens,+ ReservedOutputTokens: model.ReservedOutputTokens,+ ModelCompatibility: shared.ModelCompatibility{+ HasImageSupport: model.HasImageSupport,+ },+ },++ BaseModelProviderConfig: shared.BaseModelProviderConfig{+ ModelName: model.ModelName,+ ModelProviderConfigSchema: shared.ModelProviderConfigSchema{+ Provider: model.Provider,+ CustomProvider: model.CustomProvider,+ BaseUrl: model.BaseUrl,+ ApiKeyEnvVar: model.ApiKeyEnvVar,+ },},},Description: model.Description,diff --git a/app/server/model/client.go b/app/server/model/client.goindex 68520a5d..95b02f09 100644--- a/app/server/model/client.go+++ b/app/server/model/client.go@@ -126,7 +126,7 @@ func CreateChatCompletionStream(req.IncludeReasoning = true}- return withStreamingRetries(ctx, func(numTotalRetry int, modelErr *shared.ModelError, stripCacheControl bool) (*ExtendedChatCompletionStream, shared.FallbackResult, error) {+ return withStreamingRetries(ctx, func(numTotalRetry int, modelErr *shared.ModelError) (*ExtendedChatCompletionStream, shared.FallbackResult, error) {fallbackRes := modelConfig.GetFallbackForModelError(numTotalRetry, modelErr)resolvedModelConfig := fallbackRes.ModelRoleConfig@@ -149,7 +149,7 @@ func CreateChatCompletionStream(}}- if stripCacheControl {+ if modelErr.Kind == shared.ErrCacheSupport {for i := range req.Messages {for j := range req.Messages[i].Content {if req.Messages[i].Content[j].CacheControl != nil {diff --git a/app/server/model/client_stream.go b/app/server/model/client_stream.goindex b54f7155..503f6f78 100644--- a/app/server/model/client_stream.go+++ b/app/server/model/client_stream.go@@ -41,7 +41,7 @@ func CreateChatCompletionWithInternalStream(// Force streaming mode since we're using the streaming APIreq.Stream = true- return withStreamingRetries(ctx, func(numTotalRetry int, modelErr *shared.ModelError, stripCacheControl bool) (resp *types.ModelResponse, fallbackRes shared.FallbackResult, err error) {+ return withStreamingRetries(ctx, func(numTotalRetry int, modelErr *shared.ModelError) (resp *types.ModelResponse, fallbackRes shared.FallbackResult, err error) {fallbackRes = modelConfig.GetFallbackForModelError(numTotalRetry, modelErr)resolvedModelConfig := fallbackRes.ModelRoleConfig@@ -65,17 +65,6 @@ func CreateChatCompletionWithInternalStream(}modelConfig = resolvedModelConfig-- if stripCacheControl {- for i := range req.Messages {- for j := range req.Messages[i].Content {- if req.Messages[i].Content[j].CacheControl != nil {- req.Messages[i].Content[j].CacheControl = nil- }- }- }- }-resp, err = processChatCompletionStream(resolvedModelConfig, opClient, resolvedModelConfig.BaseModelConfig.BaseUrl, ctx, req, onStream, reqStarted)if err != nil {return nil, fallbackRes, err@@ -228,7 +217,7 @@ func processChatCompletionStream(func withStreamingRetries[T any](ctx context.Context,- operation func(numRetry int, modelErr *shared.ModelError, stripCacheControl bool) (resp *T, fallbackRes shared.FallbackResult, err error),+ operation func(numRetry int, modelErr *shared.ModelError) (resp *T, fallbackRes shared.FallbackResult, err error),onContextDone func(resp *T, err error),) (*T, error) {var resp *T@@ -236,7 +225,6 @@ func withStreamingRetries[T any](var numFallbackRetry intvar fallbackRes shared.FallbackResultvar modelErr *shared.ModelError- var hadCacheSupportErr boolfor {if ctx.Err() != nil {@@ -259,7 +247,7 @@ func withStreamingRetries[T any](log.Printf("withStreamingRetries - will run operation")- resp, fallbackRes, err = operation(numTotalRetry, modelErr, hadCacheSupportErr)+ resp, fallbackRes, err = operation(numTotalRetry, modelErr)if err == nil {return resp, nil}@@ -280,14 +268,7 @@ func withStreamingRetries[T any](log.Printf("Error in streaming operation: %v, isFallback: %t, numTotalRetry: %d, numFallbackRetry: %d, numRetry: %d, compareRetries: %d, maxRetries: %d\n", err, isFallback, numTotalRetry, numFallbackRetry, numRetry, compareRetries, maxRetries)classifyRes := classifyBasicError(err)-- isCacheSupportErr := classifyRes.Kind == shared.ErrCacheSupport-- if isCacheSupportErr {- hadCacheSupportErr = true- } else {- modelErr = &classifyRes- }+ modelErr = &classifyResnewFallback := falseif !modelErr.Retriable {@@ -324,16 +305,9 @@ func withStreamingRetries[T any](log.Printf("withStreamingRetries - retrying stream in %v seconds", retryDelay)time.Sleep(retryDelay)- if !isCacheSupportErr {- numTotalRetry++- if isFallback && !newFallback {- numFallbackRetry++- }- }-- // if we had a cache support error previously but now we're doing a new fallback, reset the hadCacheSupportErr flag since the fallback will use a new model- if hadCacheSupportErr && newFallback {- hadCacheSupportErr = false+ numTotalRetry+++ if isFallback && !newFallback {+ numFallbackRetry++}}}diff --git a/app/shared/ai_models_available.go b/app/shared/ai_models_available.goindex 1df879ff..cd1ea1dd 100644--- a/app/shared/ai_models_available.go+++ b/app/shared/ai_models_available.go@@ -1,6 +1,8 @@package sharedimport (+ "strings"+"github.com/davecgh/go-spew/spew")@@ -59,84 +61,36 @@ var BuiltInModels = []*BaseModelConfigSchema{SystemPromptDisabled: true,RoleParamsDisabled: true,ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortHigh,StopDisabled: true,},+ RequiresVariantOverrides: []string{+ "ReasoningEffort",+ },+Variants: []BaseModelConfigVariant{{VariantTag: "high",Description: "high",- },- },-- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderOpenAI,- ModelName: "o3",- },- {- Provider: ModelProviderAzureOpenAI,- ModelName: "o3",- },- {- Provider: ModelProviderOpenRouter,- ModelName: "openai/o3",- },- },- },- {- PriceId: "openai/o3",- ModelTag: "openai/o3-medium",- Description: "OpenAI o3-medium",- DefaultMaxConvoTokens: 15000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatXml,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortMedium,- StopDisabled: true,- },-- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderOpenAI,- ModelName: "o3",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortHigh,+ },},{- Provider: ModelProviderAzureOpenAI,- ModelName: "o3",+ VariantTag: "medium",+ Description: "medium",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortMedium,+ },},{- Provider: ModelProviderOpenRouter,- ModelName: "openai/o3",+ VariantTag: "low",+ Description: "low",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortLow,+ },},},- },- {- PriceId: "openai/o3",- ModelTag: "openai/o3-low",- Description: "OpenAI o3-low",- DefaultMaxConvoTokens: 15000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatXml,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortLow,- StopDisabled: true,- },Providers: []BaseModelUsesProvider{{@@ -153,11 +107,9 @@ var BuiltInModels = []*BaseModelConfigSchema{},},},-{- PriceId: "openai/o4-mini",- ModelTag: "openai/o4-mini-high",- Description: "OpenAI o4-mini-high",+ ModelTag: "openai/o4-mini",+ Description: "OpenAI o4-mini",DefaultMaxConvoTokens: 10000,BaseModelShared: BaseModelShared{@@ -173,75 +125,35 @@ var BuiltInModels = []*BaseModelConfigSchema{StopDisabled: true,},- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderOpenAI,- ModelName: "o4-mini",- },- {- Provider: ModelProviderAzureOpenAI,- ModelName: "o4-mini",- },- {- Provider: ModelProviderOpenRouter,- ModelName: "openai/o4-mini",- },- },- },-- {- PriceId: "openai/o4-mini",- ModelTag: "openai/o4-mini-medium",- Description: "OpenAI o4-mini-medium",- DefaultMaxConvoTokens: 10000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortMedium,- StopDisabled: true,+ RequiresVariantOverrides: []string{+ "ReasoningEffort",},- Providers: []BaseModelUsesProvider{+ Variants: []BaseModelConfigVariant{{- Provider: ModelProviderOpenAI,- ModelName: "o4-mini",+ VariantTag: "high",+ Description: "high",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortHigh,+ },},{- Provider: ModelProviderAzureOpenAI,- ModelName: "o4-mini",+ VariantTag: "medium",+ Description: "medium",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortMedium,+ ReservedOutputTokens: 30000, // 15k for reasoning, 15k for output+ },},{- Provider: ModelProviderOpenRouter,- ModelName: "openai/o4-mini",+ VariantTag: "low",+ Description: "low",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortLow,+ ReservedOutputTokens: 20000, // 5-10k for reasoning, 5-10k for output+ },},},- },-- {- PriceId: "openai/o4-mini",- ModelTag: "openai/o4-mini-low",- Description: "OpenAI o4-mini-low",- DefaultMaxConvoTokens: 10000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortLow,- StopDisabled: true,- },Providers: []BaseModelUsesProvider{{@@ -260,7 +172,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "openai/gpt-4.1",ModelTag: "openai/gpt-4.1",Description: "OpenAI gpt-4.1",DefaultMaxConvoTokens: 75000,@@ -290,7 +201,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "openai/gpt-4.1-mini",ModelTag: "openai/gpt-4.1-mini",Description: "OpenAI gpt-4.1-mini",DefaultMaxConvoTokens: 75000,@@ -320,7 +230,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "openai/gpt-4.1-nano",ModelTag: "openai/gpt-4.1-nano",Description: "OpenAI gpt-4.1-nano",DefaultMaxConvoTokens: 75000,@@ -350,9 +259,8 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "openai/o3-mini",- ModelTag: "openai/o3-mini-high",- Description: "OpenAI o3-mini-high",+ ModelTag: "openai/o3-mini",+ Description: "OpenAI o3-mini",DefaultMaxConvoTokens: 10000,BaseModelShared: BaseModelShared{@@ -368,75 +276,35 @@ var BuiltInModels = []*BaseModelConfigSchema{StopDisabled: true,},- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderOpenAI,- ModelName: "o3-mini",- },- {- Provider: ModelProviderAzureOpenAI,- ModelName: "o3-mini",- },- {- Provider: ModelProviderOpenRouter,- ModelName: "openai/o3-mini",- },- },- },-- {- PriceId: "openai/o3-mini",- ModelTag: "openai/o3-mini-medium",- Description: "OpenAI o3-mini-medium",- DefaultMaxConvoTokens: 10000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortMedium,- StopDisabled: true,+ RequiresVariantOverrides: []string{+ "ReasoningEffort",},- Providers: []BaseModelUsesProvider{+ Variants: []BaseModelConfigVariant{{- Provider: ModelProviderOpenAI,- ModelName: "o3-mini",+ VariantTag: "high",+ Description: "high",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortHigh,+ },},{- Provider: ModelProviderAzureOpenAI,- ModelName: "o3-mini",+ VariantTag: "medium",+ Description: "medium",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortMedium,+ ReservedOutputTokens: 30000, // 15k for reasoning, 15k for output+ },},{- Provider: ModelProviderOpenRouter,- ModelName: "openai/o3-mini",+ VariantTag: "low",+ Description: "low",+ Overrides: BaseModelShared{+ ReasoningEffort: ReasoningEffortLow,+ ReservedOutputTokens: 20000, // 5-10k for reasoning, 5-10k for output+ },},},- },-- {- PriceId: "openai/o3-mini",- ModelTag: "openai/o3-mini-low",- Description: "OpenAI o3-mini-low",- DefaultMaxConvoTokens: 10000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 100000,- ReservedOutputTokens: 40000, // 25k for reasoning, 15k for output- ModelCompatibility: fullCompatibility,- PreferredModelOutputFormat: ModelOutputFormatToolCallJson,- SystemPromptDisabled: true,- RoleParamsDisabled: true,- ReasoningEffortEnabled: true,- ReasoningEffort: ReasoningEffortLow,- StopDisabled: true,- },Providers: []BaseModelUsesProvider{{@@ -455,7 +323,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "anthropic/claude-3.7-sonnet",ModelTag: "anthropic/claude-3.7-sonnet",Description: "Anthropic Claude 3.7 Sonnet",DefaultMaxConvoTokens: 15000,@@ -470,80 +337,34 @@ var BuiltInModels = []*BaseModelConfigSchema{TokenEstimatePaddingPct: 0.10,},- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderAnthropic,- ModelName: "claude-3.7-sonnet",- },- {- Provider: ModelProviderGoogleVertex,- ModelName: "claude-3-7-sonnet@20250219",- },+ Variants: []BaseModelConfigVariant{{- Provider: ModelProviderAmazonBedrock,- ModelName: "anthropic.claude-3-7-sonnet-20250219-v1:0",+ IsBaseVariant: true,},{- Provider: ModelProviderOpenRouter,- ModelName: "anthropic/claude-3.7-sonnet",- },- },- },-- {- PriceId: "anthropic/claude-3.7-sonnet",- ModelTag: "anthropic/claude-3.7-sonnet:thinking-hidden",- Description: "Anthropic Claude 3.7 Sonnet (thinking—reasoning hidden) ",- DefaultMaxConvoTokens: 15000,+ VariantTag: "thinking",+ Description: "thinking",+ Overrides: BaseModelShared{+ ReasoningBudget: AnthropicMaxReasoningBudget,+ },- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 128000,- ReservedOutputTokens: 20000,- SupportsCacheControl: true,- PreferredModelOutputFormat: ModelOutputFormatXml,- SingleMessageNoSystemPrompt: true,- TokenEstimatePaddingPct: 0.10,- IncludeReasoning: false,- ReasoningBudget: AnthropicMaxReasoningBudget,- },-- Providers: []BaseModelUsesProvider{- {- Provider: ModelProviderAnthropic,- ModelName: "claude-3.7-sonnet",+ Variants: []BaseModelConfigVariant{+ {+ VariantTag: "visible",+ Description: "visible",+ Overrides: BaseModelShared{+ IncludeReasoning: true,+ },+ },+ {+ VariantTag: "hidden",+ Description: "hidden",+ Overrides: BaseModelShared{+ IncludeReasoning: false,+ },+ },+ },},- {- Provider: ModelProviderGoogleVertex,- ModelName: "claude-3-7-sonnet@20250219",- },- {- Provider: ModelProviderAmazonBedrock,- ModelName: "anthropic.claude-3-7-sonnet-20250219-v1:0",- },- {- Provider: ModelProviderOpenRouter,- ModelName: "anthropic/claude-3.7-sonnet",- },- },- },-- {- PriceId: "anthropic/claude-3.7-sonnet",- ModelTag: "anthropic/claude-3.7-sonnet:thinking-visible",- Description: "Anthropic Claude 3.7 Sonnet (thinking—reasoning visible) ",- DefaultMaxConvoTokens: 15000,-- BaseModelShared: BaseModelShared{- MaxTokens: 200000,- MaxOutputTokens: 128000,- ReservedOutputTokens: 20000,- SupportsCacheControl: true,- PreferredModelOutputFormat: ModelOutputFormatXml,- SingleMessageNoSystemPrompt: true,- TokenEstimatePaddingPct: 0.10,- IncludeReasoning: true,- ReasoningBudget: AnthropicMaxReasoningBudget,},Providers: []BaseModelUsesProvider{@@ -567,7 +388,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "anthropic/claude-3.5-sonnet",ModelTag: "anthropic/claude-3.5-sonnet",Description: "Anthropic Claude 3.5 Sonnet",DefaultMaxConvoTokens: 15000,@@ -603,7 +423,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "anthropic/claude-3.5-haiku",ModelTag: "anthropic/claude-3.5-haiku",Description: "Anthropic Claude 3.5 Haiku",DefaultMaxConvoTokens: 15000,@@ -639,7 +458,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "google/gemini-pro-1.5",ModelTag: "google/gemini-pro-1.5",Description: "Google Gemini 1.5 Pro",DefaultMaxConvoTokens: 75000,@@ -659,7 +477,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "google/gemini-2.5-pro-preview",ModelTag: "google/gemini-2.5-pro-preview",Description: "Google Gemini 2.5 Pro (Preview)",DefaultMaxConvoTokens: 75000,@@ -679,7 +496,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "google/gemini-2.5-pro-exp",ModelTag: "google/gemini-2.5-pro-exp",Description: "Google Gemini 2.5 Pro (Experimental)",DefaultMaxConvoTokens: 75000,@@ -698,7 +514,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "google/gemini-2.5-flash-preview",ModelTag: "google/gemini-2.5-flash-preview",Description: "Google Gemini 2.5 Flash (Preview)",DefaultMaxConvoTokens: 75000,@@ -719,7 +534,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "deepseek/v3-0324",ModelTag: "deepseek/v3-0324",Description: "DeepSeek V3 (0324)",DefaultMaxConvoTokens: 7500,@@ -737,9 +551,8 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "deepseek/r1",ModelTag: "deepseek/r1",- Description: "DeepSeek R1 (includes reasoning)",+ Description: "DeepSeek R1",DefaultMaxConvoTokens: 7500,BaseModelShared: BaseModelShared{@@ -747,26 +560,23 @@ var BuiltInModels = []*BaseModelConfigSchema{MaxOutputTokens: 8192,ReservedOutputTokens: 8192,PreferredModelOutputFormat: ModelOutputFormatXml,- IncludeReasoning: true,},- Providers: []BaseModelUsesProvider{- {Provider: ModelProviderDeepSeek, ModelName: "deepseek-reasoner"},- {Provider: ModelProviderOpenRouter, ModelName: "deepseek/deepseek-r1"},- },- },- {- PriceId: "deepseek/r1",- ModelTag: "deepseek/r1:reasoning-hidden",- Description: "DeepSeek R1 (reasoning hidden)",- DefaultMaxConvoTokens: 7500,-- BaseModelShared: BaseModelShared{- MaxTokens: 64000,- MaxOutputTokens: 8192,- ReservedOutputTokens: 8192,- PreferredModelOutputFormat: ModelOutputFormatXml,- IncludeReasoning: false,+ Variants: []BaseModelConfigVariant{+ {+ VariantTag: "reasoning-visible",+ Description: "(reasoning visible)",+ Overrides: BaseModelShared{+ IncludeReasoning: true,+ },+ },+ {+ VariantTag: "reasoning-hidden",+ Description: "(reasoning hidden)",+ Overrides: BaseModelShared{+ IncludeReasoning: false,+ },+ },},Providers: []BaseModelUsesProvider{@@ -776,9 +586,8 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "perplexity/r1-1776",ModelTag: "perplexity/r1-1776",- Description: "Perplexity R1-1776 (includes reasoning)",+ Description: "Perplexity R1-1776",DefaultMaxConvoTokens: 7500,BaseModelShared: BaseModelShared{@@ -786,7 +595,23 @@ var BuiltInModels = []*BaseModelConfigSchema{MaxOutputTokens: 128000,ReservedOutputTokens: 30000,PreferredModelOutputFormat: ModelOutputFormatXml,- IncludeReasoning: true,+ },++ Variants: []BaseModelConfigVariant{+ {+ VariantTag: "reasoning-visible",+ Description: "(reasoning visible)",+ Overrides: BaseModelShared{+ IncludeReasoning: true,+ },+ },+ {+ VariantTag: "reasoning-hidden",+ Description: "(reasoning hidden)",+ Overrides: BaseModelShared{+ IncludeReasoning: false,+ },+ },},Providers: []BaseModelUsesProvider{@@ -795,9 +620,8 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "perplexity/sonar-reasoning",ModelTag: "perplexity/sonar-reasoning",- Description: "Perplexity Sonar Reasoning (includes reasoning)",+ Description: "Perplexity Sonar Reasoning",DefaultMaxConvoTokens: 7500,BaseModelShared: BaseModelShared{@@ -805,7 +629,23 @@ var BuiltInModels = []*BaseModelConfigSchema{MaxOutputTokens: 127000,ReservedOutputTokens: 30000,PreferredModelOutputFormat: ModelOutputFormatXml,- IncludeReasoning: true,+ },++ Variants: []BaseModelConfigVariant{+ {+ VariantTag: "visible",+ Description: "(reasoning visible)",+ Overrides: BaseModelShared{+ IncludeReasoning: true,+ },+ },+ {+ VariantTag: "hidden",+ Description: "(reasoning hidden)",+ Overrides: BaseModelShared{+ IncludeReasoning: false,+ },+ },},Providers: []BaseModelUsesProvider{@@ -815,7 +655,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "qwen/qwen-2.5-coder-32b-instruct",ModelTag: "qwen/qwen-2.5-coder-32b-instruct",Description: "Qwen 2.5 Coder 32B (Instruct)",DefaultMaxConvoTokens: 10000,@@ -833,7 +672,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},{- PriceId: "qwen/qwen3-32b",ModelTag: "qwen/qwen3-32b",Description: "Qwen 3-32B (Experimental)",DefaultMaxConvoTokens: 15000, // feel free to raise once you test@@ -854,7 +692,6 @@ var BuiltInModels = []*BaseModelConfigSchema{},},{- PriceId: "qwen/qwen3-8b",ModelTag: "qwen/qwen3-8b",Description: "Qwen 3-8B (Experimental)",DefaultMaxConvoTokens: 7500,@@ -872,809 +709,43 @@ var BuiltInModels = []*BaseModelConfigSchema{},}-var AvailableModels = []*AvailableModel{}+var BuiltInModelProvidersByModelId = map[ModelId][]BaseModelUsesProvider{}-// var AvailableModels = []*AvailableModel{-// // Direct OpenAI models-// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },-// {-// 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,-// },-// },-// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// Description: "OpenAI o3-mini-high",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenAI,-// ModelName: "o3-mini",-// ModelId: "openai/o3-mini-high",-// MaxTokens: 200000,-// MaxOutputTokens: 100000,-// ReservedOutputTokens: 30000,-// ApiKeyEnvVar: OpenAIEnvVar,-// ModelCompatibility: fullCompatibility,-// BaseUrl: OpenAIV1BaseUrl,-// PreferredModelOutputFormat: ModelOutputFormatToolCallJson,-// RoleParamsDisabled: true,-// ReasoningEffortEnabled: true,-// ReasoningEffort: ReasoningEffortHigh,-// },-// },-// {-// Description: "OpenAI o3-mini-medium",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenAI,-// ModelName: "o3-mini",-// ModelId: "openai/o3-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,-// },-// },-// {-// Description: "OpenAI o3-mini-low",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenAI,-// ModelName: "o3-mini",-// ModelId: "openai/o3-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,-// },-// },--// // OpenRouter models-// {-// 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,-// },-// },-// {-// 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,-// },-// },-// {-// 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,-// },-// },-// {-// Description: "Anthropic Claude 3.7 Sonnet",-// DefaultMaxConvoTokens: 15000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderAnthropic,-// ModelName: "claude-3-7-sonnet-latest",-// ModelId: "claude-3-7-sonnet-latest",-// MaxTokens: 200000,-// MaxOutputTokens: 128000,-// ReservedOutputTokens: 20000,-// SupportsCacheControl: true,-// ApiKeyEnvVar: ApiKeyByProvider[ModelProviderAnthropic],-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderAnthropic],-// PreferredModelOutputFormat: 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,-// },-// },-// {-// 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,-// },-// },-// {-// 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,-// },-// },--// {-// Description: "Google Gemini Pro 2.5 Preview via OpenRouter",-// DefaultMaxConvoTokens: 75000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenRouter,-// ModelName: "google/gemini-2.5-pro-preview-03-25",-// ModelId: "google/gemini-2.5-pro-preview-03-25",-// MaxTokens: 1048576,-// MaxOutputTokens: 65535,-// ReservedOutputTokens: 65535,-// ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],-// PreferredModelOutputFormat: ModelOutputFormatXml,-// SupportsCacheControl: true,-// },-// },--// {-// 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,-// },-// },--// {-// Description: "Google Gemini Pro 1.5 Pro via AI Studio",-// DefaultMaxConvoTokens: 75000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderGoogleAIStudio,-// ModelName: "models/gemini-1.5-pro",-// ModelId: "models/gemini-1.5-pro",-// MaxTokens: 2000000,-// MaxOutputTokens: 8192,-// ReservedOutputTokens: 8192,-// ApiKeyEnvVar: GoogleAIStudioApiKeyEnvVar,-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderGoogleAIStudio],-// PreferredModelOutputFormat: 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// // 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },-// {-// 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,-// },-// },-// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// 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,-// },-// },--// {-// Description: "OpenAI o3-mini-high via OpenRouter",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenRouter,-// ModelName: "openai/o3-mini",-// ModelId: "openai/o3-mini-high",-// MaxTokens: 200000,-// MaxOutputTokens: 100000,-// ReservedOutputTokens: 40000,-// ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],-// PreferredModelOutputFormat: ModelOutputFormatToolCallJson,-// SystemPromptDisabled: true,-// RoleParamsDisabled: true,-// ReasoningEffortEnabled: true,-// ReasoningEffort: ReasoningEffortHigh,-// },-// },-// {-// Description: "OpenAI o3-mini-medium via OpenRouter",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenRouter,-// ModelName: "openai/o3-mini",-// ModelId: "openai/o3-mini-medium",-// MaxTokens: 200000,-// MaxOutputTokens: 100000,-// ReservedOutputTokens: 40000,-// ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],-// PreferredModelOutputFormat: ModelOutputFormatToolCallJson,-// SystemPromptDisabled: true,-// RoleParamsDisabled: true,-// ReasoningEffortEnabled: true,-// ReasoningEffort: ReasoningEffortMedium,-// },-// },-// {-// Description: "OpenAI o3-mini-low via OpenRouter",-// DefaultMaxConvoTokens: 10000,-// BaseModelConfig: BaseModelConfig{-// Provider: ModelProviderOpenRouter,-// ModelName: "openai/o3-mini",-// ModelId: "openai/o3-mini-low",-// MaxTokens: 200000,-// MaxOutputTokens: 100000,-// ReservedOutputTokens: 40000,-// ApiKeyEnvVar: ApiKeyByProvider[ModelProviderOpenRouter],-// ModelCompatibility: fullCompatibility,-// BaseUrl: BaseUrlByProvider[ModelProviderOpenRouter],-// PreferredModelOutputFormat: ModelOutputFormatToolCallJson,-// SystemPromptDisabled: true,-// RoleParamsDisabled: true,-// ReasoningEffortEnabled: true,-// ReasoningEffort: ReasoningEffortLow,-// },-// },-// }+var AvailableModels = []*AvailableModel{}var AvailableModelsByComposite = map[string]*AvailableModel{}func init() {+ for _, model := range BuiltInModels {+ AvailableModels = append(AvailableModels, model.ToAvailableModels()...)++ var addVariants func(variants []BaseModelConfigVariant, baseId ModelId)+ addVariants = func(variants []BaseModelConfigVariant, baseId ModelId) {+ for _, variant := range variants {+ var modelId ModelId+ if variant.IsBaseVariant {+ modelId = baseId+ } else {+ modelId = ModelId(strings.Join([]string{string(baseId), string(variant.VariantTag)}, "-"))+ }++ if len(variant.Variants) > 0 {+ addVariants(variant.Variants, modelId)+ continue+ }++ BuiltInModelProvidersByModelId[modelId] = model.Providers+ }+ }++ if len(model.Variants) > 0 {+ addVariants(model.Variants, ModelId(string(model.ModelTag)))+ } else {+ BuiltInModelProvidersByModelId[ModelId(string(model.ModelTag))] = model.Providers+ }+ }++ spew.Dump(BuiltInModelProvidersByModelId)for _, model := range AvailableModels {if model.Description == "" {diff --git a/app/shared/ai_models_data_models.go b/app/shared/ai_models_data_models.goindex 7ed95ee9..f576e0cf 100644--- a/app/shared/ai_models_data_models.go+++ b/app/shared/ai_models_data_models.go@@ -45,7 +45,7 @@ type BaseModelShared struct {// for openai, token estimate padding percentageTokenEstimatePaddingPct float64 `json:"tokenEstimatePaddingPct"`- ModelCompatibility ModelCompatibility `json:"modelCompatibility"`+ ModelCompatibility}type BaseModelProviderConfig struct {@@ -54,7 +54,8 @@ type BaseModelProviderConfig struct {}type BaseModelConfig struct {- ModelId ModelId `json:"modelId"`+ ModelId ModelId `json:"modelId"`+ ModelTag ModelTag `json:"modelTag"`BaseModelSharedBaseModelProviderConfig}@@ -70,15 +71,20 @@ type BaseModelConfigSchema struct {Description string `json:"description"`DefaultMaxConvoTokens int `json:"defaultMaxConvoTokens"`BaseModelShared++ RequiresVariantOverrides []string `json:"requiresVariantOverrides"`+Variants []BaseModelConfigVariant `json:"variants"`Providers []BaseModelUsesProvider `json:"providers"`}type BaseModelConfigVariant struct {- VariantTag VariantTag `json:"variantTag"`- Description string `json:"description"`- BaseModelShared- Variants []BaseModelConfigVariant `json:"variants"`+ 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 {@@ -90,18 +96,121 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {panic(fmt.Sprintf("provider %s not found", provider.Provider))}- avail = append(avail, &AvailableModel{- Description: b.Description,- DefaultMaxConvoTokens: b.DefaultMaxConvoTokens,- BaseModelConfig: BaseModelConfig{- ModelId: ModelId(strings.Join([]string{string(provider.Provider), string(provider.ModelName)}, "/")),- BaseModelShared: b.BaseModelShared,- BaseModelProviderConfig: BaseModelProviderConfig{- ModelProviderConfigSchema: providerConfig,- ModelName: provider.ModelName,+ 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,+ },},- },- })+ })+ }++ 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,+ BaseModelShared: mergedOverrides,+ BaseModelProviderConfig: BaseModelProviderConfig{+ ModelProviderConfigSchema: providerConfig,+ ModelName: provider.ModelName,+ },+ },+ })+ }++ 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,+ },+ },+ })+ }+ }++ addVariants(b.Variants, variantParams{+ BaseId: ModelId(string(b.ModelTag)),+ BaseDescription: b.Description,+ Overrides: b.BaseModelShared,+ RequiresVariantOverrides: b.RequiresVariantOverrides,+ })+ }}return avail}@@ -157,8 +266,8 @@ type ModelRoleModelConfig struct {}type ModelRoleConfigSchema struct {- Role ModelRole `json:"role"`- ModelTag ModelTag `json:"modelTag"`+ Role ModelRole `json:"role"`+ ModelId ModelId `json:"modelId"`Temperature float32 `json:"temperature"`TopP float32 `json:"topP"`@@ -172,10 +281,12 @@ type ModelRoleConfigSchema struct {}func (m *ModelRoleConfigSchema) ToModelRoleConfig() ModelRoleConfig {+ providers, ok := BuiltInModelProvidersByModelId[m.ModelId]+ if !ok {+ panic(fmt.Sprintf("no providers found for model %s", m.ModelId))+ }- modelSchema-- return m.toModelRoleConfig()+ return m.toModelRoleConfig(providers)}func (m *ModelRoleConfigSchema) toModelRoleConfig(providers []BaseModelUsesProvider) ModelRoleConfig {@@ -186,7 +297,7 @@ func (m *ModelRoleConfigSchema) toModelRoleConfig(providers []BaseModelUsesProviprovider := providers[0]tail := providers[1:]- config := GetAvailableModel(m.Provider, m.ModelTag)+ config := GetAvailableModel(provider.Provider, m.ModelId)var largeContextFallback *ModelRoleConfigif m.LargeContextFallback != nil {@@ -209,6 +320,13 @@ func (m *ModelRoleConfigSchema) toModelRoleConfig(providers []BaseModelUsesProvistrongModel = &c}+ // recursive waterfall fallback chain for all the provider options+ var missingKeyFallback *ModelRoleConfig+ if len(tail) > 0 {+ c := m.toModelRoleConfig(tail)+ missingKeyFallback = &c+ }+return ModelRoleConfig{Role: m.Role,BaseModelConfig: config.BaseModelConfig,@@ -220,6 +338,7 @@ func (m *ModelRoleConfigSchema) toModelRoleConfig(providers []BaseModelUsesProviLargeOutputFallback: largeOutputFallback,ErrorFallback: errorFallback,StrongModel: strongModel,+ MissingKeyFallback: missingKeyFallback,}}@@ -272,17 +391,59 @@ func (p PlannerRoleConfig) Value() (driver.Value, error) {}type ModelPackSchema struct {- Name string `json:"name"`- Description string `json:"description"`- Planner PlannerRoleConfig `json:"planner"`- Coder *ModelRoleConfig `json:"coder"`- PlanSummary ModelRoleConfig `json:"planSummary"`- Builder ModelRoleConfig `json:"builder"`- WholeFileBuilder *ModelRoleConfig `json:"wholeFileBuilder"` // optional, defaults to builder model — access via GetWholeFileBuilder()- Namer ModelRoleConfig `json:"namer"`- CommitMsg ModelRoleConfig `json:"commitMsg"`- ExecStatus ModelRoleConfig `json:"execStatus"`- Architect *ModelRoleConfig `json:"contextLoader"`+ Name string `json:"name"`+ Description string `json:"description"`+ Planner ModelRoleConfigSchema `json:"planner"`+ Coder *ModelRoleConfigSchema `json:"coder"`+ PlanSummary ModelRoleConfigSchema `json:"planSummary"`+ Builder ModelRoleConfigSchema `json:"builder"`+ WholeFileBuilder *ModelRoleConfigSchema `json:"wholeFileBuilder"` // optional, defaults to builder model — access via GetWholeFileBuilder()+ Namer ModelRoleConfigSchema `json:"namer"`+ CommitMsg ModelRoleConfigSchema `json:"commitMsg"`+ ExecStatus ModelRoleConfigSchema `json:"execStatus"`+ Architect *ModelRoleConfigSchema `json:"contextLoader"`+}++func (m *ModelPackSchema) ToModelPack() ModelPack {+ var (+ coder *ModelRoleConfig+ wholeFileBuilder *ModelRoleConfig+ architect *ModelRoleConfig+ )++ if m.Coder != nil {+ c := m.Coder.ToModelRoleConfig()+ coder = &c+ }++ if m.WholeFileBuilder != nil {+ c := m.WholeFileBuilder.ToModelRoleConfig()+ wholeFileBuilder = &c+ }++ if m.Architect != nil {+ c := m.Architect.ToModelRoleConfig()+ architect = &c+ }++ return ModelPack{+ Name: m.Name,+ Description: m.Description,+ Planner: PlannerRoleConfig{+ ModelRoleConfig: m.Planner.ToModelRoleConfig(),+ PlannerModelConfig: PlannerModelConfig{+ MaxConvoTokens: m.Planner.MaxConvoTokens,+ },+ },+ Coder: coder,+ PlanSummary: m.PlanSummary.ToModelRoleConfig(),+ Builder: m.Builder.ToModelRoleConfig(),+ WholeFileBuilder: wholeFileBuilder,+ Namer: m.Namer.ToModelRoleConfig(),+ CommitMsg: m.CommitMsg.ToModelRoleConfig(),+ ExecStatus: m.ExecStatus.ToModelRoleConfig(),+ Architect: architect,+ }}type ModelPack struct {diff --git a/app/shared/ai_models_packs.go b/app/shared/ai_models_packs.goindex 576cc7c3..b0060b68 100644--- a/app/shared/ai_models_packs.go+++ b/app/shared/ai_models_packs.go@@ -33,424 +33,274 @@ var BuiltInModelPacks = []*ModelPack{var DefaultModelPack *ModelPack = &DailyDriverModelPack-func init() {- DailyDriverModelPack = ModelPack{- 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: *claude37Sonnet(ModelRolePlanner, &modelConfig{- largeContextFallback: geminipro25preview(ModelRolePlanner, &modelConfig{- largeContextFallback: gemini15pro(ModelRolePlanner, nil),- errorFallback: gemini15pro(ModelRolePlanner, nil),- }),- }),- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet"),- },- Coder: claude37Sonnet(ModelRoleCoder, &modelConfig{- largeContextFallback: openaigpt41(ModelRoleCoder, nil),- }),- Architect: claude37Sonnet(ModelRoleArchitect, &modelConfig{- largeContextFallback: geminipro25preview(ModelRolePlanner, &modelConfig{- largeContextFallback: gemini15pro(ModelRolePlanner, nil),- errorFallback: gemini15pro(ModelRolePlanner, nil),- }),- }),- PlanSummary: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- }),- WholeFileBuilder: openaio4miniMediumWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- ReasoningModelPack = ModelPack{- 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: PlannerRoleConfig{- ModelRoleConfig: *claude37SonnetThinkingHidden(ModelRolePlanner, nil),- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet:thinking-hidden"),- },- Coder: claude37SonnetThinkingHidden(ModelRoleCoder, nil),- PlanSummary: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- }),-- WholeFileBuilder: openaio4miniMediumWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- StrongModelPack = ModelPack{- 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: PlannerRoleConfig{- ModelRoleConfig: *openaio3high(ModelRolePlanner, nil),- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenAI, "openai/o3-high"),- },- Architect: openaio3high(ModelRoleArchitect, nil),- Coder: claude37SonnetThinkingHidden(ModelRoleCoder, nil),- PlanSummary: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- WholeFileBuilder: openaio4miniHighWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniMediumWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- CheapModelPack = ModelPack{- 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: *openaio4miniMediumWitho3MiniFallback(ModelRolePlanner, nil),- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenAI, "openai/o3-mini-medium"),- },- Coder: openaigpt41(ModelRoleCoder, nil),- PlanSummary: *openaigpt41mini(ModelRolePlanSummary, nil),- Builder: *openaio4miniLowWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, nil),- }),- WholeFileBuilder: openaio4miniLowWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),+func getModelRoleConfig(role ModelRole, modelId ModelId, fns ...func(*ModelRoleConfigSchema)) ModelRoleConfigSchema {+ c := ModelRoleConfigSchema{+ Role: role,+ ModelId: modelId,+ Temperature: DefaultConfigByRole[role].Temperature,+ TopP: DefaultConfigByRole[role].TopP,}-- OSSModelPack = ModelPack{- 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),- }-- OpenAIModelPack = ModelPack{- 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: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- }),- WholeFileBuilder: openaio4miniMediumWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- AnthropicModelPack = ModelPack{- 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: PlannerRoleConfig{- ModelRoleConfig: *claude37Sonnet(ModelRolePlanner, nil),- PlannerModelConfig: getPlannerModelConfig(ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet"),- },- PlanSummary: *claude35haiku(ModelRolePlanSummary, nil),- Builder: *claude37Sonnet(ModelRoleBuilder, nil),- WholeFileBuilder: claude37Sonnet(ModelRoleWholeFileBuilder, nil),- Namer: *claude35haiku(ModelRoleName, nil),- CommitMsg: *claude35haiku(ModelRoleCommitMsg, nil),- ExecStatus: *claude37Sonnet(ModelRoleExecStatus, nil),- }-- GeminiPreviewModelPack = ModelPack{- 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-03-25"),- },- Coder: geminipro25preview(ModelRoleCoder, nil),- PlanSummary: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- }),- WholeFileBuilder: openaio4miniMediumWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- GeminiExperimentalModelPack = ModelPack{- 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: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, &modelConfig{- strongModel: openaio4miniHighWitho3MiniFallback(ModelRoleBuilder, nil),- }),- WholeFileBuilder: openaio4miniMediumWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniLowWitho3MiniFallback(ModelRoleExecStatus, nil),+ for _, f := range fns {+ f(&c)}-- R1PlannerModelPack = ModelPack{- 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: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, nil),- WholeFileBuilder: openaio4miniLowWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniMediumWitho3MiniFallback(ModelRoleExecStatus, nil),- }-- PerplexityPlannerModelPack = ModelPack{- 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: *openaio4miniLowWitho3MiniFallback(ModelRolePlanSummary, nil),- Builder: *openaio4miniMediumWitho3MiniFallback(ModelRoleBuilder, nil),- WholeFileBuilder: openaio4miniLowWitho3MiniFallback(ModelRoleWholeFileBuilder, nil),- Namer: *openaigpt41mini(ModelRoleName, nil),- CommitMsg: *openaigpt41mini(ModelRoleCommitMsg, nil),- ExecStatus: *openaio4miniMediumWitho3MiniFallback(ModelRoleExecStatus, nil),- }-+ return c}-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{}- }-- 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 getLargeContextFallback(modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {+ return func(c *ModelRoleConfigSchema) {+ n := getModelRoleConfig(c.Role, modelId)+ for _, f := range fns {+ f(&n)+ }+ c.LargeContextFallback = &n}}-func claude37Sonnet(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- return getModelConfig(role, ModelProviderOpenRouter, "anthropic/claude-3.7-sonnet", 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 getErrorFallback(modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {+ return func(c *ModelRoleConfigSchema) {+ n := getModelRoleConfig(c.Role, modelId)+ for _, f := range fns {+ f(&n)+ }+ c.ErrorFallback = &n+ }}-func gemini15pro(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-pro-1.5", fallbacks)+func getStrongModelFallback(modelId ModelId, fns ...func(*ModelRoleConfigSchema)) func(*ModelRoleConfigSchema) {+ return func(c *ModelRoleConfigSchema) {+ n := getModelRoleConfig(c.Role, modelId)+ for _, f := range fns {+ f(&n)+ }+ c.StrongModel = &n+ }}-func gemini25propreview(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- return getModelConfig(role, ModelProviderOpenRouter, "google/gemini-2.5-pro-preview-03-25", fallbacks)+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 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 init() {+ defaultBuilder := getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium",+ getStrongModelFallback("openai/o4-mini-high"),+ getErrorFallback("openai/o3-mini-medium"),+ )-func openaigpt41mini(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/gpt-4.1-mini", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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.",+ Planner: getModelRoleConfig(ModelRolePlanner, "anthropic/claude-3.7-sonnet",+ getLargeContextFallback("google/gemini-2.5-pro-preview",+ getLargeContextFallback("google/gemini-pro-1.5"),+ getErrorFallback("google/gemini-pro-1.5"),+ ),+ ),+ Architect: Pointer(getModelRoleConfig(ModelRoleArchitect, "anthropic/claude-3.7-sonnet",+ getLargeContextFallback("google/gemini-2.5-pro-preview",+ getLargeContextFallback("google/gemini-pro-1.5"),+ getErrorFallback("google/gemini-pro-1.5"),+ ),+ )),+ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet",+ getLargeContextFallback("openai/gpt-4.1"),+ )),+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low",+ getErrorFallback("openai/o3-mini-low"),+ ),+ Builder: defaultBuilder,+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium",+ getErrorFallback("openai/o3-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",+ getErrorFallback("openai/o3-mini-low"),+ ),}- 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{}+ ReasoningPackSchema = ModelPackSchema{+ Name: "reasoning",+ Description: "Same blend, but Sonnet:thinking with visible reasoning.",+ 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", getErrorFallback("openai/o3-mini-low")),+ Builder: defaultBuilder,+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium",+ getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-low")),}- 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{}+ 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", getErrorFallback("openai/o3-mini-low")),+ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-high", getErrorFallback("openai/o3-mini-high")),+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-high", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-medium")),}- 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{}+ 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", getErrorFallback("openai/o3-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", getStrongModelFallback("openai/o4-mini-medium"), getErrorFallback("openai/o3-mini-low")),+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-low")),}- 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{}+ 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/deepseek-r1-reasoning"),+ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "deepseek/deepseek-chat-v3-0324")),+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "deepseek/deepseek-r1-no-reasoning"),+ Builder: getModelRoleConfig(ModelRoleBuilder, "deepseek/deepseek-r1-no-reasoning"),+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "deepseek/deepseek-r1-no-reasoning")),+ Namer: getModelRoleConfig(ModelRoleName, "qwen/qwen-2.5-coder-32b-instruct"),+ CommitMsg: getModelRoleConfig(ModelRoleCommitMsg, "qwen/qwen-2.5-coder-32b-instruct"),+ ExecStatus: getModelRoleConfig(ModelRoleExecStatus, "deepseek/deepseek-r1-no-reasoning"),}- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-low", fallbacks)-}-func _openaio3miniHigh(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-mini-high", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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", getErrorFallback("openai/o3-mini-low")),+ Builder: defaultBuilder,+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-low")),}- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-mini-high", fallbacks)-}-func _openaio3miniMedium(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-mini-medium", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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"),}- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-mini-medium", fallbacks)-}-func _openaio3miniLow(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o3-mini-low", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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", getErrorFallback("openai/o3-mini-low")),+ Builder: defaultBuilder,+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-low")),}- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o3-mini-low", fallbacks)-}-func openaio4miniHigh(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-high", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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-03-25"),+ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "google/gemini-2.5-pro-exp-03-25")),+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low", getErrorFallback("openai/o3-mini-low")),+ Builder: defaultBuilder,+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-medium", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-low")),}- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-high", fallbacks)-}-func openaio4miniHighWitho3MiniFallback(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-high", nil)- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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/deepseek-r1-reasoning"),+ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet")),+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low", getErrorFallback("openai/o3-mini-low")),+ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium", getErrorFallback("openai/o3-mini-medium")),+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-medium")),}- fallbacks.errorFallback = _openaio3miniHigh(role, nil)- fallbacks.missingKeyFallback = openrouterFallback- return getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-high", fallbacks)-}-func openaio4miniMediumWitho3MiniFallback(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-medium", nil)-- if fallbacks == nil {- fallbacks = &modelConfig{}+ 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"),+ Coder: Pointer(getModelRoleConfig(ModelRoleCoder, "anthropic/claude-3.7-sonnet")),+ PlanSummary: getModelRoleConfig(ModelRolePlanSummary, "openai/o4-mini-low", getErrorFallback("openai/o3-mini-low")),+ Builder: getModelRoleConfig(ModelRoleBuilder, "openai/o4-mini-medium", getErrorFallback("openai/o3-mini-medium")),+ WholeFileBuilder: Pointer(getModelRoleConfig(ModelRoleWholeFileBuilder, "openai/o4-mini-low", getErrorFallback("openai/o3-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", getErrorFallback("openai/o3-mini-medium")),}- fallbacks.errorFallback = _openaio3miniMedium(role, nil)- fallbacks.missingKeyFallback = openrouterFallback- res := getModelConfig(role, ModelProviderOpenAI, "openai/o4-mini-medium", fallbacks)- return res-}-func openaio4miniLowWitho3MiniFallback(role ModelRole, fallbacks *modelConfig) *ModelRoleConfig {- openrouterFallback := getModelConfig(role, ModelProviderOpenRouter, "openai/o4-mini-low", 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 = _openaio3miniLow(role, nil)- 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-03-25", fallbacks)+ DefaultModelPack = &DailyDriverModelPack}diff --git a/app/shared/utils.go b/app/shared/utils.goindex 6c57c8ef..23495203 100644--- a/app/shared/utils.go+++ b/app/shared/utils.go@@ -4,13 +4,16 @@ import ("bytes""crypto/rand""fmt"- "reflect""regexp""strings""time""unicode/utf8")+func Pointer[T any](v T) *T {+ return &v+}+const TsFormat = "2006-01-02T15:04:05.999Z"func StringTs() string {@@ -140,19 +143,6 @@ func ReplaceReverse(s, old, new string, n int) string {return res}-func Merge[T any](base T, ov T) T {- rvBase := reflect.ValueOf(&base).Elem() // addressable- rvOv := reflect.ValueOf(ov)-- for i := 0; i < rvBase.NumField(); i++ {- fOv := rvOv.Field(i)- if !fOv.IsZero() { // ← built-in zero test- rvBase.Field(i).Set(fOv)- }- }- return base-}-func NormalizeEOL(data []byte) []byte {if !looksTextish(data) {return datadiff --git a/app/shared/utils_struct.go b/app/shared/utils_struct.gonew file mode 100644index 00000000..e9de0a24--- /dev/null+++ b/app/shared/utils_struct.go@@ -0,0 +1,38 @@+package shared++import "reflect"++func Merge[T any](base T, ov T) T {+ rvBase := reflect.ValueOf(&base).Elem() // addressable+ rvOv := reflect.ValueOf(ov)++ for i := 0; i < rvBase.NumField(); i++ {+ fOv := rvOv.Field(i)+ if !fOv.IsZero() { // ← built-in zero test+ rvBase.Field(i).Set(fOv)+ }+ }+ return base+}++// FieldsDefined reports whether every name in fields is present on the+// (possibly-pointer) struct v. It returns the first missing field name+// (empty string means all present).+func FieldsDefined(v any, fields []string) (ok bool, missing string) {+ rv := reflect.ValueOf(v)+ if rv.Kind() == reflect.Pointer {+ rv = rv.Elem()+ }+ if rv.Kind() != reflect.Struct {+ panic("FieldsDefined: value must be a struct or *struct")+ }++ rt := rv.Type() // a reflect.Type is a bit faster for look-ups++ for _, name := range fields {+ if _, found := rt.FieldByName(name); !found {+ return false, name+ }+ }+ return true, ""+}