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
fixes for custom models and custom providers. fix to only use :nitro suffix with openrouter if not using another special suffix (like :free)
Dane Schneider74ee022cdf49
diff --git a/app/cli/cmd/model_packs.go b/app/cli/cmd/model_packs.goindex a7035680..ab311ba2 100644--- a/app/cli/cmd/model_packs.go+++ b/app/cli/cmd/model_packs.go@@ -127,11 +127,19 @@ func showModelPack(cmd *cobra.Command, args []string) {term.StartSpinner("")customModelPacks, apiErr := api.Client.ListModelPacks()- term.StopSpinner()-if apiErr != nil {term.OutputErrorAndExit("Error fetching models: %v", apiErr)}+ customModels, err := api.Client.ListCustomModels()+ if err != nil {+ term.OutputErrorAndExit("Error fetching custom models: %v", err)+ }+ customModelsById := make(map[shared.ModelId]*shared.CustomModel)+ for _, m := range customModels {+ customModelsById[m.ModelId] = m+ }++ term.StopSpinner()modelPacks := []*shared.ModelPack{}modelPacks = append(modelPacks, customModelPacks...)@@ -186,7 +194,7 @@ func showModelPack(cmd *cobra.Command, args []string) {return}- renderModelPack(modelPack, allProperties)+ renderModelPack(modelPack, customModelsById, allProperties)fmt.Println()diff --git a/app/cli/cmd/models.go b/app/cli/cmd/models.goindex 0ab22a2d..0813e590 100644--- a/app/cli/cmd/models.go+++ b/app/cli/cmd/models.go@@ -361,7 +361,7 @@ func renderSettings(settings *shared.PlanSettings, allProperties bool) {modelPack := settings.GetModelPack()color.New(color.Bold, term.ColorHiCyan).Println("🎛️ Current Model Pack")- renderModelPack(modelPack, allProperties)+ renderModelPack(modelPack, settings.CustomModelsById, allProperties)if allProperties {color.New(color.Bold, term.ColorHiCyan).Println("🧠 Planner Defaults")@@ -369,15 +369,15 @@ func renderSettings(settings *shared.PlanSettings, allProperties bool) {table.SetAutoWrapText(false)table.SetHeader([]string{"Max Tokens", "Max Convo Tokens"})table.Append([]string{- fmt.Sprintf("%d", modelPack.Planner.GetFinalLargeContextFallback().GetSharedBaseConfig().MaxTokens),- fmt.Sprintf("%d", modelPack.Planner.GetMaxConvoTokens()),+ fmt.Sprintf("%d", modelPack.Planner.GetFinalLargeContextFallback().GetSharedBaseConfig(settings).MaxTokens),+ fmt.Sprintf("%d", modelPack.Planner.GetMaxConvoTokens(settings)),})table.Render()fmt.Println()}}-func renderModelPack(modelPack *shared.ModelPack, allProperties bool) {+func renderModelPack(modelPack *shared.ModelPack, customModelsById map[shared.ModelId]*shared.CustomModel, allProperties bool) {table := tablewriter.NewWriter(os.Stdout)table.SetAutoFormatHeaders(false)table.SetAutoWrapText(true)@@ -428,7 +428,9 @@ func renderModelPack(modelPack *shared.ModelPack, allProperties bool) {var topP float32var disabled bool- if config.GetSharedBaseConfig().RoleParamsDisabled {+ sharedBaseConfig := config.GetSharedBaseConfigWithCustomModels(customModelsById)++ if sharedBaseConfig.RoleParamsDisabled {temp = 1topP = 1disabled = true@@ -457,7 +459,7 @@ func renderModelPack(modelPack *shared.ModelPack, allProperties bool) {row = append(row, []string{tempStr,topPStr,- fmt.Sprintf("%d 🪙", config.GetSharedBaseConfig().MaxTokens-config.GetReservedOutputTokens()),+ fmt.Sprintf("%d 🪙", sharedBaseConfig.MaxTokens-config.GetReservedOutputTokens(customModelsById)),}...)}table.Append(row)diff --git a/app/server/db/settings_helpers.go b/app/server/db/settings_helpers.goindex 5ede4d3b..f30a757f 100644--- a/app/server/db/settings_helpers.go+++ b/app/server/db/settings_helpers.go@@ -18,19 +18,14 @@ func GetPlanSettings(plan *Plan) (settings *shared.PlanSettings, err error) {planDir := getPlanDir(plan.OrgId, plan.Id)settingsPath := filepath.Join(planDir, "settings.json")- customModelPacks, err := ListModelPacks(plan.OrgId)+ result, err := GetApiCustomModels(plan.OrgId)if err != nil {- return nil, fmt.Errorf("error getting custom model packs: %v", err)- }-- apiModelPacks := make([]*shared.ModelPack, len(customModelPacks))- for i, modelPack := range customModelPacks {- apiModelPacks[i] = modelPack.ToApi()+ return nil, fmt.Errorf("error getting custom models: %v", err)}defer func() {if settings != nil {- settings.Configure(apiModelPacks, os.Getenv("PLANDEX_CLOUD") != "")+ settings.Configure(result.CustomModelPacks, result.CustomModels, result.CustomProviders, os.Getenv("PLANDEX_CLOUD") != "")}}()@@ -103,19 +98,14 @@ func StorePlanSettings(plan *Plan, settings shared.PlanSettings) error {}func GetOrgDefaultSettings(orgId string) (settings *shared.PlanSettings, err error) {- customModelPacks, err := ListModelPacks(orgId)+ result, err := GetApiCustomModels(orgId)if err != nil {- return nil, fmt.Errorf("error getting custom model packs: %v", err)- }-- apiModelPacks := make([]*shared.ModelPack, len(customModelPacks))- for i, modelPack := range customModelPacks {- apiModelPacks[i] = modelPack.ToApi()+ return nil, fmt.Errorf("error getting custom models: %v", err)}defer func() {if settings != nil {- settings.Configure(apiModelPacks, os.Getenv("PLANDEX_CLOUD") != "")+ settings.Configure(result.CustomModelPacks, result.CustomModels, result.CustomProviders, os.Getenv("PLANDEX_CLOUD") != "")}}()@@ -141,19 +131,14 @@ func GetOrgDefaultSettings(orgId string) (settings *shared.PlanSettings, err err}func GetOrgDefaultSettingsForUpdate(orgId string, tx *sqlx.Tx) (settings *shared.PlanSettings, err error) {- customModelPacks, err := ListModelPacks(orgId)+ result, err := GetApiCustomModels(orgId)if err != nil {- return nil, fmt.Errorf("error getting custom model packs: %v", err)- }-- apiModelPacks := make([]*shared.ModelPack, len(customModelPacks))- for i, modelPack := range customModelPacks {- apiModelPacks[i] = modelPack.ToApi()+ return nil, fmt.Errorf("error getting custom models: %v", err)}defer func() {if settings != nil {- settings.Configure(apiModelPacks, os.Getenv("PLANDEX_CLOUD") != "")+ settings.Configure(result.CustomModelPacks, result.CustomModels, result.CustomProviders, os.Getenv("PLANDEX_CLOUD") != "")}}()@@ -193,3 +178,74 @@ func StoreOrgDefaultSettings(orgId string, settings *shared.PlanSettings, tx *sqreturn nil}++type GetCustomModelsResult struct {+ CustomModels []*shared.CustomModel+ CustomProviders []*shared.CustomProvider+ CustomModelPacks []*shared.ModelPack+}++func GetApiCustomModels(orgId string) (result *GetCustomModelsResult, err error) {+ var customModels []*CustomModel+ var customProviders []*CustomProvider+ var customModelPacks []*ModelPack++ errCh := make(chan error, 3)++ go func() {+ res, err := ListModelPacks(orgId)+ if err != nil {+ errCh <- fmt.Errorf("error getting custom model packs: %v", err)+ }+ customModelPacks = res+ errCh <- nil+ }()++ go func() {+ res, err := ListCustomModels(orgId)+ if err != nil {+ errCh <- fmt.Errorf("error getting custom models: %v", err)+ }+ customModels = res+ errCh <- nil+ }()++ go func() {+ res, err := ListCustomProviders(orgId)+ if err != nil {+ errCh <- fmt.Errorf("error getting custom providers: %v", err)+ }+ customProviders = res+ errCh <- nil+ }()++ for i := 0; i < 3; i++ {+ err := <-errCh+ if err != nil {+ return nil, err+ }+ }++ apiModelPacks := make([]*shared.ModelPack, len(customModelPacks))+ for i, modelPack := range customModelPacks {+ apiModelPacks[i] = modelPack.ToApi()+ }++ apiCustomModels := make([]*shared.CustomModel, len(customModels))+ for i, model := range customModels {+ apiCustomModels[i] = model.ToApi()+ }++ apiCustomProviders := make([]*shared.CustomProvider, len(customProviders))+ for i, provider := range customProviders {+ apiCustomProviders[i] = provider.ToApi()+ }++ result = &GetCustomModelsResult{+ CustomModels: apiCustomModels,+ CustomProviders: apiCustomProviders,+ CustomModelPacks: apiModelPacks,+ }++ return result, nil+}diff --git a/app/server/handlers/client_helper.go b/app/server/handlers/client_helper.goindex ae6ca9a3..dd3839f3 100644--- a/app/server/handlers/client_helper.go+++ b/app/server/handlers/client_helper.go@@ -8,6 +8,7 @@ import ("plandex-server/hooks""plandex-server/model""plandex-server/types"+ shared "plandex-shared")type initClientsParams struct {@@ -19,7 +20,8 @@ type initClientsParams struct {authVars map[string]string- plan *db.Plan+ plan *db.Plan+ settings *shared.PlanSettings}type initClientsResult struct {@@ -29,6 +31,7 @@ type initClientsResult struct {func initClients(params initClientsParams) initClientsResult {w := params.w+ settings := params.settingsvar authVars map[string]stringif params.authVars != nil {@@ -63,7 +66,7 @@ func initClients(params initClientsParams) initClientsResult {return initClientsResult{}}- clients := model.InitClients(authVars)+ clients := model.InitClients(authVars, settings)return initClientsResult{clients: clients,diff --git a/app/server/handlers/context_helper.go b/app/server/handlers/context_helper.goindex 5aab14de..a4ef5b89 100644--- a/app/server/handlers/context_helper.go+++ b/app/server/handlers/context_helper.go@@ -110,6 +110,7 @@ func loadContexts(openAIOrgId: context.OpenAIOrgId,authVars: context.AuthVars,plan: plan,+ settings: settings,},)@@ -123,7 +124,7 @@ func loadContexts(// ensure image compatibility if we're loading an imagefor _, context := range *loadReq {if context.ContextType == shared.ContextImageType {- if !settings.GetModelPack().Planner.GetSharedBaseConfig().HasImageSupport {+ if !settings.GetModelPack().Planner.GetSharedBaseConfig(settings).HasImageSupport {log.Printf("Error loading context: %s does not support images in context\n", settings.GetModelPack().Planner.ModelId)http.Error(w, fmt.Sprintf("Error loading context: %s does not support images in context", settings.GetModelPack().Planner.ModelId), http.StatusBadRequest)return nil, nildiff --git a/app/server/handlers/plans_changes.go b/app/server/handlers/plans_changes.goindex fa719d6d..8003bb81 100644--- a/app/server/handlers/plans_changes.go+++ b/app/server/handlers/plans_changes.go@@ -192,6 +192,7 @@ func ApplyPlanHandler(w http.ResponseWriter, r *http.Request) {openAIOrgId: requestBody.OpenAIOrgId,authVars: requestBody.AuthVars,plan: plan,+ settings: settings,},)diff --git a/app/server/handlers/plans_exec.go b/app/server/handlers/plans_exec.goindex c2e430e7..f0574b21 100644--- a/app/server/handlers/plans_exec.go+++ b/app/server/handlers/plans_exec.go@@ -39,6 +39,13 @@ func TellPlanHandler(w http.ResponseWriter, r *http.Request) {return}+ settings, err := db.GetPlanSettings(plan)+ if err != nil {+ log.Printf("Error getting plan settings: %v\n", err)+ http.Error(w, "Error getting plan settings", http.StatusInternalServerError)+ return+ }+body, err := io.ReadAll(r.Body)if err != nil {log.Printf("Error reading request body: %v\n", err)@@ -77,6 +84,7 @@ func TellPlanHandler(w http.ResponseWriter, r *http.Request) {openAIOrgId: requestBody.OpenAIOrgId,authVars: requestBody.AuthVars,plan: plan,+ settings: settings,},)err = modelPlan.Tell(modelPlan.TellParams{@@ -119,6 +127,13 @@ func BuildPlanHandler(w http.ResponseWriter, r *http.Request) {return}+ settings, err := db.GetPlanSettings(plan)+ if err != nil {+ log.Printf("Error getting plan settings: %v\n", err)+ http.Error(w, "Error getting plan settings", http.StatusInternalServerError)+ return+ }+body, err := io.ReadAll(r.Body)if err != nil {log.Printf("Error reading request body: %v\n", err)@@ -147,6 +162,7 @@ func BuildPlanHandler(w http.ResponseWriter, r *http.Request) {openAIOrgId: requestBody.OpenAIOrgId,authVars: requestBody.AuthVars,plan: plan,+ settings: settings,},)numBuilds, err := modelPlan.Build(modelPlan.BuildParams{diff --git a/app/server/model/client.go b/app/server/model/client.goindex 348db496..1ba3d8ba 100644--- a/app/server/model/client.go+++ b/app/server/model/client.go@@ -42,9 +42,9 @@ type ClientInfo struct {OpenAIOrgId string}-func InitClients(authVars map[string]string) map[string]ClientInfo {+func InitClients(authVars map[string]string, settings *shared.PlanSettings) map[string]ClientInfo {clients := make(map[string]ClientInfo)- providers := shared.GetProvidersForAuthVars(authVars)+ providers := shared.GetProvidersForAuthVars(authVars, settings)for _, provider := range providers {clients[provider.ToComposite()] = newClient(provider, authVars)@@ -105,29 +105,31 @@ func CreateChatCompletionStream(clients map[string]ClientInfo,authVars map[string]string,modelConfig *shared.ModelRoleConfig,- localProvider shared.ModelProvider,+ settings *shared.PlanSettings,ctx context.Context,req types.ExtendedChatCompletionRequest,) (*ExtendedChatCompletionStream, error) {- providerComposite := modelConfig.GetProviderComposite(authVars, localProvider)+ providerComposite := modelConfig.GetProviderComposite(authVars, settings)_, ok := clients[providerComposite]if !ok {return nil, fmt.Errorf("client not found for provider composite: %s", providerComposite)}- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, localProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, settings)// ensure the model name is set correctly on fallbacksreq.Model = baseModelConfig.ModelName- resolveReq(&req, modelConfig, baseModelConfig)+ resolveReq(&req, modelConfig, baseModelConfig, settings)// choose the fastest provider by latency/throughput on openrouterif baseModelConfig.Provider == shared.ModelProviderOpenRouter {- req.Model += ":nitro"+ if !strings.HasSuffix(string(req.Model), ":nitro") && !strings.HasSuffix(string(req.Model), ":free") && !strings.HasSuffix(string(req.Model), ":floor") {+ req.Model += ":nitro"+ }}- if baseModelConfig.ReasoningBudgetEnabled {+ if baseModelConfig.ReasoningBudget > 0 {req.ReasoningConfig = &types.ReasoningConfig{MaxTokens: baseModelConfig.ReasoningBudget,Exclude: !baseModelConfig.IncludeReasoning,@@ -144,16 +146,16 @@ func CreateChatCompletionStream(}return withStreamingRetries(ctx, func(numTotalRetry int, didProviderFallback bool, modelErr *shared.ModelError) (*ExtendedChatCompletionStream, shared.FallbackResult, error) {- fallbackRes := modelConfig.GetFallbackForModelError(numTotalRetry, didProviderFallback, modelErr, authVars, localProvider)+ fallbackRes := modelConfig.GetFallbackForModelError(numTotalRetry, didProviderFallback, modelErr, authVars, settings)resolvedModelConfig := fallbackRes.ModelRoleConfigif resolvedModelConfig == nil {return nil, fallbackRes, fmt.Errorf("model config is nil")}- providerComposite := resolvedModelConfig.GetProviderComposite(authVars, localProvider)+ providerComposite := resolvedModelConfig.GetProviderComposite(authVars, settings)- baseModelConfig := resolvedModelConfig.GetBaseModelConfig(authVars, localProvider)+ baseModelConfig := resolvedModelConfig.GetBaseModelConfig(authVars, settings)opClient, ok := clients[providerComposite]@@ -183,7 +185,7 @@ func CreateChatCompletionStream("modelConfig.ApiKeyEnvVar": baseModelConfig.ApiKeyEnvVar,})- resp, err := createChatCompletionStreamExtended(resolvedModelConfig, opClient, authVars, localProvider, ctx, req)+ resp, err := createChatCompletionStreamExtended(resolvedModelConfig, opClient, authVars, settings, ctx, req)return resp, fallbackRes, err}, func(resp *ExtendedChatCompletionStream, err error) {})}@@ -192,11 +194,11 @@ func createChatCompletionStreamExtended(modelConfig *shared.ModelRoleConfig,client ClientInfo,authVars map[string]string,- localProvider shared.ModelProvider,+ settings *shared.PlanSettings,ctx context.Context,extendedReq types.ExtendedChatCompletionRequest,) (*ExtendedChatCompletionStream, error) {- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, localProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, settings)// ensure the model name is set correctly on fallbacksextendedReq.Model = baseModelConfig.ModelName@@ -448,9 +450,9 @@ func (stream *ExtendedChatCompletionStream) Close() error {return stream.customReader.Close()}-func resolveReq(req *types.ExtendedChatCompletionRequest, modelConfig *shared.ModelRoleConfig, baseModelConfig *shared.BaseModelConfig) {+func resolveReq(req *types.ExtendedChatCompletionRequest, modelConfig *shared.ModelRoleConfig, baseModelConfig *shared.BaseModelConfig, settings *shared.PlanSettings) {// if system prompt is disabled, change the role of the system message to user- if modelConfig.GetSharedBaseConfig().SystemPromptDisabled {+ if modelConfig.GetSharedBaseConfig(settings).SystemPromptDisabled {log.Println("System prompt disabled - changing role of system message to user")for i, msg := range req.Messages {log.Println("Message role:", msg.Role)@@ -465,7 +467,7 @@ func resolveReq(req *types.ExtendedChatCompletionRequest, modelConfig *shared.Mo}}- if modelConfig.GetSharedBaseConfig().RoleParamsDisabled {+ if modelConfig.GetSharedBaseConfig(settings).RoleParamsDisabled {log.Println("Role params disabled - setting temperature and top p to 0")req.Temperature = 0req.TopP = 0diff --git a/app/server/model/client_stream.go b/app/server/model/client_stream.goindex 605a4144..6bf5467f 100644--- a/app/server/model/client_stream.go+++ b/app/server/model/client_stream.go@@ -20,21 +20,21 @@ func CreateChatCompletionWithInternalStream(clients map[string]ClientInfo,authVars map[string]string,modelConfig *shared.ModelRoleConfig,- localProvider shared.ModelProvider,+ settings *shared.PlanSettings,ctx context.Context,req types.ExtendedChatCompletionRequest,onStream OnStreamFn,reqStarted time.Time,) (*types.ModelResponse, error) {- providerComposite := modelConfig.GetProviderComposite(authVars, localProvider)+ providerComposite := modelConfig.GetProviderComposite(authVars, settings)_, ok := clients[providerComposite]if !ok {return nil, fmt.Errorf("client not found for provider composite: %s", providerComposite)}- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, localProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, settings)- resolveReq(&req, modelConfig, baseModelConfig)+ resolveReq(&req, modelConfig, baseModelConfig, settings)// choose the fastest provider by latency/throughput on openrouterif baseModelConfig.Provider == shared.ModelProviderOpenRouter {@@ -50,14 +50,14 @@ func CreateChatCompletionWithInternalStream(}return withStreamingRetries(ctx, func(numTotalRetry int, didProviderFallback bool, modelErr *shared.ModelError) (resp *types.ModelResponse, fallbackRes shared.FallbackResult, err error) {- fallbackRes = modelConfig.GetFallbackForModelError(numTotalRetry, didProviderFallback, modelErr, authVars, localProvider)+ fallbackRes = modelConfig.GetFallbackForModelError(numTotalRetry, didProviderFallback, modelErr, authVars, settings)resolvedModelConfig := fallbackRes.ModelRoleConfigif resolvedModelConfig == nil {return nil, fallbackRes, fmt.Errorf("model config is nil")}- providerComposite := resolvedModelConfig.GetProviderComposite(authVars, localProvider)+ providerComposite := resolvedModelConfig.GetProviderComposite(authVars, settings)opClient, ok := clients[providerComposite]if !ok {@@ -65,7 +65,7 @@ func CreateChatCompletionWithInternalStream(}modelConfig = resolvedModelConfig- resp, err = processChatCompletionStream(resolvedModelConfig, opClient, authVars, localProvider, ctx, req, onStream, reqStarted)+ resp, err = processChatCompletionStream(resolvedModelConfig, opClient, authVars, settings, ctx, req, onStream, reqStarted)if err != nil {return nil, fallbackRes, err}@@ -82,7 +82,7 @@ func processChatCompletionStream(modelConfig *shared.ModelRoleConfig,client ClientInfo,authVars map[string]string,- localProvider shared.ModelProvider,+ settings *shared.PlanSettings,ctx context.Context,req types.ExtendedChatCompletionRequest,onStream OnStreamFn,@@ -94,7 +94,7 @@ func processChatCompletionStream("model": modelConfig.ModelId,}))- stream, err := createChatCompletionStreamExtended(modelConfig, client, authVars, localProvider, streamCtx, req)+ stream, err := createChatCompletionStreamExtended(modelConfig, client, authVars, settings, streamCtx, req)if err != nil {cancel()diff --git a/app/server/model/model_request.go b/app/server/model/model_request.goindex 95ddfe1d..7ee4f741 100644--- a/app/server/model/model_request.go+++ b/app/server/model/model_request.go@@ -18,13 +18,13 @@ import ()type ModelRequestParams struct {- Clients map[string]ClientInfo- AuthVars map[string]string- Auth *types.ServerAuth- Plan *db.Plan- ModelConfig *shared.ModelRoleConfig- LocalProvider shared.ModelProvider- Purpose string+ Clients map[string]ClientInfo+ AuthVars map[string]string+ Auth *types.ServerAuth+ Plan *db.Plan+ ModelConfig *shared.ModelRoleConfig+ Settings *shared.PlanSettings+ Purpose stringMessages []types.ExtendedChatMessagePrediction string@@ -68,23 +68,23 @@ func ModelRequest(modelPackName := params.ModelPackNamepurpose := params.PurposesessionId := params.SessionId- localProvider := params.LocalProvider+ settings := params.Settingsif purpose == "" {return nil, fmt.Errorf("purpose is required")}- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, localProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, settings)messages = FilterEmptyMessages(messages)messages = CheckSingleSystemMessage(modelConfig, baseModelConfig, messages)inputTokensEstimate := GetMessagesTokenEstimate(messages...) + TokensPerRequest- config := modelConfig.GetRoleForInputTokens(inputTokensEstimate)+ config := modelConfig.GetRoleForInputTokens(inputTokensEstimate, settings)modelConfig = &configif params.EstimatedOutputTokens != 0 {- config = modelConfig.GetRoleForOutputTokens(params.EstimatedOutputTokens)+ config = modelConfig.GetRoleForOutputTokens(params.EstimatedOutputTokens, settings)modelConfig = &config}@@ -166,7 +166,7 @@ func ModelRequest(}}- res, err := CreateChatCompletionWithInternalStream(clients, authVars, modelConfig, localProvider, ctx, req, onStream, reqStarted)+ res, err := CreateChatCompletionWithInternalStream(clients, authVars, modelConfig, settings, ctx, req, onStream, reqStarted)if err != nil {return nil, errdiff --git a/app/server/model/name.go b/app/server/model/name.goindex 227db5b6..4cfb9362 100644--- a/app/server/model/name.go+++ b/app/server/model/name.go@@ -29,7 +29,7 @@ func GenPlanName(var tools []openai.Toolvar toolChoice *openai.ToolChoice- baseModelConfig := config.GetBaseModelConfig(authVars, settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, settings)var sysPrompt stringif baseModelConfig.PreferredOutputFormat == shared.ModelOutputFormatXml {@@ -66,17 +66,17 @@ func GenPlanName(}modelRes, err := ModelRequest(ctx, ModelRequestParams{- Clients: clients,- AuthVars: authVars,- Auth: auth,- Plan: plan,- ModelConfig: &config,- Purpose: "Plan name",- Messages: messages,- Tools: tools,- ToolChoice: toolChoice,- SessionId: sessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Clients: clients,+ AuthVars: authVars,+ Auth: auth,+ Plan: plan,+ ModelConfig: &config,+ Purpose: "Plan name",+ Messages: messages,+ Tools: tools,+ ToolChoice: toolChoice,+ SessionId: sessionId,+ Settings: settings,})if err != nil {@@ -139,7 +139,7 @@ func GenPipedDataName(var tools []openai.Toolvar toolChoice *openai.ToolChoice- baseModelConfig := config.GetBaseModelConfig(authVars, settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, settings)if baseModelConfig.PreferredOutputFormat == shared.ModelOutputFormatXml {sysPrompt = prompts.SysPipedDataNameXml@@ -175,17 +175,17 @@ func GenPipedDataName(}modelRes, err := ModelRequest(ctx, ModelRequestParams{- Clients: clients,- Auth: auth,- AuthVars: authVars,- Plan: plan,- ModelConfig: &config,- Purpose: "Piped data name",- Messages: messages,- Tools: tools,- ToolChoice: toolChoice,- SessionId: sessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Clients: clients,+ Auth: auth,+ AuthVars: authVars,+ Plan: plan,+ ModelConfig: &config,+ Purpose: "Piped data name",+ Messages: messages,+ Tools: tools,+ ToolChoice: toolChoice,+ SessionId: sessionId,+ Settings: settings,})if err != nil {@@ -235,7 +235,7 @@ func GenNoteName(var tools []openai.Toolvar toolChoice *openai.ToolChoice- baseModelConfig := config.GetBaseModelConfig(authVars, settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, settings)if baseModelConfig.PreferredOutputFormat == shared.ModelOutputFormatXml {sysPrompt = prompts.SysNoteNameXml@@ -271,17 +271,17 @@ func GenNoteName(}modelRes, err := ModelRequest(ctx, ModelRequestParams{- Clients: clients,- Auth: auth,- AuthVars: authVars,- Plan: plan,- ModelConfig: &config,- Purpose: "Note name",- Messages: messages,- Tools: tools,- ToolChoice: toolChoice,- SessionId: sessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Clients: clients,+ Auth: auth,+ AuthVars: authVars,+ Plan: plan,+ ModelConfig: &config,+ Purpose: "Note name",+ Messages: messages,+ Tools: tools,+ ToolChoice: toolChoice,+ SessionId: sessionId,+ Settings: settings,})if err != nil {diff --git a/app/server/model/plan/build_validate_and_fix.go b/app/server/model/plan/build_validate_and_fix.goindex f732fa5b..c6cb038c 100644--- a/app/server/model/plan/build_validate_and_fix.go+++ b/app/server/model/plan/build_validate_and_fix.go@@ -199,7 +199,7 @@ func (fileState *activeBuildStreamFileState) buildValidate(syntaxErrors := params.syntaxErrorsreasons := params.reasons- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, fileState.settings.GetModelPack().LocalProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, fileState.settings)// Get diff for validationlog.Printf("Getting diffs between original and updated content")@@ -292,7 +292,7 @@ func (fileState *activeBuildStreamFileState) buildValidate(WillCacheNumTokens: willCacheNumTokens,SessionId: params.sessionId,EstimatedOutputTokens: maxExpectedOutputTokens,- LocalProvider: fileState.settings.GetModelPack().LocalProvider,+ Settings: fileState.settings,})if err != nil {diff --git a/app/server/model/plan/build_whole_file.go b/app/server/model/plan/build_whole_file.goindex bc963cb5..f09f400e 100644--- a/app/server/model/plan/build_whole_file.go+++ b/app/server/model/plan/build_whole_file.go@@ -35,7 +35,7 @@ func (fileState *activeBuildStreamFileState) buildWholeFileFallback(buildCtx conreturn "", fmt.Errorf("active plan not found for plan ID %s and branch %s", planId, branch)}- baseModelConfig := config.GetBaseModelConfig(authVars, fileState.settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, fileState.settings)originalFileWithLineNums := shared.AddLineNums(originalFile)proposedContentWithLineNums := shared.AddLineNums(proposedContent)@@ -57,8 +57,8 @@ func (fileState *activeBuildStreamFileState) buildWholeFileFallback(buildCtx coninputTokens := model.GetMessagesTokenEstimate(messages...) + model.TokensPerRequestmaxExpectedOutputTokens := shared.GetNumTokensEstimate(originalFile + proposedContent)- modelConfig := config.GetRoleForInputTokens(inputTokens)- modelConfig = modelConfig.GetRoleForOutputTokens(maxExpectedOutputTokens)+ modelConfig := config.GetRoleForInputTokens(inputTokens, fileState.settings)+ modelConfig = modelConfig.GetRoleForOutputTokens(maxExpectedOutputTokens, fileState.settings)log.Println("buildWholeFile - calling model for whole file write")@@ -109,8 +109,8 @@ func (fileState *activeBuildStreamFileState) buildWholeFileFallback(buildCtx conWillCacheNumTokens: willCacheNumTokens,EstimatedOutputTokens: maxExpectedOutputTokens,- SessionId: sessionId,- LocalProvider: fileState.settings.GetModelPack().LocalProvider,+ SessionId: sessionId,+ Settings: fileState.settings,})if err != nil {diff --git a/app/server/model/plan/commit_msg.go b/app/server/model/plan/commit_msg.goindex a03064c9..2dac1aca 100644--- a/app/server/model/plan/commit_msg.go+++ b/app/server/model/plan/commit_msg.go@@ -39,7 +39,7 @@ func (state *activeTellStreamState) genPlanDescription() (*db.ConvoMessageDescri}}- baseModelConfig := config.GetBaseModelConfig(authVars, settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, settings)var sysPrompt stringvar tools []openai.Tool@@ -96,7 +96,7 @@ func (state *activeTellStreamState) genPlanDescription() (*db.ConvoMessageDescriModelStreamId: state.modelStreamId,ConvoMessageId: state.replyId,SessionId: activePlan.SessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Settings: settings,}if tools != nil {@@ -232,15 +232,15 @@ func GenCommitMsgForPendingResults(params GenCommitMsgForPendingResultsParams) (}modelRes, err := model.ModelRequest(ctx, model.ModelRequestParams{- Clients: clients,- AuthVars: authVars,- Auth: auth,- Plan: plan,- ModelConfig: &config,- Purpose: "Commit message",- Messages: messages,- SessionId: sessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Clients: clients,+ AuthVars: authVars,+ Auth: auth,+ Plan: plan,+ ModelConfig: &config,+ Purpose: "Commit message",+ Messages: messages,+ SessionId: sessionId,+ Settings: settings,})if err != nil {diff --git a/app/server/model/plan/exec_status.go b/app/server/model/plan/exec_status.goindex 99757a0f..c384e64b 100644--- a/app/server/model/plan/exec_status.go+++ b/app/server/model/plan/exec_status.go@@ -35,7 +35,7 @@ func (state *activeTellStreamState) execStatusShouldContinue(currentMessage striauthVars := state.authVarsconfig := settings.GetModelPack().ExecStatus- baseModelConfig := config.GetBaseModelConfig(authVars, settings.GetModelPack().LocalProvider)+ baseModelConfig := config.GetBaseModelConfig(authVars, settings)currentSubtask := state.currentSubtaskif currentSubtask == nil {@@ -155,7 +155,7 @@ func (state *activeTellStreamState) execStatusShouldContinue(currentMessage striModelStreamId: state.modelStreamId,ConvoMessageId: state.replyId,SessionId: sessionId,- LocalProvider: settings.GetModelPack().LocalProvider,+ Settings: settings,})if err != nil {diff --git a/app/server/model/plan/tell_exec.go b/app/server/model/plan/tell_exec.goindex 96b72147..10f3816c 100644--- a/app/server/model/plan/tell_exec.go+++ b/app/server/model/plan/tell_exec.go@@ -397,21 +397,21 @@ func execTellPlan(params execTellPlanParams) {if state.currentStage.TellStage == shared.TellStagePlanning {if state.currentStage.PlanningPhase == shared.PlanningPhaseContext {log.Println("Tell plan - isContextStage - setting modelConfig to context loader")- modelConfig = state.settings.GetModelPack().GetArchitect().GetRoleForInputTokens(requestTokens)+ modelConfig = state.settings.GetModelPack().GetArchitect().GetRoleForInputTokens(requestTokens, state.settings)log.Println("Tell plan - got modelConfig for context phase")} else if state.currentStage.PlanningPhase == shared.PlanningPhaseTasks {- modelConfig = state.settings.GetModelPack().Planner.GetRoleForInputTokens(requestTokens)+ modelConfig = state.settings.GetModelPack().Planner.GetRoleForInputTokens(requestTokens, state.settings)log.Println("Tell plan - got modelConfig for tasks phase")}} else if state.currentStage.TellStage == shared.TellStageImplementation {- modelConfig = state.settings.GetModelPack().GetCoder().GetRoleForInputTokens(requestTokens)+ modelConfig = state.settings.GetModelPack().GetCoder().GetRoleForInputTokens(requestTokens, state.settings)log.Println("Tell plan - got modelConfig for implementation stage")}// log.Println("Tell plan - modelConfig:", spew.Sdump(modelConfig))state.modelConfig = &modelConfig- baseModelConfig := modelConfig.GetBaseModelConfig(authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(authVars, state.settings)state.baseModelConfig = baseModelConfig@@ -485,11 +485,11 @@ func (state *activeTellStreamState) doTellRequest() {modelConfig := state.modelConfigactive := state.activePlan- fallbackRes := modelConfig.GetFallbackForModelError(state.numErrorRetry, state.didProviderFallback, state.modelErr, authVars, state.settings.GetModelPack().LocalProvider)+ fallbackRes := modelConfig.GetFallbackForModelError(state.numErrorRetry, state.didProviderFallback, state.modelErr, authVars, state.settings)modelConfig = fallbackRes.ModelRoleConfigstop := []string{"<PlandexFinish/>"}- baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings)if fallbackRes.FallbackType == shared.FallbackTypeProvider {state.didProviderFallback = true@@ -553,7 +553,7 @@ func (state *activeTellStreamState) doTellRequest() {state.numErrorRetry, state.numFallbackRetry, baseModelConfig.ModelName)// start the stream- stream, err := model.CreateChatCompletionStream(clients, authVars, modelConfig, state.settings.GetModelPack().LocalProvider, active.ModelStreamCtx, modelReq)+ stream, err := model.CreateChatCompletionStream(clients, authVars, modelConfig, state.settings, active.ModelStreamCtx, modelReq)if err != nil {log.Printf("Error starting reply stream: %v\n", err)go notify.NotifyErr(notify.SeverityError, fmt.Errorf("error starting reply stream: %v", err))diff --git a/app/server/model/plan/tell_load.go b/app/server/model/plan/tell_load.goindex 34fcc8b9..8449f588 100644--- a/app/server/model/plan/tell_load.go+++ b/app/server/model/plan/tell_load.go@@ -69,7 +69,7 @@ func (state *activeTellStreamState) loadTellPlan() error {defer func() {if r := recover(); r != nil {log.Printf("panic in getPlanSettings: %v\n%s", r, debug.Stack())- errCh <- fmt.Errorf("error getting plan settings: %v", r)+ errCh <- fmt.Errorf("panic getting plan settings: %v\n%s", r, debug.Stack())runtime.Goexit() // don't allow outer function to continue and double-send to channel}}()@@ -283,7 +283,7 @@ func (state *activeTellStreamState) loadTellPlan() error {defer func() {if r := recover(); r != nil {log.Printf("panic in getPlanSubtasks: %v\n%s", r, debug.Stack())- errCh <- fmt.Errorf("error getting plan subtasks: %v", r)+ errCh <- fmt.Errorf("error getting plan subtasks: %v\n%s", r, debug.Stack())runtime.Goexit() // don't allow outer function to continue and double-send to channel}}()diff --git a/app/server/model/plan/tell_stream_finish.go b/app/server/model/plan/tell_stream_finish.goindex a7d1a167..94de4870 100644--- a/app/server/model/plan/tell_stream_finish.go+++ b/app/server/model/plan/tell_stream_finish.go@@ -126,7 +126,7 @@ func (state *activeTellStreamState) handleStreamFinished() handleStreamFinishedR}}()- err := summarizeConvo(clients, authVars, settings.GetModelPack().LocalProvider, settings.GetModelPack().PlanSummary, summarizeConvoParams{+ err := summarizeConvo(clients, authVars, settings, summarizeConvoParams{auth: auth,plan: plan,branch: branch,diff --git a/app/server/model/plan/tell_stream_main.go b/app/server/model/plan/tell_stream_main.goindex 078ad58a..e9f0ba32 100644--- a/app/server/model/plan/tell_stream_main.go+++ b/app/server/model/plan/tell_stream_main.go@@ -63,7 +63,7 @@ func (state *activeTellStreamState) listenStream(stream *model.ExtendedChatCompldefer timer.Stop()streamFinished := false- baseModelConfig := state.modelConfig.GetBaseModelConfig(state.authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := state.modelConfig.GetBaseModelConfig(state.authVars, state.settings)modelProvider := baseModelConfig.ProvidermodelName := baseModelConfig.ModelNamediff --git a/app/server/model/plan/tell_stream_processor.go b/app/server/model/plan/tell_stream_processor.goindex cbdc8a8d..2143c2f9 100644--- a/app/server/model/plan/tell_stream_processor.go+++ b/app/server/model/plan/tell_stream_processor.go@@ -58,7 +58,7 @@ func (state *activeTellStreamState) processChunk(choice types.ExtendedChatCompledelta := choice.Deltacontent := delta.Content- baseModelConfig := state.modelConfig.GetBaseModelConfig(state.authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := state.modelConfig.GetBaseModelConfig(state.authVars, state.settings)if baseModelConfig.IncludeReasoning && delta.Reasoning != "" {content = delta.Reasoningdiff --git a/app/server/model/plan/tell_stream_usage.go b/app/server/model/plan/tell_stream_usage.goindex 9bb82b88..ea72feaf 100644--- a/app/server/model/plan/tell_stream_usage.go+++ b/app/server/model/plan/tell_stream_usage.go@@ -27,7 +27,7 @@ func (state *activeTellStreamState) handleUsageChunk(usage *openai.Usage) {sessionId := state.activePlan.SessionIdmodelConfig := state.modelConfig- baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings)go func() {defer func() {@@ -90,7 +90,7 @@ func (state *activeTellStreamState) execHookOnStop(sendStreamErr bool) {}modelConfig := state.modelConfig- baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings.GetModelPack().LocalProvider)+ baseModelConfig := modelConfig.GetBaseModelConfig(state.authVars, state.settings)go func() {defer func() {diff --git a/app/server/model/plan/tell_summary.go b/app/server/model/plan/tell_summary.goindex 6bf0edbf..6855c077 100644--- a/app/server/model/plan/tell_summary.go+++ b/app/server/model/plan/tell_summary.go@@ -243,7 +243,7 @@ type summarizeConvoParams struct {modelPackName string}-func summarizeConvo(clients map[string]model.ClientInfo, authVars map[string]string, localProvider shared.ModelProvider, config shared.ModelRoleConfig, params summarizeConvoParams, ctx context.Context) *shared.ApiError {+func summarizeConvo(clients map[string]model.ClientInfo, authVars map[string]string, settings *shared.PlanSettings, params summarizeConvoParams, ctx context.Context) *shared.ApiError {plan := params.planplanId := plan.Idlog.Printf("summarizeConvo: Called for plan ID %s on branch %s\n", planId, params.branch)@@ -256,6 +256,8 @@ func summarizeConvo(clients map[string]model.ClientInfo, authVars map[string]strcurrentReply := params.currentReplyactive := GetActivePlan(planId, branch)+ config := settings.GetModelPack().PlanSummary+if active == nil {log.Printf("Active plan not found for plan ID %s and branch %s\n", planId, branch)@@ -392,7 +394,7 @@ func summarizeConvo(clients map[string]model.ClientInfo, authVars map[string]str// latestSummaryCh := make(chan *db.ConvoSummary, 1)// active.LatestSummaryCh = latestSummaryCh- summary, apiErr := model.PlanSummary(clients, authVars, localProvider, config, model.PlanSummaryParams{+ summary, apiErr := model.PlanSummary(clients, authVars, settings, config, model.PlanSummaryParams{Conversation: summaryMessages,ConversationNumTokens: numTokens,LatestConvoMessageId: latestMessageId,diff --git a/app/server/model/summarize.go b/app/server/model/summarize.goindex 0a9f53ed..1e8262d5 100644--- a/app/server/model/summarize.go+++ b/app/server/model/summarize.go@@ -28,7 +28,7 @@ type PlanSummaryParams struct {SessionId string}-func PlanSummary(clients map[string]ClientInfo, authVars map[string]string, localProvider shared.ModelProvider, config shared.ModelRoleConfig, params PlanSummaryParams, ctx context.Context) (*db.ConvoSummary, *shared.ApiError) {+func PlanSummary(clients map[string]ClientInfo, authVars map[string]string, settings *shared.PlanSettings, config shared.ModelRoleConfig, params PlanSummaryParams, ctx context.Context) (*db.ConvoSummary, *shared.ApiError) {messages := []types.ExtendedChatMessage{{Role: openai.ChatMessageRoleSystem,@@ -66,7 +66,7 @@ func PlanSummary(clients map[string]ClientInfo, authVars map[string]string, locaModelStreamId: params.ModelStreamId,Messages: messages,SessionId: params.SessionId,- LocalProvider: localProvider,+ Settings: settings,})if err != nil {diff --git a/app/shared/ai_models_credentials.go b/app/shared/ai_models_credentials.goindex b3525444..ce18f3e0 100644--- a/app/shared/ai_models_credentials.go+++ b/app/shared/ai_models_credentials.go@@ -8,11 +8,18 @@ type ModelProviderOption struct {type ModelProviderOptions map[string]ModelProviderOption-func (m ModelRoleConfig) GetModelProviderOptions() ModelProviderOptions {+func (m ModelRoleConfig) GetModelProviderOptions(settings *PlanSettings) ModelProviderOptions {opts := ModelProviderOptions{}- usesProviders, ok := BuiltInModelProvidersByModelId[m.ModelId]- if !ok {+ builtInUsesProviders := BuiltInModelProvidersByModelId[m.ModelId]++ var customUsesProviders []BaseModelUsesProvider+ if settings != nil {+ customUsesProviders = settings.UsesCustomProviderByModelId[m.ModelId]+ }++ usesProviders := append(builtInUsesProviders, customUsesProviders...)+ if len(usesProviders) == 0 {return opts}@@ -38,19 +45,19 @@ func (m ModelRoleConfig) GetModelProviderOptions() ModelProviderOptions {}if m.ErrorFallback != nil {- opts = opts.Condense(m.ErrorFallback.GetModelProviderOptions())+ opts = opts.Condense(m.ErrorFallback.GetModelProviderOptions(settings))}if m.LargeContextFallback != nil {- opts = opts.Condense(m.LargeContextFallback.GetModelProviderOptions())+ opts = opts.Condense(m.LargeContextFallback.GetModelProviderOptions(settings))}if m.LargeOutputFallback != nil {- opts = opts.Condense(m.LargeOutputFallback.GetModelProviderOptions())+ opts = opts.Condense(m.LargeOutputFallback.GetModelProviderOptions(settings))}if m.StrongModel != nil {- opts = opts.Condense(m.StrongModel.GetModelProviderOptions())+ opts = opts.Condense(m.StrongModel.GetModelProviderOptions(settings))}return optsdiff --git a/app/shared/ai_models_custom.go b/app/shared/ai_models_custom.goindex 7a7ada54..b6c5a362 100644--- a/app/shared/ai_models_custom.go+++ b/app/shared/ai_models_custom.go@@ -259,3 +259,47 @@ func (input ModelsInput) ToClientModelsInput() ClientModelsInput {CustomModelPacks: clientModelPacks,}}++func (cp *CustomProvider) ToModelProviderConfigSchema() ModelProviderConfigSchema {+ return ModelProviderConfigSchema{+ Provider: ModelProviderCustom,+ CustomProvider: &cp.Name,+ BaseUrl: cp.BaseUrl,+ HasAWSAuth: cp.HasAWSAuth,+ SkipAuth: cp.SkipAuth,+ ApiKeyEnvVar: cp.ApiKeyEnvVar,+ ExtraAuthVars: cp.ExtraAuthVars,+ }+}++func (input *CustomModel) ToBaseModelConfig(authVars map[string]string, settings *PlanSettings) *BaseModelConfig {+ providers := GetProvidersForAuthVarsWithModelId(authVars, settings, input.ModelId)++ if len(providers) == 0 {+ return nil+ }++ providerSchema := providers[0]+ return input.ToBaseModelConfigForProvider(authVars, settings, &providerSchema)+}++func (input *CustomModel) ToBaseModelConfigForProvider(authVars map[string]string, settings *PlanSettings, providerSchema *ModelProviderConfigSchema) *BaseModelConfig {+ var modelName ModelName+ for _, usesProvider := range input.Providers {+ if usesProvider.Provider == providerSchema.Provider {+ modelName = usesProvider.ModelName+ break+ }+ }++ return &BaseModelConfig{+ ModelTag: ModelTag(input.ModelId),+ ModelId: input.ModelId,+ Publisher: input.Publisher,+ BaseModelShared: input.BaseModelShared,+ BaseModelProviderConfig: BaseModelProviderConfig{+ ModelProviderConfigSchema: *providerSchema,+ ModelName: modelName,+ },+ }+}diff --git a/app/shared/ai_models_data_models.go b/app/shared/ai_models_data_models.goindex fd238a82..b163053b 100644--- a/app/shared/ai_models_data_models.go+++ b/app/shared/ai_models_data_models.go@@ -59,12 +59,6 @@ type BaseModelConfig struct {Publisher ModelPublisher `json:"publisher,omitempty"`BaseModelSharedBaseModelProviderConfig- ApiKeyEnvVar string `json:"apiKeyEnvVar,omitempty"`- PreferredModelOutputFormat ModelOutputFormat `json:"preferredModelOutputFormat,omitempty"`- ReasoningBudgetEnabled bool `json:"reasoningBudgetEnabled,omitempty"`- ReasoningBudget int `json:"reasoningBudget,omitempty"`- SupportsCacheControl bool `json:"supportsCacheControl,omitempty"`- UsesOpenAIResponsesAPI bool `json:"usesOpenAIResponsesAPI,omitempty"`}type BaseModelUsesProvider struct {@@ -125,7 +119,6 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {ModelProviderConfigSchema: providerConfig,ModelName: provider.ModelName,},- ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,},})}@@ -134,7 +127,7 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {BaseVariant *BaseModelConfigVariantBaseId ModelIdBaseDescription string- Overrides BaseModelShared+ CumulativeOverrides BaseModelSharedRequiresVariantOverrides []string}@@ -145,7 +138,7 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {}baseDescription := params.BaseVariant.DescriptionbaseId := params.BaseId- mergedOverrides := params.Overrides+ mergedOverrides := params.CumulativeOverridesavail = append(avail, &AvailableModel{Description: baseDescription,@@ -159,7 +152,6 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {ModelProviderConfigSchema: providerConfig,ModelName: provider.ModelName,},- ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,},})}@@ -193,18 +185,23 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {baseDescription = b.Description}- modelId := ModelId(strings.Join([]string{string(baseId), string(variant.VariantTag)}, "-"))+ var modelId ModelId+ if variant.IsDefaultVariant {+ modelId = baseId+ } else {+ modelId = ModelId(strings.Join([]string{string(baseId), string(variant.VariantTag)}, "-"))+ }description := strings.Join([]string{baseDescription, variant.Description}, " ")- merged := Merge(b.BaseModelShared, variant.Overrides)+ merged := Merge(baseParams.CumulativeOverrides, variant.Overrides)if len(variant.Variants) > 0 {addVariants(variant.Variants, variantParams{BaseVariant: &variant,BaseId: modelId,BaseDescription: description,- Overrides: merged,+ CumulativeOverrides: merged,RequiresVariantOverrides: variant.RequiresVariantOverrides,})}@@ -219,7 +216,6 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {ModelProviderConfigSchema: providerConfig,ModelName: provider.ModelName,},- ApiKeyEnvVar: providerConfig.ApiKeyEnvVar,},})}@@ -228,7 +224,7 @@ func (b *BaseModelConfigSchema) ToAvailableModels() []*AvailableModel {addVariants(b.Variants, variantParams{BaseId: ModelId(string(b.ModelTag)),BaseDescription: b.Description,- Overrides: b.BaseModelShared,+ CumulativeOverrides: b.BaseModelShared,RequiresVariantOverrides: b.RequiresVariantOverrides,})}@@ -539,23 +535,40 @@ func (m ModelRoleConfig) GetModelId() ModelId {return m.ModelId}-func (m ModelRoleConfig) GetBaseModelConfig(authVars map[string]string, localProvider ModelProvider) *BaseModelConfig {+func (m ModelRoleConfig) GetBaseModelConfig(authVars map[string]string, settings *PlanSettings) *BaseModelConfig {+ foundProvider := m.GetFirstProviderForAuthVars(authVars, settings)+ if foundProvider == nil {+ return nil+ }++ return m.GetBaseModelConfigForProvider(authVars, settings, foundProvider)+}++func (m ModelRoleConfig) GetBaseModelConfigForProvider(authVars map[string]string, settings *PlanSettings, providerSchema *ModelProviderConfigSchema) *BaseModelConfig {if m.BaseModelConfig != nil {return m.BaseModelConfig}- foundProvider := m.GetFirstProviderForAuthVars(authVars, localProvider)- if foundProvider == nil {- return m.BaseModelConfig+ availableModel := GetAvailableModel(providerSchema.Provider, m.ModelId)+ if availableModel != nil {+ c := availableModel.BaseModelConfig+ return &c}- c := GetAvailableModel(foundProvider.Provider, m.ModelId).BaseModelConfig+ var customModel *CustomModel+ if settings != nil {+ customModel = settings.CustomModelsById[m.ModelId]+ }+ if customModel != nil {+ c := customModel.ToBaseModelConfigForProvider(authVars, settings, providerSchema)+ return c+ }- return &c+ return nil}-func (m ModelRoleConfig) GetProviderComposite(authVars map[string]string, localProvider ModelProvider) string {- baseModelConfig := m.GetBaseModelConfig(authVars, localProvider)+func (m ModelRoleConfig) GetProviderComposite(authVars map[string]string, settings *PlanSettings) string {+ baseModelConfig := m.GetBaseModelConfig(authVars, settings)if baseModelConfig == nil {return ""@@ -564,26 +577,35 @@ func (m ModelRoleConfig) GetProviderComposite(authVars map[string]string, localPreturn baseModelConfig.ToComposite()}-func (m ModelRoleConfig) GetReservedOutputTokens() int {+func (m ModelRoleConfig) GetReservedOutputTokens(customModelsById map[ModelId]*CustomModel) int {if m.ReservedOutputTokens > 0 {return m.ReservedOutputTokens}- sharedBaseConfig := m.GetSharedBaseConfig()+ sharedBaseConfig := m.GetSharedBaseConfigWithCustomModels(customModelsById)return sharedBaseConfig.ReservedOutputTokens}-func (m ModelRoleConfig) GetSharedBaseConfig() *BaseModelShared {+func (m ModelRoleConfig) GetSharedBaseConfig(settings *PlanSettings) *BaseModelShared {+ return m.GetSharedBaseConfigWithCustomModels(settings.CustomModelsById)+}++func (m ModelRoleConfig) GetSharedBaseConfigWithCustomModels(customModels map[ModelId]*CustomModel) *BaseModelShared {if m.BaseModelConfig != nil {return &m.BaseModelConfig.BaseModelShared}builtInModel := BuiltInBaseModelsById[m.ModelId]- if builtInModel == nil {- return nil+ if builtInModel != nil {+ return &builtInModel.BaseModelShared+ }++ customModel := customModels[m.ModelId]+ if customModel != nil {+ return &customModel.BaseModelShared}- return &builtInModel.BaseModelShared+ return nil}func (m *ModelRoleConfig) Scan(src interface{}) error {@@ -627,12 +649,12 @@ func (p PlannerRoleConfig) Value() (driver.Value, error) {return json.Marshal(p)}-func (p PlannerRoleConfig) GetMaxConvoTokens() int {+func (p PlannerRoleConfig) GetMaxConvoTokens(settings *PlanSettings) int {if p.MaxConvoTokens > 0 {return p.MaxConvoTokens}- return p.ModelRoleConfig.GetSharedBaseConfig().DefaultMaxConvoTokens+ return p.ModelRoleConfig.GetSharedBaseConfig(settings).DefaultMaxConvoTokens}type RoleJSON anydiff --git a/app/shared/ai_models_errors.go b/app/shared/ai_models_errors.goindex a92a2d46..c3be5ad1 100644--- a/app/shared/ai_models_errors.go+++ b/app/shared/ai_models_errors.go@@ -46,7 +46,7 @@ func (m *ModelRoleConfig) GetFallbackForModelError(didProviderFallback bool,modelErr *ModelError,authVars map[string]string,- localProvider ModelProvider,+ settings *PlanSettings,) FallbackResult {if m == nil || modelErr == nil {return FallbackResult{@@ -72,7 +72,7 @@ func (m *ModelRoleConfig) GetFallbackForModelError(} else if !didProviderFallback {log.Println("no error fallback, trying provider fallback")- providerFallback := m.GetProviderFallback(authVars, localProvider)+ providerFallback := m.GetProviderFallback(authVars, settings)log.Println(spew.Sdump(map[string]interface{}{"providerFallback": providerFallback,@@ -97,8 +97,8 @@ func (m *ModelRoleConfig) GetFallbackForModelError(// we just try a single provider fallback if all defined fallbacks are exhausted// if we've got openrouter credentials in the stack, we always use OpenRouter as the fallback since it has its own routing/fallback routing to maximize resilience// otherwise we just use the second provider in the stack-func (m ModelRoleConfig) GetProviderFallback(authVars map[string]string, localProvider ModelProvider) *ModelRoleConfig {- providers := m.GetProvidersForAuthVars(authVars, localProvider)+func (m ModelRoleConfig) GetProviderFallback(authVars map[string]string, settings *PlanSettings) *ModelRoleConfig {+ providers := m.GetProvidersForAuthVars(authVars, settings)if len(providers) < 2 {return nil@@ -121,13 +121,13 @@ func (m ModelRoleConfig) GetProviderFallback(authVars map[string]string, localPravailableModel := GetAvailableModel(provider, m.ModelId)- if availableModel == nil {- log.Printf("no available model found for provider %s and model id %s", provider, m.ModelId)- return nil+ if availableModel != nil {+ c := availableModel.BaseModelConfig+ res.BaseModelConfig = &c+ } else {+ c := m.GetBaseModelConfig(authVars, settings)+ res.BaseModelConfig = c}- c := availableModel.BaseModelConfig- res.BaseModelConfig = &c-return &res}diff --git a/app/shared/ai_models_large_context.go b/app/shared/ai_models_large_context.goindex f44efdea..3c4b8df1 100644--- a/app/shared/ai_models_large_context.go+++ b/app/shared/ai_models_large_context.go@@ -46,11 +46,11 @@ func (m ModelRoleConfig) GetFinalLargeOutputFallback() ModelRoleConfig {// note that if the token number exeeds all the fallback models, it will return the last fallback model-func (m ModelRoleConfig) GetRoleForInputTokens(inputTokens int) ModelRoleConfig {+func (m ModelRoleConfig) GetRoleForInputTokens(inputTokens int, settings *PlanSettings) ModelRoleConfig {var maxTokens intvar paddingPct float64- sharedBaseConfig := m.GetSharedBaseConfig()+ sharedBaseConfig := m.GetSharedBaseConfig(settings)maxTokens = sharedBaseConfig.MaxTokenspaddingPct = sharedBaseConfig.TokenEstimatePaddingPct@@ -75,14 +75,20 @@ func (m ModelRoleConfig) GetRoleForInputTokens(inputTokens int) ModelRoleConfigreturn currentConfig}-func (m ModelRoleConfig) GetRoleForOutputTokens(outputTokens int) ModelRoleConfig {- sharedBaseConfig := m.GetSharedBaseConfig()+func (m ModelRoleConfig) GetRoleForOutputTokens(outputTokens int, settings *PlanSettings) ModelRoleConfig {+ sharedBaseConfig := m.GetSharedBaseConfig(settings)outputTokens = int(float64(outputTokens) * (1 + sharedBaseConfig.TokenEstimatePaddingPct))var currentConfig ModelRoleConfig = m++ customModelsById := map[ModelId]*CustomModel{}+ if settings != nil {+ customModelsById = settings.CustomModelsById+ }+var n int = 0for {- if currentConfig.GetReservedOutputTokens() >= outputTokens {+ if currentConfig.GetReservedOutputTokens(customModelsById) >= outputTokens {return currentConfig}diff --git a/app/shared/ai_models_providers.go b/app/shared/ai_models_providers.goindex d08dc212..d7d9d29b 100644--- a/app/shared/ai_models_providers.go+++ b/app/shared/ai_models_providers.go@@ -211,9 +211,22 @@ func init() {}}-func GetProvidersForAuthVars(authVars map[string]string) []ModelProviderConfigSchema {+func GetProvidersForAuthVars(authVars map[string]string, settings *PlanSettings) []ModelProviderConfigSchema {var foundProviders []ModelProviderConfigSchema++ allProviders := []ModelProviderConfigSchema{}+for _, providerConfig := range BuiltInModelProviderConfigs {+ allProviders = append(allProviders, providerConfig)+ }++ if settings != nil {+ for _, customProvider := range settings.CustomProviders {+ allProviders = append(allProviders, customProvider.ToModelProviderConfigSchema())+ }+ }++ for _, providerConfig := range allProviders {if providerConfig.SkipAuth {foundProviders = append(foundProviders, providerConfig)@@ -247,19 +260,28 @@ func GetProvidersForAuthVars(authVars map[string]string) []ModelProviderConfigScreturn foundProviders}-func (m ModelRoleConfig) GetProvidersForAuthVars(authVars map[string]string, localProvider ModelProvider) []ModelProviderConfigSchema {+func GetProvidersForAuthVarsWithModelId(authVars map[string]string, settings *PlanSettings, modelId ModelId) []ModelProviderConfigSchema {+ var localProvider ModelProvider+ if settings != nil {+ modelPack := settings.GetModelPack()+ if modelPack != nil {+ localProvider = modelPack.LocalProvider+ }+ }++ builtInUsesProviders := BuiltInModelProvidersByModelId[modelId]- // localProvider can come from the model pack *or* the model role config- if localProvider == "" && m.LocalProvider != "" {- localProvider = m.LocalProvider+ var customUsesProviders []BaseModelUsesProvider+ if settings != nil {+ customUsesProviders = settings.UsesCustomProviderByModelId[modelId]}- usesProviders, ok := BuiltInModelProvidersByModelId[m.ModelId]- if !ok {+ usesProviders := append(builtInUsesProviders, customUsesProviders...)+ if len(usesProviders) == 0 {return []ModelProviderConfigSchema{}}- providers := GetProvidersForAuthVars(authVars)+ providers := GetProvidersForAuthVars(authVars, settings)providersByComposite := map[string]ModelProviderConfigSchema{}for _, provider := range providers {providersByComposite[provider.ToComposite()] = provider@@ -286,10 +308,15 @@ func (m ModelRoleConfig) GetProvidersForAuthVars(authVars map[string]string, loc}return res++}++func (m ModelRoleConfig) GetProvidersForAuthVars(authVars map[string]string, settings *PlanSettings) []ModelProviderConfigSchema {+ return GetProvidersForAuthVarsWithModelId(authVars, settings, m.ModelId)}-func (m ModelRoleConfig) GetFirstProviderForAuthVars(authVars map[string]string, localProvider ModelProvider) *ModelProviderConfigSchema {- providers := m.GetProvidersForAuthVars(authVars, localProvider)+func (m ModelRoleConfig) GetFirstProviderForAuthVars(authVars map[string]string, settings *PlanSettings) *ModelProviderConfigSchema {+ providers := m.GetProvidersForAuthVars(authVars, settings)if len(providers) == 0 {return nil}diff --git a/app/shared/plan_model_settings.go b/app/shared/plan_model_settings.goindex 68e8cd88..73456b29 100644--- a/app/shared/plan_model_settings.go+++ b/app/shared/plan_model_settings.go@@ -8,18 +8,32 @@ import ()type PlanSettings struct {- ModelPackName string `json:"modelPackName"`- ModelPack *ModelPack `json:"modelPack"`- CustomModelPacks []*ModelPack `json:"customModelPacks"`- IsCloud bool `json:"isCloud"`- Configured bool `json:"configured"`- UpdatedAt time.Time `json:"updatedAt"`+ ModelPackName string `json:"modelPackName"`+ ModelPack *ModelPack `json:"modelPack"`+ CustomModelPacks []*ModelPack `json:"customModelPacks"`+ CustomModels []*CustomModel `json:"customModels"`+ CustomModelsById map[ModelId]*CustomModel `json:"customModelsById"`+ CustomProviders []*CustomProvider `json:"customProviders"`+ UsesCustomProviderByModelId map[ModelId][]BaseModelUsesProvider `json:"usesCustomProviderByModelId"`+ IsCloud bool `json:"isCloud"`+ Configured bool `json:"configured"`+ UpdatedAt time.Time `json:"updatedAt"`}-func (p *PlanSettings) Configure(customModelPacks []*ModelPack, isCloud bool) {+func (p *PlanSettings) Configure(customModelPacks []*ModelPack, customModels []*CustomModel, customProviders []*CustomProvider, isCloud bool) {p.CustomModelPacks = customModelPacks+ p.CustomModels = customModels+ p.CustomProviders = customProvidersp.IsCloud = isCloudp.Configured = true+ p.CustomModelsById = map[ModelId]*CustomModel{}+ p.UsesCustomProviderByModelId = map[ModelId][]BaseModelUsesProvider{}++ for _, customModel := range customModels {+ p.CustomModelsById[customModel.ModelId] = customModel+ p.UsesCustomProviderByModelId[customModel.ModelId] = customModel.Providers+ }+}func (p PlanSettings) GetModelPack() *ModelPack {@@ -91,56 +105,56 @@ func (ps PlanSettings) GetPlannerMaxTokens() int {modelPack := ps.GetModelPack()planner := modelPack.Plannerfallback := planner.GetFinalLargeContextFallback()- baseConfig := fallback.GetSharedBaseConfig()+ baseConfig := fallback.GetSharedBaseConfig(&ps)return baseConfig.MaxTokens}func (ps PlanSettings) GetPlannerMaxReservedOutputTokens() int {modelPack := ps.GetModelPack()planner := modelPack.Planner- return planner.GetFinalLargeContextFallback().GetReservedOutputTokens()+ return planner.GetFinalLargeContextFallback().GetReservedOutputTokens(ps.CustomModelsById)}func (ps PlanSettings) GetArchitectMaxTokens() int {modelPack := ps.GetModelPack()architect := modelPack.GetArchitect()fallback := architect.GetFinalLargeContextFallback()- return fallback.GetSharedBaseConfig().MaxTokens+ return fallback.GetSharedBaseConfig(&ps).MaxTokens}func (ps PlanSettings) GetArchitectMaxReservedOutputTokens() int {modelPack := ps.GetModelPack()architect := modelPack.GetArchitect()fallback := architect.GetFinalLargeContextFallback()- return fallback.GetReservedOutputTokens()+ return fallback.GetReservedOutputTokens(ps.CustomModelsById)}func (ps PlanSettings) GetCoderMaxTokens() int {modelPack := ps.GetModelPack()coder := modelPack.GetCoder()fallback := coder.GetFinalLargeContextFallback()- return fallback.GetSharedBaseConfig().MaxTokens+ return fallback.GetSharedBaseConfig(&ps).MaxTokens}func (ps PlanSettings) GetCoderMaxReservedOutputTokens() int {modelPack := ps.GetModelPack()coder := modelPack.GetCoder()fallback := coder.GetFinalLargeContextFallback()- return fallback.GetReservedOutputTokens()+ return fallback.GetReservedOutputTokens(ps.CustomModelsById)}func (ps PlanSettings) GetWholeFileBuilderMaxTokens() int {modelPack := ps.GetModelPack()builder := modelPack.GetWholeFileBuilder()fallback := builder.GetFinalLargeContextFallback()- return fallback.GetSharedBaseConfig().MaxTokens+ return fallback.GetSharedBaseConfig(&ps).MaxTokens}func (ps PlanSettings) GetWholeFileBuilderMaxReservedOutputTokens() int {modelPack := ps.GetModelPack()builder := modelPack.GetWholeFileBuilder()fallback := builder.GetFinalLargeOutputFallback()- return fallback.GetReservedOutputTokens()+ return fallback.GetReservedOutputTokens(ps.CustomModelsById)}func (ps PlanSettings) GetPlannerMaxConvoTokens() int {@@ -152,7 +166,7 @@ func (ps PlanSettings) GetPlannerMaxConvoTokens() int {return planner.MaxConvoTokens}- return planner.GetSharedBaseConfig().DefaultMaxConvoTokens+ return planner.GetSharedBaseConfig(&ps).DefaultMaxConvoTokens}func (ps PlanSettings) GetPlannerEffectiveMaxTokens() int {@@ -192,16 +206,16 @@ func (ps PlanSettings) GetModelProviderOptions() ModelProviderOptions {}opts = opts.Condense(- ms.Planner.GetModelProviderOptions(),- ms.Builder.GetModelProviderOptions(),- ms.PlanSummary.GetModelProviderOptions(),- ms.Namer.GetModelProviderOptions(),- ms.CommitMsg.GetModelProviderOptions(),- ms.ExecStatus.GetModelProviderOptions(),+ ms.Planner.GetModelProviderOptions(&ps),+ ms.Builder.GetModelProviderOptions(&ps),+ ms.PlanSummary.GetModelProviderOptions(&ps),+ ms.Namer.GetModelProviderOptions(&ps),+ ms.CommitMsg.GetModelProviderOptions(&ps),+ ms.ExecStatus.GetModelProviderOptions(&ps),// optional roles- getOptionalModelProviderOptions(ms.WholeFileBuilder),- getOptionalModelProviderOptions(ms.Architect),- getOptionalModelProviderOptions(ms.Coder),+ getOptionalModelProviderOptions(&ps, ms.WholeFileBuilder),+ getOptionalModelProviderOptions(&ps, ms.Architect),+ getOptionalModelProviderOptions(&ps, ms.Coder),)return opts@@ -214,6 +228,8 @@ func (ps *PlanSettings) Equals(other *PlanSettings) bool {func (ps PlanSettings) ForCompare() PlanSettings {ps.UpdatedAt = time.Time{}ps.CustomModelPacks = nil+ ps.CustomModels = nil+ ps.CustomProviders = nilps.IsCloud = falseps.Configured = falsereturn ps@@ -232,9 +248,9 @@ func (ps PlanSettings) DeepCopy() (*PlanSettings, error) {return ©, nil}-func getOptionalModelProviderOptions(cfg *ModelRoleConfig) ModelProviderOptions {+func getOptionalModelProviderOptions(settings *PlanSettings, cfg *ModelRoleConfig) ModelProviderOptions {if cfg == nil {return ModelProviderOptions{}}- return cfg.GetModelProviderOptions()+ return cfg.GetModelProviderOptions(settings)}