fix: skip GenerationCommonConfig for Ollama (rejects maxOutputTokens)

Ollama's Genkit plugin validates config against its own schema which
does not include maxOutputTokens. Sending GenerationCommonConfig causes
"Additional property maxOutputTokens is not allowed" error.

Add skipCommonCfg flag to genkitProvider, set for Ollama. Other
providers (Anthropic, OpenAI, Gemini) continue to send maxOutputTokens.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: intel352

genkit/adapter.go

@@ -12,11 +12,12 @@ import (
 
 // genkitProvider adapts a Genkit model to provider.Provider.
 type genkitProvider struct {
-	g         *gk.Genkit
-	modelName string // "provider/model" format e.g. "anthropic/claude-sonnet-4-6"
-	name      string
-	authInfo  provider.AuthModeInfo
-	maxTokens int // 0 means use model default
+	g              *gk.Genkit
+	modelName      string // "provider/model" format e.g. "anthropic/claude-sonnet-4-6"
+	name           string
+	authInfo       provider.AuthModeInfo
+	maxTokens      int  // 0 means use model default
+	skipCommonCfg  bool // true for providers that reject GenerationCommonConfig (e.g. Ollama)
 
 	mu           sync.Mutex
 	definedTools map[string]bool // tracks which tool names are registered
@@ -61,11 +62,12 @@ func (p *genkitProvider) resolveToolRefs(tools []provider.ToolDef) []ai.ToolRef
 }
 
 // generationConfig returns a WithConfig option when maxTokens is configured.
+// Returns nil for providers that don't support GenerationCommonConfig (e.g. Ollama).
 func (p *genkitProvider) generationConfig() ai.GenerateOption {
-	if p.maxTokens > 0 {
-		return ai.WithConfig(&ai.GenerationCommonConfig{MaxOutputTokens: p.maxTokens})
+	if p.skipCommonCfg || p.maxTokens <= 0 {
+		return nil
 	}
-	return nil
+	return ai.WithConfig(&ai.GenerationCommonConfig{MaxOutputTokens: p.maxTokens})
 }
 
 // Chat sends a non-streaming request and returns the complete response.

genkit/providers.go

@@ -124,10 +124,11 @@ func NewOllamaProvider(ctx context.Context, model, serverAddress string, maxToke
 	p := &ollamaPlugin.Ollama{ServerAddress: serverAddress, Timeout: 300} // 5 min — model loading can be slow
 	g := initGenkitWithPlugin(ctx, gk.WithPlugins(p))
 	return &genkitProvider{
-		g:         g,
-		modelName: "ollama/" + model,
-		name:      "ollama",
-		maxTokens: maxTokens,
+		g:             g,
+		modelName:     "ollama/" + model,
+		name:          "ollama",
+		maxTokens:     maxTokens,
+		skipCommonCfg: true, // Ollama rejects GenerationCommonConfig (maxOutputTokens)
 		authInfo: provider.AuthModeInfo{
 			Mode:        "none",
 			DisplayName: "Ollama (local)",