fix: use MaxTokens not ModelContextLength for output limit

internal/plugins/ai/gemini/gemini.go

@@ -214,10 +214,14 @@ func parseThinkingConfig(level domain.ThinkingLevel) (*genai.ThinkingConfig, boo
 func (o *Client) buildGenerateContentConfig(opts *domain.ChatOptions) (*genai.GenerateContentConfig, error) {
 	temperature := float32(opts.Temperature)
 	topP := float32(opts.TopP)
+	var maxTokens int32
+	if opts.MaxTokens > 0 {
+		maxTokens = int32(opts.MaxTokens)
+	}
 	cfg := &genai.GenerateContentConfig{
 		Temperature:     &temperature,
 		TopP:            &topP,
-		MaxOutputTokens: int32(opts.ModelContextLength),
+		MaxOutputTokens: maxTokens,
 	}
 
 	if opts.Search {

internal/plugins/ai/vertexai/vertexai.go

@@ -127,14 +127,11 @@ func (c *Client) Send(ctx context.Context, msgs []*chat.ChatCompletionMessage, o
 	return c.sendClaude(ctx, msgs, opts)
 }
 
-// getMaxTokens returns the max tokens to use for a request
+// getMaxTokens returns the max output tokens to use for a request
 func getMaxTokens(opts *domain.ChatOptions) int64 {
 	if opts.MaxTokens > 0 {
 		return int64(opts.MaxTokens)
 	}
-	if opts.ModelContextLength > 0 {
-		return int64(opts.ModelContextLength)
-	}
 	return int64(defaultMaxTokens)
 }
 

internal/plugins/ai/vertexai/vertexai_test.go

@@ -310,24 +310,19 @@ func TestGetMaxTokens(t *testing.T) {
 		expected int64
 	}{
 		{
-			name:     "MaxTokens takes priority",
-			opts:     &domain.ChatOptions{MaxTokens: 8192, ModelContextLength: 4096},
+			name:     "MaxTokens specified",
+			opts:     &domain.ChatOptions{MaxTokens: 8192},
 			expected: 8192,
 		},
 		{
-			name:     "ModelContextLength used when MaxTokens is 0",
-			opts:     &domain.ChatOptions{MaxTokens: 0, ModelContextLength: 16384},
-			expected: 16384,
-		},
-		{
-			name:     "Default used when both are 0",
-			opts:     &domain.ChatOptions{MaxTokens: 0, ModelContextLength: 0},
+			name:     "Default when MaxTokens is 0",
+			opts:     &domain.ChatOptions{MaxTokens: 0},
 			expected: int64(defaultMaxTokens),
 		},
 		{
-			name:     "MaxTokens negative treated as 0",
-			opts:     &domain.ChatOptions{MaxTokens: -1, ModelContextLength: 2048},
-			expected: 2048,
+			name:     "Default when MaxTokens is negative",
+			opts:     &domain.ChatOptions{MaxTokens: -1},
+			expected: int64(defaultMaxTokens),
 		},
 	}
 
@@ -406,9 +401,9 @@ func TestBuildGeminiConfig(t *testing.T) {
 
 	t.Run("basic config with temperature and TopP", func(t *testing.T) {
 		opts := &domain.ChatOptions{
-			Temperature:        0.7,
-			TopP:               0.9,
-			ModelContextLength: 8192,
+			Temperature: 0.7,
+			TopP:        0.9,
+			MaxTokens:   8192,
 		}
 		config := client.buildGeminiConfig(opts)