feat: add agent options to model config (#6383)

Signed-off-by: Ettore Di Giacinto <[email protected]>

Author: mudler

core/config/model_config.go

@@ -74,14 +74,22 @@ type ModelConfig struct {
 	Options   []string `yaml:"options" json:"options"`
 	Overrides []string `yaml:"overrides" json:"overrides"`
 
-	MCP MCPConfig `yaml:"mcp" json:"mcp"`
+	MCP   MCPConfig   `yaml:"mcp" json:"mcp"`
+	Agent AgentConfig `yaml:"agent" json:"agent"`
 }
 
 type MCPConfig struct {
 	Servers string `yaml:"remote" json:"remote"`
 	Stdio   string `yaml:"stdio" json:"stdio"`
 }
 
+type AgentConfig struct {
+	MaxAttempts        int  `yaml:"max_attempts" json:"max_attempts"`
+	MaxIterations      int  `yaml:"max_iterations" json:"max_iterations"`
+	EnableReasoning    bool `yaml:"enable_reasoning" json:"enable_reasoning"`
+	EnableReEvaluation bool `yaml:"enable_re_evaluation" json:"enable_re_evaluation"`
+}
+
 func (c *MCPConfig) MCPConfigFromYAML() (MCPGenericConfig[MCPRemoteServers], MCPGenericConfig[MCPSTDIOServers]) {
 	var remote MCPGenericConfig[MCPRemoteServers]
 	var stdio MCPGenericConfig[MCPSTDIOServers]

core/http/endpoints/openai/mcp.go

@@ -95,19 +95,35 @@ func MCPCompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader,
 		// we satisfy to just call internally ComputeChoices
 		defaultLLM := cogito.NewOpenAILLM(config.Name, apiKey, "http://127.0.0.1:"+port)
 
-		f, err := cogito.ExecuteTools(
-			defaultLLM, fragment,
+		cogitoOpts := []cogito.Option{
 			cogito.WithStatusCallback(func(s string) {
 				log.Debug().Msgf("[model agent] [model: %s] Status: %s", config.Name, s)
 			}),
 			cogito.WithContext(ctx),
-			// TODO: move these to configs
-			cogito.EnableToolReEvaluator,
-			cogito.WithIterations(3),
-			cogito.WithMaxAttempts(3),
-			cogito.WithTools(
-				cogitoTools...,
-			),
+			cogito.WithTools(cogitoTools...),
+			cogito.WithIterations(3),  // default to 3 iterations
+			cogito.WithMaxAttempts(3), // default to 3 attempts
+		}
+
+		if config.Agent.EnableReasoning {
+			cogitoOpts = append(cogitoOpts, cogito.EnableToolReasoner)
+		}
+
+		if config.Agent.EnableReEvaluation {
+			cogitoOpts = append(cogitoOpts, cogito.EnableToolReEvaluator)
+		}
+
+		if config.Agent.MaxIterations != 0 {
+			cogitoOpts = append(cogitoOpts, cogito.WithIterations(config.Agent.MaxIterations))
+		}
+
+		if config.Agent.MaxAttempts != 0 {
+			cogitoOpts = append(cogitoOpts, cogito.WithMaxAttempts(config.Agent.MaxAttempts))
+		}
+
+		f, err := cogito.ExecuteTools(
+			defaultLLM, fragment,
+			cogitoOpts...,
 		)
 		if err != nil && !errors.Is(err, cogito.ErrNoToolSelected) {
 			return err

docs/content/docs/features/mcp.md

@@ -27,6 +27,8 @@ The Model Context Protocol is a standard for connecting AI models to external to
 - **⚡ Cached Connections**: Efficient tool caching for better performance
 - **🔒 Secure Authentication**: Support for bearer token authentication
 - **🎯 OpenAI Compatible**: Uses the familiar `/mcp/v1/chat/completions` endpoint
+- **🧠 Advanced Reasoning**: Configurable reasoning and re-evaluation capabilities
+- **⚙️ Flexible Agent Control**: Customizable execution limits and retry behavior
 
 ## Configuration
 
@@ -73,6 +75,13 @@ mcp:
         }
       }
     }
+
+# Agent Configuration
+agent:
+  max_attempts: 3        # Maximum number of tool execution attempts
+  max_iterations: 3     # Maximum number of reasoning iterations
+  enable_reasoning: true # Enable tool reasoning capabilities
+  enable_re_evaluation: false # Enable tool re-evaluation
 ```
 
 ### Configuration Options
@@ -90,6 +99,14 @@ Configure local command-based MCP servers:
 - **`args`**: Array of command-line arguments
 - **`env`**: Environment variables (optional)
 
+#### Agent Configuration (`agent`)
+Configure agent behavior and tool execution:
+
+- **`max_attempts`**: Maximum number of tool execution attempts (default: 3)
+- **`max_iterations`**: Maximum number of reasoning iterations (default: 3)
+- **`enable_reasoning`**: Enable tool reasoning capabilities (default: false)
+- **`enable_re_evaluation`**: Enable tool re-evaluation (default: false)
+
 ## Usage
 
 ### API Endpoint
@@ -148,8 +165,31 @@ mcp:
         }
       }
     }
+
+agent:
+  max_attempts: 5
+  max_iterations: 5
+  enable_reasoning: true
+  enable_re_evaluation: true
 ```
 
+## Agent Configuration Details
+
+The `agent` section controls how the AI model interacts with MCP tools:
+
+### Execution Control
+- **`max_attempts`**: Limits how many times a tool can be retried if it fails. Higher values provide more resilience but may increase response time.
+- **`max_iterations`**: Controls the maximum number of reasoning cycles the agent can perform. More iterations allow for complex multi-step problem solving.
+
+### Reasoning Capabilities
+- **`enable_reasoning`**: When enabled, the agent uses advanced reasoning to better understand tool results and plan next steps.
+- **`enable_re_evaluation`**: When enabled, the agent can re-evaluate previous tool results and decisions, allowing for self-correction and improved accuracy.
+
+### Recommended Settings
+- **Simple tasks**: `max_attempts: 2`, `max_iterations: 2`, `enable_reasoning: false`
+- **Complex tasks**: `max_attempts: 5`, `max_iterations: 5`, `enable_reasoning: true`, `enable_re_evaluation: true`
+- **Development/Debugging**: `max_attempts: 1`, `max_iterations: 1`, `enable_reasoning: true`, `enable_re_evaluation: true`
+
 ## How It Works
 
 1. **Tool Discovery**: LocalAI connects to configured MCP servers and discovers available tools