feat(mcp): add planning and reevaluation (#6541)

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

Author: mudler

core/config/model_config.go

@@ -85,10 +85,13 @@ type MCPConfig struct {
 }
 
 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"`
+	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"`
+	EnablePlanning        bool `yaml:"enable_planning" json:"enable_planning"`
+	EnableMCPPrompts      bool `yaml:"enable_mcp_prompts" json:"enable_mcp_prompts"`
+	EnablePlanReEvaluator bool `yaml:"enable_plan_re_evaluator" json:"enable_plan_re_evaluator"`
 }
 
 func (c *MCPConfig) MCPConfigFromYAML() (MCPGenericConfig[MCPRemoteServers], MCPGenericConfig[MCPSTDIOServers]) {

core/http/endpoints/openai/mcp.go

@@ -93,6 +93,18 @@ func MCPCompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader,
 			cogitoOpts = append(cogitoOpts, cogito.EnableToolReasoner)
 		}
 
+		if config.Agent.EnablePlanning {
+			cogitoOpts = append(cogitoOpts, cogito.EnableAutoPlan)
+		}
+
+		if config.Agent.EnableMCPPrompts {
+			cogitoOpts = append(cogitoOpts, cogito.EnableMCPPrompts)
+		}
+
+		if config.Agent.EnablePlanReEvaluator {
+			cogitoOpts = append(cogitoOpts, cogito.EnableAutoPlanReEvaluator)
+		}
+
 		if config.Agent.EnableReEvaluation {
 			cogitoOpts = append(cogitoOpts, cogito.EnableToolReEvaluator)
 		}

docs/content/docs/features/mcp.md

@@ -28,6 +28,9 @@ The Model Context Protocol is a standard for connecting AI models to external to
 - **🔒 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
+- **📋 Auto-Planning**: Break down complex tasks into manageable steps
+- **🎯 MCP Prompts**: Specialized prompts for better MCP server interaction
+- **🔄 Plan Re-evaluation**: Dynamic plan adjustment based on results
 - **⚙️ Flexible Agent Control**: Customizable execution limits and retry behavior
 
 ## Configuration
@@ -82,6 +85,9 @@ agent:
   max_iterations: 3     # Maximum number of reasoning iterations
   enable_reasoning: true # Enable tool reasoning capabilities
   enable_re_evaluation: false # Enable tool re-evaluation
+  enable_planning: false # Enable auto-planning capabilities
+  enable_mcp_prompts: false # Enable MCP prompts
+  enable_plan_re_evaluator: false # Enable plan re-evaluation
 ```
 
 ### Configuration Options
@@ -106,6 +112,9 @@ Configure agent behavior and tool execution:
 - **`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)
+- **`enable_planning`**: Enable auto-planning capabilities (default: false)
+- **`enable_mcp_prompts`**: Enable MCP prompts (default: false)
+- **`enable_plan_re_evaluator`**: Enable plan re-evaluation (default: false)
 
 ## Usage
 
@@ -171,6 +180,9 @@ agent:
   max_iterations: 5
   enable_reasoning: true
   enable_re_evaluation: true
+  enable_planning: true
+  enable_mcp_prompts: true
+  enable_plan_re_evaluator: true
 ```
 
 ## Agent Configuration Details
@@ -185,10 +197,16 @@ The `agent` section controls how the AI model interacts with MCP tools:
 - **`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.
 
+### Planning Capabilities
+- **`enable_planning`**: When enabled, the agent uses auto-planning to break down complex tasks into manageable steps and execute them systematically. The agent will automatically detect when planning is needed.
+- **`enable_mcp_prompts`**: When enabled, the agent uses specialized prompts exposed by the MCP servers to interact with the exposed tools.
+- **`enable_plan_re_evaluator`**: When enabled, the agent can re-evaluate and adjust its execution plan based on intermediate results.
+
 ### 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`
+- **Simple tasks**: `max_attempts: 2`, `max_iterations: 2`, `enable_reasoning: false`, `enable_planning: false`
+- **Complex tasks**: `max_attempts: 5`, `max_iterations: 5`, `enable_reasoning: true`, `enable_re_evaluation: true`, `enable_planning: true`, `enable_mcp_prompts: true`
+- **Advanced planning**: `max_attempts: 5`, `max_iterations: 5`, `enable_reasoning: true`, `enable_re_evaluation: true`, `enable_planning: true`, `enable_mcp_prompts: true`, `enable_plan_re_evaluator: true`
+- **Development/Debugging**: `max_attempts: 1`, `max_iterations: 1`, `enable_reasoning: true`, `enable_re_evaluation: true`, `enable_planning: true`
 
 ## How It Works
 

go.mod

@@ -33,7 +33,7 @@ require (
 	github.com/mholt/archiver/v3 v3.5.1
 	github.com/microcosm-cc/bluemonday v1.0.27
 	github.com/modelcontextprotocol/go-sdk v1.0.0
-	github.com/mudler/cogito v0.2.0
+	github.com/mudler/cogito v0.3.0
 	github.com/mudler/edgevpn v0.31.0
 	github.com/mudler/go-processmanager v0.0.0-20240820160718-8b802d3ecf82
 	github.com/nikolalohinski/gonja/v2 v2.4.1

go.sum

@@ -515,6 +515,8 @@ github.com/mr-tron/base58 v1.2.0 h1:T/HDJBh4ZCPbU39/+c3rRvE0uKBQlU27+QI8LJ4t64o=
 github.com/mr-tron/base58 v1.2.0/go.mod h1:BinMc/sQntlIE1frQmRFPUoPA1Zkr8VRgBdjWI2mNwc=
 github.com/mudler/cogito v0.2.0 h1:UzowMlP6kiDLnuwQikac9yUOhI6Qe2tW1jZP5gHQvaY=
 github.com/mudler/cogito v0.2.0/go.mod h1:abMwl+CUjCp87IufA2quZdZt0bbLaHHN79o17HbUKxU=
+github.com/mudler/cogito v0.3.0 h1:NbVAO3bLkK5oGSY0xq87jlz8C9OIsLW55s+8Hfzeu9s=
+github.com/mudler/cogito v0.3.0/go.mod h1:abMwl+CUjCp87IufA2quZdZt0bbLaHHN79o17HbUKxU=
 github.com/mudler/edgevpn v0.31.0 h1:CXwxQ2ZygzE7iKGl1J+vq9pL5PvsW2uc3qI/zgpNpp4=
 github.com/mudler/edgevpn v0.31.0/go.mod h1:DKgh9Wu/NM3UbZoPyheMXFvpu1dSLkXrqAOy3oKJN3I=
 github.com/mudler/go-piper v0.0.0-20241023091659-2494246fd9fc h1:RxwneJl1VgvikiX28EkpdAyL4yQVnJMrbquKospjHyA=