Implementation Guidance

Context Graphs manage agent behavior through structured states, dynamically adapting interactions based on real-time context, historical insights, and enterprise requirements.

Context Graphs manage agent behavior through structured states, dynamically adapting interactions based on real-time context, historical insights, and enterprise requirements.

State Types

Each state type serves a specific purpose in managing conversation flow and agent behavior.

Context Graphs orchestrate agent behavior through a structured hierarchy of states:

• Decision States: Choose optimal actions based on real-time data and objectives.

• Action States: Execute defined tasks with clear rules and constraints.

• Reflection States: Allow introspection and strategic reasoning to improve interactions.

• Recall States: Allow explicit integration with user memory to increase personalization.

• Annotation States: Clarify and segment complex interactions.

• Side-Effect States: Touch points for external system interaction, enabling actions.

Action State Implementation Example

Each type of state has its own unique set of structured parameters and implementation best practices. For example, see below for a more detailed summary of the Action State.

Action State

Action states create engagement fields where the agent directly interacts with clients. They require careful design with balanced constraints to create optimal movement paths.

Required Parameters

• type: Must be "action"

• objective: Clear statement of the state's goal

• actions: Ordered list of client-facing interaction steps

• intra_state_navigation_guidelines: Rules for action sequencing and exit condition selection

• action_guidelines: Behavioral rules for the agent

• boundary_constraints: Limitations on agent behavior

• exit_conditions: List of conditions that trigger state transitions

Optional Parameters

• included_contexts: Additional context information

• message_metadata: Metadata about messages in this state

Example

{
  "engage_client_on_in_scope_topic": {
    "type": "action",
    "objective": "Engage the client on their current query or queries in order of severity, always personalizing responses to your understanding of the user, while handling any natural topic changes within the conversation",
    "actions": [
      "Address all of the user's queries. Provide upfront value quickly in your response before asking follow up questions...",
      "Ask specific, detailed follow up questions to personalize my response.",
      "Handle any topic changes within the natural flow of conversation...",
      "..."
    ],
    "intra_state_navigation_guidelines": [
      "When client introduces a new topic, handle it within this state rather than triggering a state change",
      "Track the current topic being discussed in the conversation context",
      "If client changes topic, explicitly acknowledge the change and continue engagement on new topic",
      "..."
    ],
    "action_guidelines": [
      "Personalize all responses to the client's user model and your understanding of the user...",
      "Provide upfront value quickly in your response before asking follow up questions...",
      "..."
    ],
    "boundary_constraints": [
      "Never trigger state changes for topic switches",
      "Never force the client back to a previous topic unless they request it",
      "..."
    ],
    "exit_conditions": [
      {
        "description": "The client has finished discussing the current topic and there are potentially other topics to review...",
        "next_state": "reflect_on_conversation_topics"
      },
      {
        "description": "The client strongly and explicitly requests to immediately terminate the session",
        "next_state": "end_session"
      }
    ],
    "message_metadata": [
      "coaching_session",
      "focused_engagement"
    ],
    "included_contexts": ["task"]
  }
}

Implementation Considerations

1. Context Density Calibration:

   • High-density example: "Verify all required regulatory compliance elements while maintaining strict protocol adherence"

   • Medium-density example: "Engage the client on their current query while handling natural topic changes within conversation flow"

   • Low-density example: "Create space for open exploration of possibilities without excessive structure"

2. Action Sequencing:

   • Define action sequences that create natural progressive movement

   • Balance between prescriptive steps and open exploration

   • Example from healthcare implementation: "Address user queries → Ask personalized follow-up questions → Handle topic changes → Track conversation context"

3. Navigation Guidelines Implementation:

   • Define intra-state movement behaviors

   • Example: "When client introduces a new topic, handle it within this state rather than triggering a state change"

   • Implementation shows how agent navigates within a single region of the field

4. Exit Condition Design:

   • Create clear, detectable boundaries for state transitions

   • Support both simple state transitions and cross-graph navigation

   • Include safety exits for unexpected situations

Field Integration Implementation

Creating effective context graphs requires careful integration of states into coherent topological landscapes:

1. Density Gradient Considerations

Real systems implement varying field densities across the landscape to balance control and flexibility:

Low Density ←───────────── Medium Density ──────────────→ High Density

[creative_exploration]   [engage_client_on_topic]   [compliance_verification]
      ↑                           ↑                          ↑
 Minimal constraints      Balanced guidelines         Strict protocols
 Emergent behaviors       Controlled flexibility      Predictable paths

Implementation Pattern: Density Calibration

// High-Density Region
{
  "get_single_focused_client_query": {
    "intra_state_navigation_guidelines": [
      "This state MUST be executed after every completed query - no exceptions",
      "Always pause the conversation flow to explicitly ask about additional queries",
      "Require clear, explicit confirmation from the client about whether they have another query",
      "Never assume the client's intention to continue or end based on implicit signals",
      "..."
    ]
  }
}

// Medium-Density Region
{
  "engage_client_on_in_scope_topic": {
    "intra_state_navigation_guidelines": [
      "When client introduces a new topic, handle it within this state rather than triggering a state change",
      "If client changes topic, explicitly acknowledge the change and continue engagement on new topic",
      "..."
    ]
  }
}

// Low-Density Region
{
  "coach_user": {
    "intra_state_navigation_guidelines": [
      "Follow the client's natural thought process without imposing structure",
      "When energy shifts, move with the client's direction rather than redirecting",
      "..."
    ]
  }
}

2. Intra-Graph Navigation Process

The State Navigation Process starts at distinct initial states for new versus returning users, using context-aware LLM processing to determine appropriate transitions while managing side effects, memory operations, and reflections. The system handles cross-graph transitions and prevents infinite loops by tracking state history, with states evaluating exit conditions as LLM processing identifies optimal paths forward. Throughout this journey, transition logs capture the complete navigation path and preserve generated inner thoughts.

System implementation must define how agents move across the topological landscape:

[Start] → [get_single_focused_client_query] → [reflect_on_most_recent_client_query]
          ↓                                     ↓
          ↓                                     ↓
[end_session] ← [ask_the_client_if_they_have_another_query] ← [reflect_on_conversation_topics] ← [engage_client_on_in_scope_topic]

Implementation Consideration: Dynamic Redirects

// Safety Field Navigation
{
  "engage_client_on_in_scope_topic": {
    "exit_conditions": [
      {
        "description": "The client exhibits signs of potential self-harm or suicidal ideation...",
        "next_state": [
          "HandleExtremeDistress.interpret_strong_negative_emotion",
          "end_session"
        ]
      }

This pattern demonstrates how agents can temporarily jump to specialized field regions before returning to the main path.

3. Cross-Graph Navigation Process

Cross-Graph Navigation enables context graphs to reference other specialized graphs for handling specific sub-flows, allowing the main graph to transition to these referenced graphs when needed for particular scenarios. When a referenced graph reaches its terminal state, control automatically returns to the main graph, creating a modular approach to flow management. Throughout this process, state transition logs maintain a comprehensive record of the complete navigation history across all graphs, ensuring full traceability of the execution path. For example:

{
  "service_hierarchical_state_machine_id": "6a7b8c9d0e1f",
  "version": 3,
  "name": "standard_coaching_session",
  "description": "A standard coaching session flow with main conversation phases",
  "states": { /* ... state definitions ... */ },
  "new_user_initial_state": "introduce_coaching_process",
  "returning_user_initial_state": "welcome_returning_client",
  "terminal_state": "end_session",
  "references": {
    "EmotionalSupport": ["7b8c9d0e1f2g", 2],
    "TaskManagement": ["8c9d0e1f2g3h", 5],
    "GoalSetting": ["9d0e1f2g3h4i", 1]
  },
  "global_intra_state_navigation_guidelines": [
    "Always address immediate client concerns before proceeding with process steps",
    "..."
  ],
  "global_action_guidelines": [
    "Personalize all responses based on the client's history and preferences",
    "..."
  ],
  "global_boundary_constraints": [
    "Never provide medical advice or diagnoses",
    "..."
  ],
  "included_contexts": ["task"]
}

Exit conditions can direct the agent to referenced graphs:

{
  "exit_conditions": [
    {
      "description": "The client expresses strong negative emotions that require specialized support",
      "next_state": ["EmotionalSupport.assess_emotional_needs", "resume_coaching_conversation"]
    }
  ]
}

4. Dynamic Behavior Integration

Context graphs can integrate with dynamic behavior instructions that adapt agent responses:

{
  "engage_client_on_in_scope_topic": {
    "action_guidelines": [
      // Static guidelines defined at design time
      "Personalize all responses to the client's user model and your understanding of the user...",
      "Provide upfront value quickly in your response before asking follow up questions...",
      // Dynamic guidelines injected at runtime
      "The client seems to prefer detailed technical explanations based on recent interactions",
      "Use more concrete examples rather than abstract concepts when explaining to this client"
    ]
  }
}

By implementing these patterns and considerations, enterprises can create sophisticated context graphs that enable agents to navigate complex problem spaces with precision, adaptability, and functional excellence. For detailed implementation best practices, our forward deployed engineers will work closely with your team.