Round RobinInstall

Overview

The Round Robin node is designed to manage and orchestrate multi-participant conversations, especially for LLM (Large Language Model) workflows such as AI chatbots, role-based simulations, or collaborative dialogue systems. It provides a robust mechanism for storing, retrieving, running, and managing conversation turns between multiple roles (e.g., User, Assistant, System), with support for round tracking, memory recall, and flexible output formatting.

Common scenarios:

• Building conversational loops where each participant (role) takes turns in a structured manner.
• Managing persistent conversation state across workflow executions.
• Preparing conversation history in formats compatible with various LLM providers (OpenAI, Anthropic, Google Gemini, etc.).
• Supporting advanced use cases like multi-agent simulations, customer support bots, or interview roleplays.

Practical examples:

• Simulating a ChatGPT-like experience with User, Assistant, and System roles.
• Running a meeting simulation with multiple participants, each contributing messages in turn.
• Storing and recalling conversation memory for debugging or analysis.

Properties

Below are the main input properties relevant to the "Default" Resource and Operation:

Output

The structure of the json output field depends on the selected operation and output format:

1. Setup Conversation

• Outputs an object for each participant (if roles are defined), or a generic object if not.
• Example:

{
  "status": "success",
  "message": "Conversation initialized for Assistant",
  "operation": "setup",
  "storageId": "my-support-chat",
  "conversation_id": "my-support-chat",
  "RR_title": "Support Chat",
  "conversation_context": "...",
  "talking_points": ["..."],
  "timestamp": "2024-06-01T12:00:00Z",
  "number_of_participants": 3,
  "total_rounds": 5,
  "roleCount": 3,
  "currentRole": "Assistant",
  "currentRoleIndex": 1,
  "roleInfo": { /* role details */ },
  "binaryProperty": "data",
  "isCurrentTurn": false,
  "participant_id": ""
}

2. Store Message

• Outputs updated conversation state, including round count, max rounds, message count, and last updated timestamp.
• Example:

{
  "roundCount": 2,
  "maxRounds": 5,
  "roundsRemaining": 3,
  "messageCount": 6,
  "lastUpdated": "2024-06-01T12:05:00Z",
  "binaryProperty": "data"
}

• If the round limit is reached:

{
  "status": "limit_reached",
  "message": "Maximum conversation rounds (5) reached. Consider clearing the conversation to start over.",
  "roundCount": 5,
  "maxRounds": 5,
  "messageCount": 15
}

3. Run Conversation

• Outputs the result of processing a turn, indicating whether the round or turn is complete, and information about the next role.
• Example:

{
  "status": "turn_complete",
  "operation": "run",
  "title": "Support Chat",
  "conversationId": "my-support-chat",
  "currentRoundCount": 2,
  "maxRounds": 5,
  "nextRole": "User",
  "nextRoleIndex": 0,
  "messageCount": 7,
  "isComplete": false,
  "isRoundComplete": false,
  "lastUpdated": "2024-06-01T12:10:00Z"
}

• If the maximum rounds are reached:

{
  "status": "complete",
  "message": "Conversation has reached the maximum number of rounds (5).",
  "title": "Support Chat",
  "conversationId": "my-support-chat",
  "currentRoundCount": 5,
  "maxRounds": 5,
  "isComplete": true
}

4. Retrieve Conversation

• Output varies based on the selected format:
  • Conversation History for LLM: Array of {role, content} objects, optionally formatted for OpenAI, Anthropic, Google, or generic LLMs.
  • Message Array: Simple array of all messages with role and content.
  • Grouped by Role: Object grouping messages by role.
• Common fields:
  • messageCount, lastUpdated, roundCount, maxRounds, roundsRemaining, binaryProperty.
• Example (OpenAI format):

{
  "conversationHistory": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello!"},
    {"role": "assistant", "content": "Hi, how can I help you?"}
  ],
  "messageCount": 3,
  "roundCount": 1,
  "maxRounds": 5,
  "roundsRemaining": 4,
  "binaryProperty": "data"
}

5. Recall Memory

• Returns raw memory state, optionally including messages, roles, and metadata.
• Example:

{
  "status": "success",
  "operation": "recall",
  "timestamp": "2024-06-01T12:20:00Z",
  "memoryId": "my-support-chat",
  "conversation_id": "my-support-chat",
  "memory": {
    "messages": [ ... ],
    "roles": [ ... ],
    "spotCount": 3,
    "roundCount": 2,
    "maxRounds": 5,
    "lastUpdated": "2024-06-01T12:19:00Z",
    "title": "Support Chat",
    "context": "...",
    "talkingPoints": [ ... ],
    "isNewConversation": false
  }
}

6. Clear Conversation

• Confirms that the conversation was cleared.
• Example:

{
  "success": true,
  "operation": "clear",
  "timestamp": "2024-06-01T12:30:00Z",
  "roundCount": 0,
  "maxRounds": 0,
  "messageCount": 0,
  "status": "success",
  "message": "Conversation cleared successfully",
  "binaryProperty": "data"
}

Binary Data

• The node uses n8n's binary data storage to persist conversation state. The binary property name is configurable and must be consistent across related nodes.

Dependencies

• External Storage: Uses internal n8n binary data storage via a helper (createStorageManager). No external API keys required.
• n8n Configuration: Ensure binary data is passed correctly between nodes using the specified binary property name (default: "data").
• Environment Variables: None required.

Troubleshooting

Common Issues:

• Missing Binary Data:
  • Error: "No conversation data found. Please connect this to a setup node first."
    Solution: Ensure a "Setup Conversation" or "Store Message" operation runs before "Run Conversation" or "Retrieve Conversation".
• Invalid Role Index:
  • Error: "Invalid role index: X. Must be between 0 and N."
    Solution: Check that the "Current Role Index" or "Current Participant" is within the valid range.
• Corrupted or Missing Binary Property:
  • Error: "Binary property 'data' not found. Check that your workflow correctly passes binary data."
    Solution: Verify that the correct binary property name is used and that upstream nodes output binary data under that property.
• Max Rounds Reached:
  • Status: "limit_reached" or "complete"
    Solution: Either clear the conversation or increase the "Maximum Conversation Rounds".

Other Notes:

• If roles are not defined or detected, the node falls back to default roles (User, Assistant, System).
• When using custom roles or input roles, ensure the input data structure matches expectations.

Links and References

• n8n Documentation
• OpenAI Chat Format
• Anthropic Claude API
• Google Gemini API