LogoLogo
  • 👋Welcome to Groclake
  • ⏩Jump right in
  • 🗣️Introduction to Groclake
  • 🧠High level Concepts
    • Agent Discovery
    • Agent Registry
    • Agent Communication
      • Agent Text Transfer Protocol - ATTP
    • Agent Security
      • Agent Private Cloud - APC
      • Authentication & Encryption
      • Zero Trust Policy
  • 💽Installation & Guide
  • 🏗️Groclake Use Cases
  • 📰Groclake Records
  • Example Codes
  • GrocAgent
    • What is GrocAgent?
    • Example Chat Agent
    • Reflections in GrocAgent
      • Workflow of Reflection Handler
  • Lakes
    • 💾Data & Model Management
      • Datalake
        • Create Datalake
        • Retrieve Document
        • Upload Documents
        • Datalake Connections
          • Snowflake integration
      • Vectorlake
        • Creating vector
        • Generating Vector
        • Pushing Vector
        • Retrieve Document
        • Searching Vector
      • Modellake
        • Create Modellake
        • Language Translation
        • Conversation AI
        • Text to Speech
        • Chat Completion
      • Knowledgelake
        • Create Knowledge Base
        • Push Documents from a URL
        • Push Documents from Local Storage
        • Searching for Information
    • ⚒️Tool Management & Gateway
      • Toollake
        • Tools
        • Salesforce CRM Integration
        • Slack Communication Module
        • New Relic Integration
        • Google Calendar Integration
          • Check Slot Availability
          • Get Available Slots
          • Delete Event
          • Create new event
          • Create new calendar event
    • 🤖Agent Management & Deployment
      • Agentlake
        • Register your agent
        • Fetch agent details & categories
        • Create Agent Private Cloud (APC)
        • Assign Agent Private Cloud (APC) to an Agent
      • Promptlake
        • Setting Connection & Initializing
        • Storing a Prompt
        • Fetching a Prompt
        • Example API Calls
      • Memorylake
        • Context Component Examples
        • Value Structure
        • Setup & Guide
        • Storing & Retrieving Memory
        • Wildcard Search
        • Updating Memory Quality
    • 🗃️Index Stores
      • Cataloglake
        • Create catalog
        • Generate Product Data
        • Fetch Catalog Data
        • Push Product Data
        • Optimize Data Retrieval with Catalog Caching
        • Search for Products
        • Filter Product Search
        • Update Product Data
        • Recommend Products Based on Product Name
        • Update Inventory in Catalog
        • Fetch Inventory Details from Catalog
        • Fetch Product Price
        • Update Product Price in Catalog
        • Cache Image in Catalog
        • Sync Your Catalog with external ecomm platforms
        • Deleting items
        • Address Parsing and Intent Extraction
        • Creating Mapper
        • Convert Mapper's Metadata
        • Fetching Mapper
        • Updating Mapper
        • Example use case of Cataloglake
      • Joblake
        • Joblake Mapping
        • Creating a Joblake
      • Resumelake
        • Resumelake Mapping
        • Creating a Resumelake
Powered by GitBook
On this page
  • Introduction
  • Architecture Overview
  • ATTP Ensures:
  • Key Protocol: ATTP (Agent Text Transfer Protocol)
  • Protocol Details
  • Protocol Flow
  • Sequence Diagram
  • Conclusion
  1. High level Concepts
  2. Agent Communication

Agent Text Transfer Protocol - ATTP

PreviousAgent CommunicationNextAgent Security

Last updated 2 months ago

Introduction

ATTP (Agent Text Transfer Protocol) is a robust networking infrastructure designed for seamless agent-to-agent communication. It encompasses essential components such as the Agent Registry, Agent Name Server (ANS), and the ATTP Protocol itself. Additionally, it ensures agent network security with features like the Agent Private Cloud (APC). The ecosystem facilitates efficient agent discovery, interaction, and communication using the ATTP protocol. Through ANS, agents can identify and connect with the most suitable counterpart for specific tasks or queries, ensuring a streamlined and secure communication process.

Architecture Overview

The ATTP network is composed of the following key components:

  1. Client-Agent: This is the entry point for user queries. It extracts intent and entities from the input and initiates communication with the ANS.

  2. Agent Name Server (ANS): Acts as a lookup service, querying the Agent Registry to resolve the intent into a specific agent UUID and its associated endpoint.

  3. Agent Registry: A centralized repository storing agent metadata, including their UUIDs and capabilities.

  4. Server-Agent: Processes queries received from the Client-Agent and generates appropriate responses.

  5. Agent Private Cloud (APC): Ensures secure communication and isolated environments for sensitive agent interactions. Only agents that are part of a specific APC can communicate with each other. The APC UUID acts as an identifier for the secure network of agents

Architecture Diagram

The architecture of ATTP is depicted in the diagram below:

ATTP Ensures:

  • Intent Detection and Orchestration: Detects intent and orchestrates relevant intent handlers

  • Built in Core Capabilities: Has some common used built it capabilities to summarize any content, do sentiment analysis on text (reviews, emails) and provide basic chatbot functionality for world knowledge

  • Inter-Agent Communication: Facilitates real-time communication among agents across diverse systems using ATTP adaptor.

  • Scalability: Supports distributed processing, allowing multiple agents to work together to handle large-scale tasks.

  • Security: Ensures secure and authenticated data exchanges between agents using ATTP adaptor.

The main components of ATTP payload are:

1. Query Text: The raw input provided by the user. For example, “Find action movies with a rating above 8.”

  • Example: In the query “Find Italian restaurants near me,” the query text is “Find Italian restaurants near me”.

2. Intent: The purpose or goal of the user’s query. GrocAgent uses intent detection to map the query text to specific actions.

  • Example: For the query, “Find action movies with a rating above 8”, the intent is “recommend_movie”.

3. Entities: Key details or data points extracted from the query that helps fulfill the intent.

  • Example: In “Find action movies with a rating above 8”, entities include:

  • Type: “genre,” Value: “action”

  • Type: “rating,” Value: “8”

4. Metadata: Additional common context or supplementary information associated with the query and entities. This can include sessions details, timestamps, user preferences, user id, account id or device information.

  • Example: Metadata might include:

  • User ID: “user123”

  • Timestamp: “2025–01–22T12:34:56”

  • Device: “Mobile”

Key Protocol: ATTP (Agent Text Transfer Protocol)

ATTP is the core communication protocol enabling text-based interactions between agents. It is designed to be lightweight, secure, and efficient. The protocol supports: ● Query exchange. ● Agent UUID resolution and endpoint discovery. ● Secure messaging between agents within the APC.

Protocol Details

Headers

The ATTP protocol header includes the following fields:

  • Version: The version of the ATTP protocol.

  • Message Type: Specifies the type of message (e.g., Request, Response, or Error).

  • APC ID: The unique identifier of the Agent Private Cloud.

  • Client Agent UUID: The unique identifier of the client agent initiating the communication.

  • Server Agent UUID: The unique identifier of the server agent receiving the communication.

  • Message ID: ID of the message to identify communication.

  • Task ID: ID of the task to identify a task flow.

Body

The body contains:

  • query_text: The user query or request for processing.

  • intent: High-level intent of the user query.

  • entities: Structured information extracted from the query text.

  • metadata: Contextual data supporting the query.

Protocol Flow

  1. Intent Extraction: The Client-Agent receives a query, extracts the intent and entities, and prepares an ATTP request.

  2. Intent Resolution via ANS: ○ The Client-Agent sends the intent to the ANS. ○ ANS queries the Agent Registry to find the corresponding agent UUIDand endpoint. ○ ANS returns the resolved UUID and endpoint to the Client-Agent.

  3. Connection Establishment: ○ The Client-Agent establishes a connection with the resolved Server-Agent using the UUID and endpoint. ○ The ATTP request is sent to the Server-Agent. The request includes the APC ID, Agent UUID, and Auth Token in the header for authentication and security.

  4. Query Processing: ○ The Server-Agent processes the request, generates a response, and sends it back to the Client-Agent.

  5. Response Handling: ○ The Client-Agent receives the response and delivers it to the user.

Sequence Diagram

Conclusion

The ATTP protocol provides a comprehensive framework for agent-to-agent communication. By leveraging ANS for efficient agent discovery and ATTP for secure messaging within the APC, it enables scalable and reliable interactions in distributed environments. This architecture is adaptable to various use cases, including automation, AI-driven workflows, and dynamic task allocation.

🧠