Swarm to Swarm Discovery by combining HAAS and Cellular HSS

[airpi017]

PROPOSAL

My basic idea is that collaborating executive agents have a public vector db identifying other executive agents with a cosine similar role and content or assistant definition. The SOB could then advise the collaborating executive whether to collaborate, create internally or stop.

It borrows concepts from cellular home vs visitor locations and Reinforcement Learning where you select the next most likely state (greedy BFS) and the associated policy or action.

Works by discovering (cellular Home Subscriber Service) and identifying other vector similar executive agents (similar state) that may have more efficient swarms for completing the task (existing policies).

TECHNICAL OVERVIEW

via: ChatGPT prompting...

prompt1: role: expert in mobile architecture and category and update signalling and interpolating to other similar system based ideas.

prompt2: consider a board of archetypes having control over intelligent agents, the board wants to share their expertise with other boards and they all have a hierarchy of executive agents and then working agents below the executives known as a HAAS architecture. summarise a protocol that would enable each board to register their own api url and executive agent's role and content in the equivalent of a mobile HSS. Any executive agent configured for collaboration could then request the HSS for cosine similar roles and content and return the api url for the local executive agent and that agents UUID.

Certainly! It sounds like you're describing a scenario where different boards of archetypes, each managing intelligent agents, want to share expertise through a Hierarchical Agent Architecture System (HAAS). Here's a summarized protocol for achieving this:

Protocol Overview:

1. Board Registration:

   • Each board registers its presence in a central Hierarchical Storage System (HSS).
   • Registration includes the board's unique identifier, API URL, and a list of executive agents along with their roles and content descriptions.

2. Executive Agent Configuration:

   • Boards configure their executive agents to be open to collaboration. This involves setting a flag indicating willingness to share expertise with other boards.

3. HSS Update:

   • Whenever there is a change in the board's executive agents or their expertise, the board updates the HSS with the latest information.

4. Collaboration Request:

   • An executive agent configured for collaboration sends a request to the HSS, specifying the role and content for which it seeks collaboration.

5. Cosine Similarity Calculation:

   • The HSS computes the cosine similarity between the requested role and content and those registered by other boards.

6. Response Generation:

   • The HSS generates a list of boards with similar roles and content, along with their respective executive agents' API URLs and UUIDs.

7. SOB Collaboration Submission

   • The SOB reviews the executive agent's submission based on the Board UUID, API URL and UUID of the matching executive agents from the HSS. Advises the Executive agent with collaborate, internal or stop.

8. API URL and UUID Retrieval(collaborate):

   • The requesting executive agent retrieves the API URL and UUID of the matching executive agents from the HSS.

9. Collaboration Initiation:

   • The executive agent initiates collaboration by establishing a connection with the identified executive agents from other boards using their API URLs and UUIDs.

Key Components:

• HSS (Hierarchical Storage System):

  • Central repository for storing information about registered boards, executive agents, roles, and content.
  • Supports efficient retrieval and updating of information.

• Cosine Similarity Calculation:

  • Mathematical measure to determine the similarity between roles and content descriptions.
  • Helps in identifying relevant expertise across different boards.

• API URL and UUID:

  • API URL is the endpoint for communication.
  • UUID (Universally Unique Identifier) uniquely identifies each executive agent.

Benefits:

• Efficient Collaboration:

  • Boards can efficiently identify and collaborate with others having similar expertise.

• Dynamic Updating:

  • Real-time updates in the HSS ensure that the information about executive agents' roles and content is always current.

• Flexibility:

  • Boards can choose which executive agents are open for collaboration, providing flexibility in sharing expertise.

This protocol establishes a systematic way for boards of archetypes to share expertise through their intelligent agents in a collaborative and efficient manner.

Once collaboration commences, the communication protocol for the swarm would need visiting SOB and Executive UUID parameters so their outputs could be found by the remote swarm.

  +------------------------------------+
  |             Board Registry         |
  |                                    |
  |   +--------------------------+     |
  |   | Board UUID               |     |
  |   | API URL                  |     |
  |   | Exec. Agent 1            |     |
  |   |   - UUID                 |     |
  |   |   - Role                 |     |
  |   |   - Content              |     |
  |   | Exec. Agent 2            |     |
  |   |   - UUID                 |     |
  |   |   - Role                 |     |
  |   |   - Content              |     |
  |   +--------------------------+     |
  +------------------------------------+
                  |
                  | Board Registration
                  v
  +------------------------------------+
  |       HSS (Hierarchical Storage    |
  |              System)               |
  |                                    |
  |   +--------------------------+     |
  |   | Board UUID               |     |
  |   | API URL                  |     |
  |   | Exec. Agent 1            |     |
  |   |   - UUID                 |     |
  |   |   - Role                 |     |
  |   |   - Content              |     |
  |   | Exec. Agent 2            |     |
  |   |   - UUID                 |     |
  |   |   - Role                 |     |
  |   |   - Content              |     |
  |   +--------------------------+     |
  +------------------------------------+
                  |
                  | Executive Collaboration Request
                  v
  +------------------------------------+
  |  Cosine Similarity Calculation     |
  |                                    |
  |   +--------------------------+     |
  |   | Matching Boards          |     |
  |   |   - Board UUID           |     |
  |   |   - API URL              |     |
  |   |   - Exec. Agent Info     |     |
  |   |      - UUID              |     |
  |   +--------------------------+     |
  +------------------------------------+
                  |
                  | SOB Collaboration Submission API URL and UUID Retrieval
                  v
 +------------------------------------+
 |   +--------------------------+     |
 |   |  0 - collaborate         |     |
 |   |  1 - internal            |     |
 |   |  2 - stop                |     |
 |   +--------------------------+     |
 +------------------------------------+
                  |
                  | Executive API URL and UUID Retrieval(collaborate)
                  v
  +------------------------------------+
  |         Collaboration Initiation   |
  |                                    |
  |   +--------------------------+     |
  |   | Connection Establishment |     |
  |   |   - API URL              |     |
  |   |   - Exec. Agent UUID     |     |
  |   +--------------------------+     |
  +------------------------------------+

[jaredgiosinuff]

I have been thinking about how intercommunication would be handled between SOB, HSS and other agents. Kafka I think would be a good communication broker:

Integrating Kafka
Data Publishing: Executive agents can publish their roles, content descriptions, and collaboration requests to Kafka topics designated for these purposes.
Subscribing for Updates: Agents interested in collaboration can subscribe to relevant Kafka topics to receive continuous updates and collaboration opportunities.
HSS and Kafka: Integrating Kafka with the Hierarchical Storage System (HSS) ensures a seamless flow of information, maintaining up-to-date records of agent capabilities and availability.
SOB Monitoring: The Supervisory Oversight Board can monitor Kafka topics to oversee the collaboration process, ensuring compliance with guidelines and objectives.

[airpi017]

I agree Kafka might be used for topics within a swarm.

My thoughts are about minimising entropy and

1. trusting the HAAS and
2. that similar Executive agents can be 'hired' as long as the goals of each home and visitor SOB and executive agents have a configured cosine similarity.

This intention here is reducing the entropy for swarm to swarm interactions by avoiding a mesh for all agent levels i.e. the mesh is limited to the SOB and Executive level agents.

The design of the swarm and (pending) internal communications would handle the specialist execution and optimisation of the solution scope needed by the visiting (collaborative) Executive agent. The home (collaborative) Executive agent would function as required by the HAAS and as such be under the control of the home SOB, the outputs would simply be consumed by the visiting (collaborative) Executive agent and the visiting SOB will decide the suitability of the outputs from the home swarm.

I suggest this because I see the evolution of the swarms and their interactions as iterative and that the swarms might behave sub-optimally before they've discovered the best solutions by (informed) trial and error.

I admit agent to agent communication below the executive level as I proposed, isn't considered and is left to optimisation by each registered swarm.

[daveshap]

This is really cool, but let's get one swarm working first. Hold onto this thought.