Our open-source framework for

Adaptive Intelligence

Adaptive intelligence fits users and use cases

Policy-Driven

Combining rules with content generation and inference optimises outcomes.

Modular

Modular design for flexible implementation

Web3 Native Architecture

Intelligent Agents

Multi-Agent Architecture

Coupling LLMs with knowledge graphs enables new levels of hyper-personalisation and intelligence. Continuous learning enable agents to deliver context-aware interactions.

Secure Messaging

Decentralised Messaging

OOAI's core is a secure, decentralised messaging framework that ensures private communication between users and agents. End-to-End encryption powered by Ockam.

Data Exchange

Connective Intelligence

Real-time data exchange is crucial for hydrating intelligent content. Our data exchange core integrates IAM and verifiable credentials, in a secure zero-trust manner.

Use OOAI for

Instant Digital Payments

Self-Custodial Wallets

Virtually no transaction fees

Personalised Assistants

Privacy-Preserving

Hyperpersonalised assistants

Automated Scheduling

Custom Notifications

Preference and event-based scheduling

Data Sovereignty

Verifiable Credentials

Privacy-Preserving attestations

Event Delegation

Data Routing

Intelligent event and data distribution

Micropayments

Send Payments

Instant, permissionless payments

Social Media

Engage Communities

Real-Time intelligent interaction

What Makes OOAI Unique

OOAI is the world's first (and only) open source AI governance framework.

OOAI is not designed to replace Large Language Model (LLM) or AI service providers. Instead, as a message brokering framework with foundational tools, it aims to empower users with complete control and sovereignty over their data in an era where AI plays an increasingly significant role in our lives.

By prioritising user privacy, trust, and intelligent interactions, OOAI creates a decentralised ecosystem that puts the power of AI in the hands of individuals, addressing the challenges of data centralisation and user empowerment in the digital age.

OOAI will support Solana at launch. Solana was chosen for its low cost, secure and fast transaction settlements. Not all transactions are recorded on-chain. We believe a hybrid approach is required to mitigate the high cost associated with on-chain transactions.

Our roadmap explores the possibility of launching a dedicated network on Substrate for better autonomy. Such a migration would only benefit the community if a significant economy of scale is achieved (e.g. 5 million+ MAU).

Click to find it on github.

Agents and Agency

Given the complexity of digital ecosystems and the multitude of endpoints users interact with daily, it's impractical for individuals to manage their data securely and efficiently. To address this, OOAI has developed a sophisticated suite of agent types that users can dynamically control and customize. These agents are not mere bots, but intelligent entities with foundational flows and advanced capabilities baked in.

Our agent ecosystem is designed to assist users with authentication, authorisation, data filtering, budget/attention allocation, and much more. They interact through collaborative and adversarial interfaces, enabling complex inter-agent communications and negotiations on behalf of users.

OOAI implements several agent types:

Base Agent: Defined in base_agent.rs, it encapsulates core agent functionality including state management, Q-table for action-value estimation, knowledge graph, domain and skill tracking, and provider metadata handling. It supports JSON serialization/deserialization and implements methods for knowledge graph manipulation.
Data Agent: Combines Base Agent, Knowledge Agent, and Q-Learning Agent functionalities. It integrates with DataExchangeStruct for event processing and uses a MqttKafkaDataBridge for message handling. It implements both CollaborativeAgent and AdversarialAgent traits, enabling complex inter-agent interactions.
Knowledge Agent: Extends Base Agent with advanced NLP capabilities. It uses SpacyModule for text processing, implements a sophisticated knowledge graph, and includes sentiment analysis, entity recognition, and topic extraction. It also integrates with PersonalisationGraph and supports RLHF configurations.
Q-Learning Agent: Implements reinforcement learning using the Q-learning algorithm. It features experience replay with a priority queue, eligibility traces for temporal credit assignment, and softmax action selection. It supports dynamic exploration rate adjustment and Q-table persistence using verifiable credentials.

These agents interact through the CollaborativeAgent and AdversarialAgent traits, enabling message exchange, collaboration proposals, knowledge sharing, and competitive interactions. The AgentRouter facilitates inter-agent communication using an AgentGraph for message routing and adapts routing based on accumulated experiences.

The Base Agent is extensible allowing users and developers to continue expanding on the ecosystem with new and innovative Agents.

The key distinction of our agent system is its commitment to user empowerment through complete transparency. By leveraging verifiable credentials, secure data exchange protocols, and our unique OOAI token economy, agents provide users with unprecedented agency in their digital interactions. This approach not only enhances security and efficiency but also gives users greater control and insight into their data management and third-party interactions in an increasingly complex digital landscape.

Distributed vs Decentralised

OOAI's messaging infrastructure incorporates both distributed and decentralised elements, leveraging the strengths of each approach to create a robust, scalable, and secure communication system.

The distributed aspect of OOAI's messaging is evident in its use of multiple technologies and protocols to handle message routing and storage. The MessageRouter struct in route_classifier.rs demonstrates this by integrating Kafka for high-throughput message streaming and MQTT for lightweight, real-time messaging. This distributed approach allows for efficient message classification and routing based on content and metadata, ensuring optimal performance across various network conditions.

On the decentralised front, OOAI implements a consensus layer, as seen in consensus.rs, which utilises a combination of TiKV for distributed key-value storage and local storage for message replication. This decentralised architecture enhances data resilience and availability, crucial for a web3 ecosystem. The ConsensusLayer struct implements methods for message validation, replication, and synchronisation across the network, ensuring consistency and integrity of the messaging system without relying on a central authority.

OOAI implements end-to-end encryption using the Ockam framework, as seen in the message_encryption.rs file. The EncryptedMessageHandler and KeyExchangeHandler manage secure message transmission and key exchange. Additionally, the PIIHandler in pii_handler.rs employs sophisticated techniques for masking and unmasking personally identifiable information (PII) within messages. These measures ensure message confidentiality and secure handling of sensitive data, even in a decentralised environment.

Data Exchange

A Seamless Data Fusion Engine

OOAI's data exchange system is a sophisticated, modular architecture designed to facilitate seamless communication and data transfer across diverse protocols and platforms. At its core, the system leverages a flexible DataBridge trait, allowing for the implementation of various data bridging solutions, such as the MqttKafkaDataBridge, which combines MQTT and Kafka messaging systems.

The DataExchange struct serves as the central component, abstracting the data exchange process and connecting incoming data sources to processing sinks. It utilizes CloudEvents for standardized event management and employs asynchronous stream processing for efficient handling of data flows.

In OOAI, messaging is not just a communication mechanism but an integral part of its data exchange capabilities. The tight integration between messaging and data exchange creates a flexible, secure, and efficient system capable of handling diverse data transfer requirements in the decentralized AI and web3 ecosystem that OOAI aims to serve.

OOAI supports multiple connection types (gRPC, Kafka, and MQTT) through the ConnectionType enum in exchange_interfaces.rs. It incorporates robust error handling with custom error types like DataExchangeError, ensuring graceful failure management across different protocols. The DataExchangeProcessor in exchange_interfaces.rs dynamically routes messages based on provider configurations, allowing for easy addition and modification of data exchange providers.

OOAI implements encrypted communications and distributed data storage, as mentioned in data_bridging.rs. It uses verifiable credentials and a token-based economy, referenced in exchange_graphql.rs, to enhance security and trust. The system integrates with GraphQL through the HasuraQueryRoot in exchange_graphql.rs, providing a flexible API layer that allows clients to query payment providers and initiate data exchanges using a unified interface.

Messaging in OOAI is intricately linked with its data exchange system, forming a cohesive framework for efficient, secure, and flexible communication across the platform. The relationship between messaging and data exchange is evident in several key aspects of the OOAI architecture.

At the core of this relationship is the Data Exchange module, which serves as the backbone for high-performance, asynchronous message processing. This module leverages CloudEvents for standardized event management, ensuring compatibility and seamless integration across diverse systems within the OOAI ecosystem.

The MessageRouter struct in route_classifier.rs plays a crucial role in bridging messaging and data exchange. It utilizes both Kafka and MQTT clients, demonstrating the system's ability to handle different messaging protocols. This flexibility is essential for efficient data exchange, as it allows OOAI to adapt to various communication requirements and network conditions.

The Message struct in message.rs is designed to accommodate a wide range of data types and metadata, making it suitable for diverse data exchange scenarios. Its inclusion of fields like entity_graph and metadata allows for rich, context-aware messaging that can carry complex data structures, facilitating sophisticated data exchange operations.

The AppState struct in app_state.rs serves as a shared resource for managing routing information, which is crucial for both messaging and data exchange. It allows the system to dynamically adapt its routing strategies based on the current state of the network and the nature of the data being exchanged.

The integration of various data exchange providers through the DataExchangeProcessor and the support for multiple connection types (gRPC, Kafka, MQTT) in the data exchange system complement the messaging infrastructure. This allows OOAI to handle a wide range of data exchange scenarios, from real-time streaming to batch processing, all while leveraging the underlying messaging capabilities.

The GraphQL integration, implemented through structures like HasuraQueryRoot, provides a unified interface for initiating data exchanges and querying payment providers. This integration leverages the messaging system to facilitate these operations, demonstrating how messaging underpins complex data exchange processes in OOAI.