ARIA: The Cognitive Continuity Layer

ARIA is an advanced cognitive architecture designed to reconstruct and preserve human intent during digital research. Unlike traditional browser history or bookmarking tools, ARIA builds a deterministic knowledge graph that maps the evolution of a user's mental "rabbit holes," ensuring continuity across fragmented browsing sessions.

Executive Summary

The primary objective of ARIA is Continuity Reconstruction. In modern research workflows, users often traverse complex paths across disparate domains. ARIA captures the underlying relationships between these activities, identifying hierarchical structures (Sets and Subsets) and temporal associations. By integrating local and cloud-based Large Language Models (LLMs) with automated headless browsing, ARIA transforms passive history into an active, queryable knowledge base.

System Architecture

The following diagram illustrates the high-level interaction between the ARIA components:

graph TD
    subgraph Client_Environment [Client Environment]
        Extension[Chrome Extension]
        UI[React Research Workspace]
    end

    subgraph ARIA_Core [ARIA Core Backend]
        API[Node.js Express API]
        Extractor[Entity & Intent Extraction]
        GraphOrch[Graph Orchestrator]
        Agent[Playwright Search Agent]
    end

    subgraph Data_Layer [Data Layer]
        DB[(PostgreSQL + Graph Schema)]
    end

    subgraph Intelligence_Layer [Intelligence Layer]
        Groq[Groq / Llama 3.3]
        Local[Ollama / Local Inference]
    end

    Extension -->|Content Extraction| API
    API --> Extractor
    Extractor -->|Analysis| Intelligence_Layer
    Intelligence_Layer -->|Structured Entities| GraphOrch
    GraphOrch --> DB
    GraphOrch -->|Identify Knowledge Gaps| Agent
    Agent -->|Automated Discovery| API
    UI -->|Contextual Query| API

Loading

Research Ingestion Flow

The process of transforming raw browsing activity into structured knowledge follows a rigorous pipeline:

sequenceDiagram
    participant User
    participant Extension
    participant Backend
    participant Graph as Knowledge Graph
    participant ResearchAgent as Playwright Agent

    User->>Extension: Browses Research Topic
    Extension->>Backend: Transmits Sanitized Content
    Backend->>Backend: Entity Extraction & Intent Analysis
    Backend->>Graph: Map to Current Investigation (Subset/Superset)
    Note right of Backend: Evaluate Knowledge Gaps
    Backend->>ResearchAgent: Trigger Deep Search (Optional)
    ResearchAgent->>Backend: Return Grounded Sources
    Backend->>Graph: Ingest External Sources
    User->>Backend: Query Investigation Context
    Backend->>Graph: Retrieve Relevant Subsets & Evidence
    Backend->>User: Provide Grounded Continuity Report

Loading

Core Capabilities

Deterministic Knowledge Graph

ARIA models research as a series of interconnected nodes and edges. It automatically categorizes pages into "Investigations" based on entity overlap and temporal proximity. This allows for complex structural understanding, such as recognizing that "Mercer F1 Technical Specs" is a subset of "Formula 1 Research."

Temporal Train-of-Thought Preservation

By enforcing strict temporal association windows, ARIA preserves the logical sequence of a user's research. Concepts browsed in quick succession are linked within the graph, allowing the system to reconstruct the mental leaps made during a session.

Agentic Gap-Filling

When the system detects a gap in the current knowledge graph, it can autonomously deploy a headless browser (Playwright) to perform targeted searches. These results are not merely presented to the user but are ingested back into the permanent memory, improving the fidelity of future queries.

Multi-Model Routing Architecture

ARIA utilizes a tiered intelligence strategy to balance performance and reasoning depth:

• Cloud Providers (Groq/OpenRouter): High-reasoning tasks such as final synthesis and complex relationship mapping.
• Local Inference (Ollama): Privacy-sensitive entity extraction and preliminary categorization.

Deployment and Configuration

Prerequisites

• Docker and Docker Compose
• API access for Groq or compatible LLM providers

Environment Setup

A .env configuration is required in the backend directory with the following parameters:

• GROQ_API_KEY: Authentication for cloud inference.
• DATABASE_URL: Connection string for the PostgreSQL instance.
• PORT: Service listener port (default: 3000).

Execution

The system is fully containerized for simplified deployment:

docker-compose up -d --build

Client Integration

The Chrome Extension must be loaded in developer mode from the build directory. Once active, it automatically synchronizes with the local backend to begin cognitive ingestion.

Privacy and Data Sovereignty

ARIA is built on the principle of local-first data ownership. All browsing data, extracted entities, and the resulting knowledge graph are stored within the user's private Docker environment. This architecture ensures that sensitive research activity remains entirely under the user's control.