Skip to content

A-Proxy Documentation

A-Proxy is a persona-driven web archiving system that enables targeted web crawling from specific user perspectives while preserving LLM interactions for research.

Core Goals

A-Proxy serves two primary archival objectives:

  1. Persona-Driven Web Crawling: Create detailed persona representations to enable targeted web crawling from specific demographic, behavioral, and contextual perspectives.

  2. LLM Evolution Documentation: Archive chat interactions to study how large language models respond to different persona contexts over time.

How It Works

Workflow

  1. Persona Creation: Archivists create personas with four attribute categories (demographic, psychographic, behavioral, contextual) through the web interface.

  2. Chat-Assisted Development: Optional conversations with Claude AI help explore persona characteristics. Archivists then manually update persona profiles based on insights from the conversation.

  3. Targeted Browsing: Browse websites with browser variables configured to reflect persona characteristics (location via VPN, language, user-agent).

  4. Archiving: Capture web pages as they appear to the configured persona, preserving both content and context.

  5. Research Value: Preserved archives document personalized web experiences; chat logs document LLM interactions with different persona contexts.

Key Features

Feature Description
Persona Management Create and manage user personas with 4-dimensional attributes
Journey Tracking Record sequences of web interactions as "journeys" with "waypoints"
Web Archiving Archive webpages with persona-specific context (language, geolocation)
Claude AI Chat Chat with Claude AI as or about a persona
VPN Integration Browse from different geographic regions using NordVPN

Current Capabilities

Working:

  • Full CRUD for personas with all 4 dimensions
  • Journey and waypoint management
  • Claude AI chat (requires API key)
  • VPN integration (requires NordVPN setup)
  • Web page archiving with screenshots
  • Internet Archive submission

Manual processes:

  • Persona attribute updates after chat sessions
  • Archive comparison (no built-in diff tool)

Use Cases

  • Targeted web archiving from diverse demographic perspectives
  • Documenting how different user groups experience websites
  • Preserving persona-LLM interactions for future research
  • Creating archival records that reflect multiple viewpoints

Getting Started