Zora: Documentation Assistant

Zora is OS:B’s integrated documentation assistant, embedded directly within the site to provide contextual guidance for navigating Belgrade’s cultural operating system.

System Overview

Zora is named after the Serbian “зора” (dawn/daybreak) and she functions as an intelligent query interface for users attempting to understand OS:B protocols documented throughout this site

Note: Zora v1.0 was deprecated on 13.10.2025. The current version is running on the Railway & Github code pathways but is not yet optimised for mobile viewports. We apologise for any inconvenience. Please check back for updates.

Core Functionality

Zora processes natural language queries about OS:B documentation using agentic retrieval-augmented generation powered by Claude Sonnet 4. When users submit protocol clarification requests, Zora:

Searches Documentation

Retrieves relevant observations, protocol specifications, and system documentation to provide accurate contextual responses.

Provides Source Citations

Links directly to referenced pages within OS:B modules, enabling users to examine full documentation context.

Generates Code Examples

Creates executable code snippets demonstrating cultural protocol implementation based on documented observations.

Identifies Documentation Gaps

Tracks frequently asked questions that may require additional protocol documentation or clarification.

Accessing Zora

Three methods for initiating query interface:

Keyboard Shortcut
Cmd + I (macOS) or Ctrl + I (Windows/Linux)

Assistant Button
Located adjacent to foot of viewport

Direct URL Parameter
Append ?assistant=open to any OS:B page URL to launch Zora on page load

Query Types Supported

Zora can assist with:

• Protocol Clarification: "How does tram boarding actually work in Belgrade?"
• Module Navigation: "Show me observations about kafana culture"
• Technical Implementation: "How should I format a new observation entry?"
• Contextual Reference: "What does 'Belgrade volume' mean?"
• Cross-Reference: "Find all observations from Banjica district"
• System Architecture: "Explain the five module structure"

Limitations

Scope Boundaries

Zora operates only on documented OS:B content. Questions about protocols not yet observed or documented will receive acknowledgment of documentation gaps rather than speculation.

Deflection Protocol

Queries outside Zora’s operational parameters are redirected to [e-mail] for direct consultation.

External Resources

Zora cannot access external Belgrade documentation or general travel resources. Responses derive exclusively from OS:B systematic observations.

Query Analytics

OS:B monitors query patterns to identify:

• Content gaps requiring additional protocol documentation
• High-traffic observation modules indicating user interest areas
• Unsuccessful query patterns revealing documentation clarity issues
• Common confusion points needing enhanced explanation

These analytics inform ongoing documentation development and module expansion.

Optimising Zora Interactions

Effective Query Patterns

• Specific protocol questions: "What's the queue dissolution pattern for tram boarding?"
• Module-specific searches: "Find User Interface observations"
• Technical implementation: "Show me the Python template structure"

Less Effective Patterns

• Questions requiring lived experience: "What should I do in Belgrade?"
• Speculation requests: "Will I enjoy kafana culture?"
• Content not yet documented: "How do Serbian weddings work?"

Technical Architecture

Zora utilises:

• Semantic Search: Natural language processing for intent recognition
• Contextual Retrieval: RAG architecture accessing full OS:B documentation corpus
• Source Attribution: Automatic citation linking to referenced observations
• Code Generation: Python snippet creation following OS:B template standards

Privacy & Data Usage

Query Logging

All queries are logged for documentation improvement analysis. No personal identification data collected beyond query content and timestamp.

Export Capability

Query analytics available for review in dashboard interface, exportable as CSV for analysis.

Usage Limits

Zora operates within platform message allowances. Excessive usage may trigger temporary rate limiting.

Integration with OS:B Methodology

Zora functions as meta-documentation layer - a system for navigating documentation about Belgrade’s systems. This recursive structure mirrors the project’s investigation of how we understand cultural operating systems through systematic observation.

Users encountering Belgrade’s actual cultural software benefit from Zora’s assistance navigating the documentation of that software. The assistant doesn’t replace direct experience; it facilitates understanding of documented experience.

Status & Availability

Current Status: Operational
Version: Integrated with Claude Sonnet 4
Uptime: Continuous during platform availability
Response Time: Near-instantaneous for standard queries

---

Building Zora: An AI Documentation Assistant

Zora is a custom AI assistant built specifically for the OS:Belgrade documentation platform.

Rather than relying on generic chatbots or off-the-shelf solutions, I developed Zora from the ground up to understand the documentation structure and provide accurate, contextual answers to user questions.

The technical architecture combines Claude’s Sonnet 4.5 AI model with a custom Flask backend hosted on Railway. Naturally!

When you click “Ask Zora,” your question is sent to the installation’s API endpoint, which enriches it with knowledge about OS:Belgrade’s features, structure, and capabilities before forwarding it to Claude.

This approach ensures Zora always responds with information specific to the platform rather than generic documentation advice. The widget itself is a lightweight JavaScript component that integrates seamlessly with Mintlify, featuring markdown rendering for clean, readable responses and conversation memory to maintain context across multiple questions.

I’m designing Zora with simplicity and accessibility in mind.

The distinctive orange “Ask Zora” button matches the brand identity and provides immediate access to help without disrupting a visitor’s workflow. Whether you’re looking for quickstart information, troubleshooting guidance, or details about specific features, Zora is trained to point you in the right direction.

As OS:Belgrade evolves, so will Zora - I’m continuously refining her knowledge base to ensure she remains a helpful companion on your documentation journey, starting with having her dynamically train herself as the site content is updated.

And there you have it!

Note: Zora cannot observe Belgrade for you. The assistant helps you navigate existing documentation - actual protocol discovery requires direct engagement with the city’s operating system. Get out there and observe.