Zorora

Local-deployment deep research engine with credibility scoring and citation graphs.

Overview

Zorora is a deep research engine that searches across academic databases, web sources, and newsroom articles, then synthesizes findings with credibility scoring and citation graphs. Built for macOS (Apple Silicon) with minimal RAM footprint, it runs directly from your computer with all content stored locally for complete privacy and control.

Key Features

Deep Research Pipeline

A 6-phase research pipeline that goes beyond simple search.

1. Parallel Source Aggregation - Searches academic (7 sources), web (Brave + DDG), and newsroom simultaneously
2. Citation Following - Multi-hop exploration of cited papers (configurable depth: 1-3)
3. Cross-Referencing - Groups claims by similarity and counts agreement across sources
4. Credibility Scoring - Rules-based scoring of source authority
5. Citation Graph Building - Visualizes relationships between sources
6. Synthesis - Generates comprehensive answers with confidence levels and citations

Local-First Architecture

All processing and storage happens on your machine.

• SQLite Database - Fast indexed queries (~/.zorora/zorora.db)
• JSON Storage - Full research findings (~/.zorora/research/findings/)
• Zero Cloud Dependencies - Core functionality works offline
• Complete Privacy - Research data never leaves your machine

Credibility Scoring

Transparent, rules-based scoring of source authority.

• Domain-Based Scoring - Nature (0.85), arXiv (0.50), etc.
• Citation Modifiers - Higher scores for well-cited sources
• Cross-Reference Agreement - Boosts for claims confirmed by multiple sources
• Predatory Publisher Detection - Flags questionable sources

Research Depth Levels

Academic Sources

Zorora searches across 7 academic databases:

• Google Scholar
• PubMed
• CORE
• arXiv
• bioRxiv
• medRxiv
• PubMed Central (PMC)

Plus web search (Brave + DuckDuckGo) and Asoba newsroom integration.

Getting Started

Prerequisites

• Python 3.8+
• LM Studio running on http://localhost:1234
  • Download: lmstudio.ai
  • Load a 4B model (e.g., Qwen3-4B)
• Brave Search API key (optional) - For enhanced web search

Installation

From GitHub:

pip install git+https://github.com/AsobaCloud/zorora.git

From source:

git clone https://github.com/AsobaCloud/zorora.git
cd zorora
pip install -e .

Run

Terminal Interface:

zorora

Web Interface:

zorora web
# Opens at http://localhost:5000

Usage

Deep Research Query

Terminal:

[1] > What are the latest developments in large language model architectures?

The system automatically detects research intent and executes the deep research workflow.

Web UI:

1. Open http://localhost:5000
2. Enter your research question
3. Select depth level (Quick/Balanced/Thorough)
4. Click “Start Research”
5. View synthesis, sources, and credibility scores

Programmatic Access

from engine.research_engine import ResearchEngine

engine = ResearchEngine()
state = engine.deep_research("Your research question", depth=1)
print(state.synthesis)

Search Past Research

# Search past research
results = engine.search_research(query="LLM architectures", limit=10)

# Load specific research
research_data = engine.load_research(results[0]['research_id'])

Slash Commands

Research Commands

• /search <query> - Force deep research workflow
• /ask <query> - Conversational mode (no web search)

Code Commands

• /code <prompt> - Code generation or file editing
• /develop <request> - Multi-step development workflow

System Commands

• /models - Interactive model selector
• /config - Show current configuration
• /history - Browse saved sessions
• /help - Show available commands

Performance

Why Local-First?

Problem: Cloud-based research tools require uploading your queries and data to external servers, creating privacy concerns for sensitive research.

Solution: Zorora runs entirely on your machine:

• Pattern matching routes queries (no LLM decision overhead)
• Hardcoded 6-phase research pipeline
• Local SQLite + JSON storage
• Zero cloud dependencies for core functionality

Result: Complete privacy, 100% reliability with 4B models, 1/3 the RAM of 8B orchestrators.

Open Source

Zorora is open source under the MIT license.

• GitHub: github.com/AsobaCloud/zorora
• Issues: Report bugs and request features
• Pull Requests: Contributions welcome

Support & Resources

Documentation

• README
• Commands Reference
• Architecture Docs

Support

• Email: support@asoba.co
• Discord: Join our community
• GitHub Issues: Report bugs

Get Help & Stay Updated