Open Source Projects
Production-ready tools that demonstrate our commitment to practical innovation
Innovation Through Open Source
We believe the best way to demonstrate our capabilities is to show, not tell. Our open-source projects aren't academic exercises or proof-of-concepts - they're production-ready tools that solve real business problems. Each project reflects our philosophy: practical solutions that deliver measurable value.
By making these tools open source, we're not just contributing to the community - we're showing you exactly how we think and work. No black boxes, no proprietary magic. Just solid engineering and smart problem-solving that you can inspect, use, and build upon.
Agent Code Factory
AI-Powered Feature Pipeline with Local LLM Support
The Development Challenge
Software development involves repetitive tasks that consume valuable engineering time: parsing feature requirements, analyzing codebases, designing implementations, generating code, running tests, and validating deployments. Traditional AI coding assistants require expensive API keys and often produce inconsistent results without proper context. Teams need a reproducible, controllable pipeline that maintains code quality while accelerating development.
Our Solution
Agent Code Factory is a reproducible AI pipeline where you describe a feature and 20+ specialized agents work together to deliver production-ready code. Running entirely on local LLMs via Ollama, it eliminates API costs while keeping your code private. The core principle: AI assists - human owns the release.
Multi-Agent Pipeline
26 specialized agents handle distinct tasks: Spec Agent parses requirements, Context Agent analyzes your repo, Design Agent proposes architecture, Implementation Agent generates code, and more.
Local LLM Powered
Runs entirely on Ollama with no API keys needed. Keep your code private and eliminate per-token costs. Supports models like Qwen3 for reasoning and code generation.
RAG Semantic Search
Vector embeddings enable semantic code search across your repository. Find conceptually related code, not just exact keyword matches, improving context for better generations.
Docker Validation
Test generated code in isolated containers with automated health checks and API probes. Catch integration issues before they reach production.
Security Scanning
Integrated Bandit analysis, YAML safety checks, and Docker security validation catch vulnerabilities early. Policy enforcement blocks secrets and critical issues.
Code Review Agent
Automated senior engineer code review evaluates architecture decisions, identifies potential issues, and provides actionable feedback before approval.
Key Features
- Approval Checkpoints: Human-in-the-loop design ensures you control what gets implemented. Review and approve at each critical stage
- Policy Enforcement: YAML-based rules for coverage thresholds, security requirements, and approval workflows
- Project Scaffolding: Create new projects from built-in or custom templates with marketplace sharing
- Git Integration: Automatic commits, branches, tags, and PR creation for seamless version control
- GitHub Actions: Auto-generated CI/CD workflows with rollback jobs and canary deployment support
- Web Dashboard: Monitor pipeline runs, view artifacts, and deploy projects through an intuitive interface
- Resume Support: Pause and resume pipeline runs without losing progress
- Code Health Check: Self-healing deprecation scanning with automatic fixes via pyupgrade and ruff
Pipeline Stages
Specification
SpecAgent parses your feature description into structured requirements. DecompositionAgent breaks complex features into manageable sub-tasks.
Context & Design
ContextAgent analyzes your repository with RAG-powered semantic search. DesignAgent proposes implementation approach with human approval checkpoint.
Implementation
ImplementationAgent generates code patches. FixAgent runs automatic validation loops to catch and correct errors iteratively.
Validation & Deploy
Docker build, health checks, API probes, security scan, code review, and policy enforcement before final approval and deployment.
Technical Implementation
Technology Stack:
- Core: Python 3.11+ with async processing
- LLM: Ollama backend with Qwen3:14b (reasoning) and Qwen3-coder:30b (code generation)
- RAG: Vector embeddings with nomic-embed-text model
- Dashboard: FastAPI web UI for monitoring and management
- Validation: Docker containers with API probes and health checks
- Security: Bandit, YAML validation, Docker security scanning
Quick Start:
# Install and setup
pip install -e ".[dev]"
ollama pull qwen3:14b
ollama pull nomic-embed-text
# Index your repository for semantic search
coding-factory index
# Run the pipeline
coding-factory run "Add login rate-limit"
# Launch web dashboard
coding-factory dashboardUse Cases
Feature Development
Describe a feature in natural language and let the pipeline generate implementation code, tests, documentation, and deployment configurations.
Code Migration
Use RAG search to understand existing patterns, then generate modernized code that follows your codebase conventions and style.
Security Hardening
Run security scans across your project, identify vulnerabilities, and generate fixes with policy enforcement ensuring issues are resolved.
Documentation Generation
RAG optimization preprocesses docs with smart chunking, Q&A extraction, and summaries for dramatically better retrieval.
benchHUB
Comprehensive Performance Intelligence for Modern Infrastructure
The Business Problem
Organizations invest millions in computational infrastructure without truly understanding its capabilities or limitations. They make critical decisions about hardware upgrades, cloud migrations, and capacity planning based on gut feeling rather than data. Performance bottlenecks hide until they cause production failures. Different teams use different benchmarks, making comparisons impossible.
Our Solution
benchHUB is a cross-platform performance benchmarking suite that provides comprehensive insights into system capabilities across six critical dimensions:
CPU Performance
Measure single and multi-core processing power with real-world workloads that reflect actual business applications, not synthetic tests.
GPU Capabilities
Evaluate GPU performance for machine learning, data processing, and visualisation tasks. Supports both NVIDIA CUDA and Apple Silicon.
Memory Bandwidth
Understand memory performance characteristics that often bottleneck real applications. Identify limitations before they impact production.
Disk I/O Operations
Test storage performance under various workloads to understand database, file processing, and backup capabilities.
Machine Learning Performance
Benchmark ML model training and inference speeds to make informed decisions about AI infrastructure investments.
Visualization & Rendering
Assess graphics and plotting capabilities for data visualisation, reporting, and dashboard performance.
Key Features
- Normalized Scoring: Compare different hardware configurations with our Reference Index system that provides apples-to-apples comparisons
- Configurable Intensity: Run light, standard, or heavy benchmarks depending on your needs and time constraints
- Interactive Dashboard: Beautiful Streamlit-based visualisation of results with historical tracking and trend analysis
- Anonymous Leaderboard: Compare your infrastructure against industry peers without revealing sensitive information
- Modular Architecture: Easily extend with custom benchmarks specific to your workloads
- CI/CD Integration: Automate performance regression testing in your development pipeline
Business Value
Infrastructure Planning
Make data-driven decisions about hardware investments. Know exactly what performance gains you'll get before spending money on upgrades.
Cost Optimisation
Identify over-provisioned resources and right-size your infrastructure. Many organisations achieve significant savings on infrastructure costs.
Performance Troubleshooting
Quickly identify performance bottlenecks and degradation. Catch problems before they impact users.
Vendor Evaluation
Objectively compare cloud providers, hardware vendors, and configuration options with standardised benchmarks.
Technical Implementation
Technology Stack:
- Core: Python 3.8+ with NumPy, SciPy for computational workloads
- GPU Support: CUDA for NVIDIA, Metal Performance Shaders for Apple Silicon
- Dashboard: Streamlit for interactive visualisation
- API: FastAPI for result submission and retrieval
- Database: SQLite for local storage, PostgreSQL for server deployment
Installation & Setup:
# Create virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Run benchmark with configurable intensity
python benchhub.py --profile standard # Options: light, standard, heavy
# Launch interactive dashboard
streamlit run dashboard.pyUnique Scoring System:
- Composite score combining CPU (40%), GPU (40%), and Memory (20%)
- Inverse timing calculation for fair comparison
- Scores normalised and capped at 1000
- Graceful degradation for systems without dedicated GPUs
Use Cases
Cloud Migration Planning
A financial services company used benchHUB to evaluate AWS, Azure, and GCP instances for their trading platform. Result: substantial cost savings by choosing the right instance types.
ML Infrastructure Optimisation
An AI startup used benchHUB to optimise their model training infrastructure, significantly reducing both training time and costs.
Performance Regression Testing
A software company integrated benchHUB into their CI/CD pipeline to catch performance regressions before deployment, preventing several potential production issues.
The Business Problem
Knowledge workers spend significant time searching for and reading documents. Critical information gets buried in lengthy reports, contracts, and technical documentation. AI tools like ChatGPT have token limits that make processing large documents expensive or impossible. Teams miss important details in contracts, requirements get overlooked in specifications, and decisions are delayed while waiting for document review.
Our Solution
Max_Agent is an intelligent document processing system that automatically summarises large documents while preserving critical information. Unlike simple summarisation tools, Max_Agent understands the importance of technical content like code blocks, equations, tables, and diagrams.
Intelligent Summarisation
Uses state-of-the-art transformer models to create concise summaries that capture the essence of documents without losing critical details.
Technical Content Preservation
Automatically identifies and preserves code blocks, mathematical equations, data tables, and technical specifications in full.
Customizable Compression
Adjust summarisation ratios based on your needs - from executive summaries to detailed abstracts.
Multi-Format Support
Process PDFs, Word documents, technical manuals, research papers, and more with format-aware extraction.
AI-Optimized Output
Generate summaries specifically optimised for LLM consumption, maximizing context window utilization.
Batch Processing
Process entire document libraries automatically with configurable workflows and quality controls.
Key Features
- Smart Extraction: Uses pdfplumber for accurate text extraction that preserves document structure and formatting
- GPU Acceleration: Optional CUDA support for much faster processing of large document batches
- Quality Metrics: Automatic quality scoring ensures summaries maintain information fidelity
- Logging & Tracking: Complete audit trail of all processed documents for compliance and quality assurance
- API Integration: RESTful API for integration with existing document management systems
- Custom Models: Support for fine-tuned models specific to your domain (legal, medical, technical)
Business Value
Time Savings
Dramatically reduce document review time. What took hours now takes minutes, freeing knowledge workers for higher-value tasks.
Improved Decision Making
Ensure critical information isn't missed. Max_Agent highlights key points and preserves all technical details.
AI Cost Reduction
Substantially reduce LLM API costs by optimizing document size while maintaining information completeness.
Compliance & Risk
Never miss critical contract terms or regulatory requirements buried in lengthy documents.
Technical Implementation
Technology Stack:
- Core: Python 3.10+ with async processing support
- PDF Processing: pdfplumber for accurate text extraction
- Summarisation: facebook/bart-large-cnn transformer model
- PDF Generation: FPDF for creating summarised documents
- GPU Support: CUDA for accelerated processing (optional)
- Configuration: YAML-based for easy customisation
Installation & Setup:
# Clone repository
git clone https://github.com/Tennisee-data/Max_Agent.git
cd Max_Agent
# Create virtual environment (recommended for isolation)
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install package in development mode
pip install -e .
# Configure settings (optional)
# Create config.yaml to customize input/output directories,
# summarization ratios, and model parameters
# Run the application
max-agent
# Or with custom config
max-agent --config config.yamlKey Capabilities:
- Preserves code blocks and mathematical equations intact
- Customizable summarisation ratios for different use cases
- Automatic text cleaning and preprocessing
- Intelligent text binning for optimal summarisation
- Comprehensive logging for audit trails
- Batch processing for document libraries
Origin Story:
"The idea came when I wanted to pack and optimise the PDFs I could use to provide context in the configuration of an OpenAI ChatGPT Agent" - François Reeves
Use Cases
Legal Document Review
A law firm uses Max_Agent to process thousands of pages of discovery documents, drastically reducing review time while ensuring no critical information is missed.
Technical Documentation
A software company uses Max_Agent to create concise versions of API documentation for AI assistants, significantly improving developer productivity.
Research & Development
A pharmaceutical company processes research papers much faster, accelerating literature reviews and competitive intelligence gathering.
Contract Management
A procurement team uses Max_Agent to quickly review vendor contracts, identifying key terms and potential risks in minutes instead of hours.
Real-World Impact
The origin story of Max_Agent demonstrates our philosophy perfectly: "The idea came when I wanted to pack and optimise the PDFs I could use to provide context in the configuration of an OpenAI ChatGPT Agent." A real problem, encountered in real work, solved with a practical solution that others can now benefit from.
This isn't theoretical - it's a tool we use daily in our own work, continuously improving based on real-world usage. When we help clients implement document processing solutions, Max_Agent often forms the foundation, customized for their specific needs.
PublicLedger
Blockchain-Based Transparency for Public Spending
The Democratic Problem
Public spending decisions happen behind closed doors. Billions in taxpayer funds are allocated with limited transparency. Meeting minutes are sanitized, records go missing, and citizens have no reliable way to track how and why their money is being spent. Corruption thrives in opacity, and even legitimate decisions suffer from lack of public trust.
Our Solution
PublicLedger is a blockchain-based accountability system that creates an immutable record of all public spending decisions. Every meeting, every vote, every allocation is permanently recorded and publicly accessible:
Immutable Meeting Records
Complete meeting recordings, transcripts, and notes stored on blockchain. No retroactive editing, no convenient deletions, no lost records.
Decision Documentation
Every funding decision includes who proposed it, who voted for it, what the rationale was, and all supporting documents. Complete accountability trail.
Public Access Interface
User-friendly portal for citizens to search, browse, and analyze public spending. Advanced search by department, official, amount, or keyword.
Automated Alerts
Subscribe to notifications about spending in areas you care about. Get alerted when new allocations are made or decisions are recorded.
Investigation Tools
Built-in analytics for journalists and investigators to identify patterns, anomalies, and potential conflicts of interest.
Smart Contract Enforcement
Automated compliance checks and spending limits enforced by smart contracts. Prevents unauthorized or irregular spending patterns.
Key Features
- Distributed Storage: No single point of failure or control - data is replicated across multiple nodes
- Cryptographic Verification: Every record is cryptographically signed and timestamped
- Open Source: Complete transparency in how the system works, not just what it records
- Privacy Protection: Personal information of citizens is protected while maintaining official accountability
- Multi-format Support: Handles documents, audio, video, and structured data
- API Access: Enables third-party tools and analysis platforms
Impact Vision
Restore Public Trust
When citizens can verify how their money is spent, trust in government institutions can begin to rebuild.
Reduce Corruption
Sunlight is the best disinfectant. Knowing that every decision is permanently recorded changes behavior.
Improve Decision Quality
Officials make better decisions when they know they'll be held accountable for them.
Enable Civic Engagement
Informed citizens can participate more effectively in democratic processes.
Development Status
Current Phase: Architecture Design & Prototype Development
Target Launch: 2025
Technology Stack (Planned):
- Blockchain: Ethereum-based or custom chain optimised for document storage
- Storage: IPFS for large file distribution
- Frontend: React-based citizen portal
- Analytics: Python-based investigation tools
- Smart Contracts: Solidity for spending rules enforcement
Get Involved
This project aims to be a public good. We're seeking:
- Government partners willing to pilot the system
- Developers interested in contributing to the codebase
- Legal experts to help navigate regulatory requirements
- Citizens and advocacy groups to help shape requirements
Watch this space for updates, or contact us to learn more about getting involved.
The Investment Challenge
Individual investors and financial professionals face an overwhelming flood of market data, news, and analysis. Making informed investment decisions requires processing vast amounts of information across multiple sources - financial statements, market trends, news sentiment, and technical indicators. Traditional tools are either too simplistic or require expensive institutional subscriptions.
Our Solution
StockIceberg.ai is a comprehensive financial analysis platform that harnesses the power of artificial intelligence to deliver institutional-grade market insights to everyone. The platform aggregates, analyzes, and presents complex financial data in actionable formats. Start free and upgrade as your needs grow.
AI-Powered Analysis
Advanced machine learning models analyze market patterns, sentiment, and fundamentals to surface insights that matter.
Comprehensive Data
Access to extensive financial data including real-time quotes, historical trends, and detailed company fundamentals.
Sentiment Analysis
Natural language processing of news, social media, and analyst reports to gauge market sentiment.
Technical Indicators
Automated technical analysis with customizable indicators and pattern recognition.
Portfolio Insights
Track and analyze portfolio performance with risk metrics and optimization suggestions.
Real-Time Alerts
Stay informed with intelligent alerts on price movements, news events, and market anomalies.
Pricing
Free Tier
Get started with core features at no cost. Access basic analysis tools, limited historical data, and essential market insights.
Premium Features
Unlock advanced AI analysis, extended data history, real-time alerts, and priority support with a premium subscription.
Better Decisions
Make investment decisions backed by comprehensive data analysis rather than gut feelings or incomplete information.
Risk Management
Identify potential risks and opportunities before they become obvious to the broader market.
The Communication Challenge
Clear, professional communication is essential in today's business environment. Yet many professionals struggle with proofreading, editing, and polishing their written content. Grammar errors, unclear phrasing, and inconsistent tone can undermine credibility and impact.
Our Solution
ProfText is a professional text enhancement platform designed to help users refine their written communications. Whether you're preparing business documents, academic papers, or professional correspondence, ProfText provides the tools to ensure your message is clear, polished, and impactful.
Professional Proofreading
Advanced grammar and spelling correction that goes beyond basic spell-check to catch subtle errors and improve readability.
Style Enhancement
Suggestions to improve clarity, conciseness, and professional tone while maintaining your authentic voice.
Context-Aware Editing
Understands the context of your writing to provide relevant suggestions tailored to your specific use case.
Multi-Format Support
Works with various document types and text formats to fit seamlessly into your existing workflow.
A Digital Winter Wonderland
GlobeShake brings the magic of snow globes into the digital age. This interactive web experience lets users shake, customize, and share their own virtual snow globe creations - a delightful blend of nostalgia and modern web technology.
The Experience
Whether you're looking for a moment of digital tranquility, a unique greeting to share with friends, or simply a playful distraction, GlobeShake offers an enchanting interactive experience that captures the whimsy of classic snow globes.
Interactive Animation
Shake your device or click to watch the snow swirl and settle in realistic physics-based motion.
Customization Options
Personalize your snow globe with different scenes, colors, and snowfall effects.
Shareable Creations
Create unique snow globe moments and share them with friends and family.
Cross-Platform
Enjoy the experience on desktop, tablet, or mobile - the magic travels with you.
Why We Build in the Open
Transparency Builds Trust
When you can see exactly how our tools work, you know there are no black boxes or vendor lock-in. What you see is what you get, and you can verify it yourself.
Community Makes Us Better
Every issue reported, every pull request submitted makes our tools better. The community's diverse use cases push us to build more robust, flexible solutions.
Proof of Capability
Our open-source projects demonstrate our technical capabilities better than any case study. You can see our code quality, architecture decisions, and problem-solving approach.
Accelerating Innovation
By sharing our tools, we help others solve problems faster. The time saved can be invested in innovation, pushing the entire industry forward.
Get Involved
Our projects are actively maintained and we welcome contributions from the community:
Find all our projects on GitHub:
github.com/Tennisee-dataNeed a Custom Solution?
Our open-source projects demonstrate our capabilities, but every business has unique challenges. We can:
Many of our client engagements start with "We saw your open-source project and need something similar but..."