SOFTWARE / OPEN SOURCE / UFFESCIENCE

Software

Open-source tools, middleware, and research artifacts developed by the UFFeScience Research Group at the Institute of Computing, UFF. They cover workflow execution, provenance data management, scheduling, and data analysis for computational science.

All repositories are public. If you use one of these tools in your research, please cite the associated paper.

FEATURED / FLAGSHIP PROJECT

AkôFlow

Actively developed
EXECUTION / MIDDLEWARE

Workflow engines and middleware

4 repositories
Java · Cloud workflows

SciCumulus

Workflow engine for running parameter-sweep experiments on distributed and cloud resources. This repository is a full architectural refactoring of the original UFF system, keeping its execution semantics with a modular, testable codebase.

Scala · Apache Spark

SAMbA-RaP

Extension of Apache Spark that runs black-box scientific programs over raw data files while capturing provenance and domain-specific data, with built-in version control of the files produced.

Python · Serverless

Denethor

Execution monitoring and provenance collection for scientific workflows whose activities run as AWS Lambda functions. Runs either locally or in serverless mode.

Java · Spring Boot

PhenoManager

REST API and worker service for managing computational experiments and their executions (workflows, executables, commands, and web services) on cloud and cluster environments.

PROVENANCE / OPEN SCIENCE / VISUALIZATION

Provenance, analysis, and open science

7 repositories
Python · Deep learning

DLProv

Provenance services for deep learning workflows, covering pre-processing, training, and evaluation. Supports online analysis in MonetDB, W3C PROV documents, and Neo4j graphs. Distributed as a Docker image.

Python · Agentic workflows

ANCHOR

Captures and analyzes the provenance of agentic workflows with an interaction-centric model of agents, tools, users, model calls, and delegations, so multi-agent runs can be traced, audited, and reproduced.

TypeScript · Open science

Triptych

Links research data, source code, and publications spread across GitHub, Zenodo, OSF, and arXiv, so the full lifecycle of a scientific project can be traced and reproduced.

Jupyter · Urban visualization

UTK-PROV

Provenance-aware extension of the Urban Toolkit. Records how grammar-based urban visualizations evolve during construction and stores the history in a relational database for querying.

Python · Federated learning

FLARE-PROV

Federated k-means clustering with scikit-learn, using MiniBatch-style aggregation, instrumented with DfAnalyzer to capture prospective and retrospective provenance of each round.

JavaScript · Federated learning

FedObserve

Web application for managing and visualizing federated learning experiments, with a t-SNE service for model inspection. Uses AkôFlow for AWS provisioning and execution monitoring.

Python · Data repositories

Dataverse Dados de Rede

Scripts that prepare, transform, and load network and machine learning datasets, with their metadata, into Dataverse repositories through an automated workflow.

SCHEDULING / METAHEURISTICS

Scheduling and optimization

3 repositories
Jupyter · Confidentiality

CYCLOPS

Schedules cloud-based workflows under data confidentiality constraints. Builds a conflict graph between data files and jointly decides activity scheduling and data placement in public clouds.

C++ · GRASP

MemoirGRASP

GRASP-based execution strategy for multisite cloud workflows that reuses previously computed intermediate data, weighing recomputation against storage cost and letting users prioritize makespan or financial cost.

Rust · Urban routing

Protector

Heuristics for the Routing of Police Vehicles in Large Urban Centers (RVP-Urb) problem: plans preventive patrol routes that maximize coverage of crime hotspots under route-length, fleet, and territorial constraints.

APPLICATIONS / REPRODUCIBILITY

Research artifacts and applications

4 repositories

Code and data that accompany recent papers, published so results can be reproduced and extended.

Python · Bioinformatics

CellLens+

Interpretable cell-type annotation for single-cell RNA-seq data. Combines graph-based community detection with a deterministic scoring engine, plus an LLM layer that explains, but never decides, the predicted identities.

JavaScript · Phylogenetics

PhyloTreeMiner

Mining of patterns in phylogenetic trees, available as a command-line workflow and as an interactive web demo.

Python · Graph databases

Graph Modeling Benchmark

Benchmark built on mgBench that measures how four graph modeling transformations affect throughput, latency, CPU, memory, and storage across seven analytical queries.

Python · LLMs for workflows

Can LLMs Engineer Parallel Workflows?

Artifacts of a study on LLM-generated and LLM-translated workflows: k-means, matrix multiplication, and lysozyme pipelines in Nextflow, Parsl, and PyCOMPSs, with human-written reference implementations.

AKÔFLOW / DRIVERS / EXAMPLES

AkôFlow ecosystem

5 repositories

Companion repositories with runnable examples and storage drivers for AkôFlow.

This page highlights a selection of repositories. The full list, including older and archived projects, is on the UFFeScience GitHub organization.