Factory Intelligence Platform

Factory Intelligence Platform is an early-stage open-source Factory Intelligence Layer for connecting manufacturing systems, contextualizing operational data, detecting process drift, and supporting human-reviewed AI-assisted workflows.

Factory Intelligence Platform is the technical center of the Open Factory Initiative mission. It is designed to help manufacturers, engineers, quality teams, validation professionals, researchers, and open-source contributors build transparent, governed, and interoperable infrastructure for the AI-native factory.

View Repository Get Involved Read the Roadmap

The Problem: Factories Have Data, But Not Shared Context

Manufacturing teams often rely on fragmented systems: PLCs, SCADA, historians, MES, QMS, CMMS, ERP, spreadsheets, validation documents, operator notes, and tribal knowledge. These systems contain valuable information, but they rarely provide shared operational context.

When process drift, downtime, deviations, or quality events occur, teams often spend too much time manually reconstructing what happened. They must gather historian trends, equipment context, alarms, work orders, batch information, quality records, and human observations from separate systems.

Factory Intelligence Platform is intended to reduce that fragmentation by creating an open-source layer for connecting systems, preserving source traceability, and supporting explainable human-reviewed decision-making.

Mission connection

OFI is building shared public infrastructure so factory intelligence can be inspectable, interoperable, and governed by people who understand the manufacturing context.

What the Platform Does

The platform is designed as a modular Factory Intelligence Layer above existing manufacturing systems, with a first emphasis on read-heavy, evidence-backed workflows.

Connect

Connects to existing factory systems and data sources using controlled, open-source-friendly integration patterns.

Contextualize

Normalizes factory events and connects process data, equipment, batches, quality events, maintenance activity, and operational context.

Detect and Explain

Supports explainable detection of process drift, excursions, and patterns that require human review.

Govern

Supports governed AI-assisted workflows with evidence, traceability, review queues, and human accountability.

Initial MVP Workflow

The first MVP is intentionally narrow. It uses a synthetic factory simulator to generate normal, drift, and excursion events. Those events are ingested into a shared factory event model, analyzed by Process Sentinel, organized into an evidence timeline, and presented for human review before any recommendation is accepted.

Step 1
Synthetic Factory Simulator
Step 2
Ingestion Worker
Step 3
Factory Event Store / Unified Namespace
Step 4
Process Sentinel Drift Detection
Step 5
Evidence Timeline
Step 6
Governed Recommendation Queue
Step 7
Web UI Workbench
Step 8
RCA / CAPA Draft Export
Step 9
Factory Memory

Current Project Components

The current repository includes an executable MVP skeleton focused on the simulator-backed Process Sentinel workflow.

Synthetic factory simulator

Deterministic normal, drift, and excursion events.

Shared event contracts

Common factory event schemas for simulator, ingestion, and analysis workflows.

Ingestion service

Event validation, storage, and dead-letter handling.

Process Sentinel

Explainable drift rules, evidence, and recommendation logic.

API service

FastAPI endpoints over stored MVP state.

Web workbench

Placeholder for the future human-review interface.

Local infrastructure

Docker and PostgreSQL path for durable storage evolution.

Repository areas currently include services/simulator, packages/factory-events, services/ingestion, services/process-sentinel, services/api, apps/web, and infra/docker.

Who This Is For

Factory Intelligence Platform is intended for the people who understand manufacturing systems and the people who can help make open-source infrastructure trustworthy.

Automation engineers
Controls engineers
Process engineers
Quality professionals
Validation engineers
Manufacturing IT/OT teams
Historian and data infrastructure engineers
AI/ML practitioners
Cybersecurity contributors
Academic and industry researchers
Open-source maintainers and documentation contributors

Development and Testing Model

The project is developed in public on GitHub using documented contribution workflows, issue tracking, code review expectations, automated tests, type checking, contract testing, and local reproducible development commands. Early releases prioritize simulator-backed workflows so contributors can test core behavior without connecting to real factory systems.

Project practices

Public GitHub development
Documented contribution process
Governance documentation
Security policy
Public roadmap
Linting and type checking
Pytest, Playwright, contract, integration, and end-to-end tests
Local reproducible development commands
Simulator-backed test data

Open-Source Ecosystem Goals

The long-term goal is to grow Factory Intelligence Platform into a sustainable open-source ecosystem for manufacturing intelligence. That means the project must become more than code. It needs public governance, contributor onboarding, security practices, validation-ready documentation, community feedback, technical advisory input, and reusable reference implementations.

Build a broad contributor community across manufacturing, automation, quality, validation, cybersecurity, AI/ML, and research.

Provide transparent governance and roadmap decision-making.

Publish reusable architecture, security, validation, and integration patterns.

Support safe, secure, and responsible adoption in quality-critical manufacturing environments.

Keep the core project open, inspectable, extensible, and community-driven.

Planned local-first AI and site-specific onboarding

The roadmap includes future local-first AI and site-specific onboarding work for adapting the Factory Intelligence Platform to a facility's operating context. This may include mapping assets, process signals, procedures, terminology, quality workflows, and review paths so SLM/LLM-assisted features can provide more relevant, evidence-grounded decision support.

Planned work includes a Site AI Package, site, area, line, and asset mapping, terminology and procedure mapping, RAG before fine-tuning, SLM/LLM evaluation workflows, model governance, human review, and traceable evidence.

Future roadmap work includes site-specific SLM/LLM training and evaluation workflows that can help adapt factory intelligence assistance to a facility's terminology, procedures, asset structure, and operational context while preserving human review, traceable evidence, and governance controls.

Site AI Package for site profiles, area/line/asset hierarchy, process context, and equipment metadata.
Site, area, line, and asset mapping tied to terminology, procedure, quality workflow, and data source context.
RAG before fine-tuning so site-specific assistance starts with cited, retrievable evidence.
SLM/LLM evaluation workflows for grounded answers, refusal behavior, cited evidence, task accuracy, latency, and resource use.
Model governance through model, prompt, tool, and dataset registry concepts.
Human review, governed recommendations, usage logging, and traceable evidence for AI-assisted outputs.

Read Roadmap Details

Current Limitations

This project is early-stage and is not intended for production use in regulated manufacturing environments. The initial implementation uses simulated data and read-only patterns. It does not perform autonomous quality decisions, closed-loop process control, electronic signature approval, or direct writeback to source systems.

Site-specific validation remains the responsibility of adopting organizations. The project can support validation-ready documentation and risk-based assurance practices, but those materials must be evaluated against each organization's intended use, procedures, controls, and operating environment.

Responsible scope control

The first MVP focuses on observable, simulator-backed investigation support. Higher-risk capabilities such as writeback, closed-loop control, and regulated approvals are intentionally outside the current scope.

Help Build Open Factory Intelligence

Contributors can help with code, documentation, manufacturing use cases, validation concepts, cybersecurity review, AI/ML evaluation, historian integration patterns, governance input, and community feedback.

View GitHub Read Contributing Guide Get Involved