OCBF

Object-Centric Belief Fusion is a Python library for reasoning about uncertain object-centric histories from heterogeneous evidence.

Supply a fixed schema and candidate universe, interpret versioned reports, resolve manual trust, and compile one probability model. Inference returns declared posterior capabilities; process queries turn those capabilities into qualified answers about duration exceptions, conformance, exposure, and priorities.

Semantic context + interpreted evidence + resolved parameters
                              ↓
                       Canonical model
                              ↓
                    Qualified posterior
                              ↓
                     Process query results

Try it

Python 3.12 or later is required. From a checkout:

git clone https://github.com/RayanJavan/ocbf.git
cd ocbf
python -m venv .venv

Activate the environment with .venv\Scripts\Activate.ps1 in PowerShell or source .venv/bin/activate in a POSIX shell, then run:

python -m pip install -e .
python -m examples.fixed_parameters

The example uses synthetic reports and the core finite reference engine. It writes readable results, manual-trust comparisons, and reproducible correction/retraction calculations under artifacts/fixed-parameters. No native solver or factory data is needed.

Learn and use

Read the documentation at ocbf.readthedocs.io. The sources for each section are also readable in this repository:

• Getting started: installation and a complete assessment.
• Concepts: the workflow, library objects, and their behavior.
• How-to guides: evidence, inference, execution, and integration.
• Capabilities: supported routes and their boundaries.
• Architecture: module responsibilities and extension seams.
• Integrate an application: application inputs and the Mammut example.
• Research: separate dated investigations and proposals.

OCBF preserves evidence gaps and model qualifications. Numerical agreement does not establish physical factory accuracy. The Mammut example requires verified source and identity bindings before it can produce a retained-data assessment.

Develop

python -m pip install -e ".[dev]"
python -m pytest -q

For documentation, use the pinned environment and PowerShell or POSIX commands in documentation maintenance. MkDocs, Material, and mkdocstrings build the guides and generated API reference together. CI checks that site separately from the library suite and uploads the ocbf-site artifact. Read the Docs publishes the same build; the hosting guide records its settings. See the development guide for validation and contribution rules.