Reason about uncertain process histories¶
OCBF takes possibly-conflicting reports about what happened in a process, weighs how much to trust each source, and answers questions about what most likely happened — while keeping the uncertainty explicit instead of guessing a single history. You call it from your own code; it is not a data pipeline, user interface, or scheduler.
Run your first assessment Explore capabilities
Try it¶
Python 3.12 or later is required. From a clone of the repository, with an activated environment (see installation):
The example uses synthetic reports and the core finite reference engine.
Find your way¶
-
Getting started
Learn by doing: install the library and follow a complete synthetic assessment.
-
Concepts
Understand the workflow, the library objects, and why results carry qualifications.
-
How-to guides
Solve one task: interpret reports, choose inference, bound execution, or integrate.
-
Reference
Look up capabilities, configuration, result fields, the glossary, and the API.
-
Development
Contribute: architecture, validation, and documentation maintenance.
One model, explicit handoffs¶
flowchart LR
C["Semantic context"] --> M["Canonical model"]
R["Versioned reports"] --> E["Interpreted evidence"]
E --> M
T["Resolved trust"] --> M
M --> I["Inference"]
I --> B["Qualified posterior"]
B --> Q["Process queries"]
N["Normative reference"] --> Q
Database access, ingestion, credentials, scheduling, and screens belong to the calling application; see the integration boundary.
Understand the answer you receive¶
Results retain their scope, denominators, scientific identities, evidence qualifications, and numerical assessment. A probability is conditional on the supplied model and evidence. An unavailable quantity remains explicit.
See capabilities and limitations and result meanings before using an estimate in a decision.