# Sources, ecosystem position and licensing ## 2026-10-23 model-mining extension The new model, discovery, synchronous-product search, XML transport, adapters, policies and cohort implementation are independent MoonBit code written with AI assistance. The MIT runtime does not call, vendor, translate or redistribute PM4Py. Its ordinary P/T firing, classical Alpha footprint/maximal pairs, prefix acceptor, and nonnegative shortest-path algorithms follow public mathematical descriptions. Behavioural references: [process discovery overview](https://processmining.org/process-discovery.html), [PM4Py discovery API](https://processintelligence.solutions/app/pm4py/api/api/pm4py.discovery.html), [XES task force description](https://www.tf-pm.org/resources/xes-standard/about-xes), [BPMN 2.0.2](https://www.omg.org/spec/BPMN/2.0.2) and PNML ordinary P/T XML conventions. No specification text, external dataset, screenshots or diagrams are redistributed. All new logs/models/fixtures are synthetic original examples, not real customer data. The optional Python harness uses the published public API of [PM4Py 2.7.23](https://github.com/process-intelligence-solutions/pm4py), AGPL-3.0, as an external development-time reference. Its source and third-party dependencies are installed in an ignored local directory or isolated environment, not included in the Mooncakes package. Retain upstream licensing when installing or redistributing that separate environment; it is not licensed by this project's MIT file. The harness and fixture generator are original test integration code, not a port of PM4Py's implementation. See [models](models.md) for the distinct comparison paths and weighted-alignment reference limitation found during testing. Basic Petri-net firing already exists in the adjacent `moonpetri` project noted below. The independent contribution here is model discovery + trace/model conformance + standard exchange + case/resource/temporal analysis, not claiming the token primitive is new. Its compact ordinary-net representation is shared by discovery and alignment to keep accepting-state, budget and witness semantics consistent; adapters can use JSON/PNML rather than duplicate another simulator. There is no claim that all unpublished projects are absent or that this reaches PM4Py's full algorithm catalogue. MoonProcessLens has an independent process-log root package: case reconstruction, empirical directly-follows discovery, finite-reference weighted edit alignment and interval performance analysis. Its code and synthetic fixtures were created with AI assistance; no external process-mining source or production dataset was imported. The retained finite-reference alignment is a direct dynamic program verified against independently enumerated common subsequences, not a port of a Petri-net alignment engine. A 2026-10-22 review of 3,221 official directory entries and selected fixed APIs did not identify an equivalent complete process-analysis library within that scope. [moonpetri](https://github.com/okMambaOut/moonpetri/tree/e5eb4ed854be19e68dcb4ddb092f8d0bf6be61f5) models/fires Petri nets; [moonflowgraph](https://github.com/AlexenderSokolov/moonflowgraph/tree/e2faa7c7105161b561111f0f31e1ef57071b713d) plans predefined task DAGs. [moon-loglens](https://github.com/sundaysebasidian-byte/moon-loglens/tree/4603ef7527332002684200ff4a9a88c313f944f4) provides general JSONL aggregation, while [moon-tracelens](https://github.com/btlqql/moon-tracelens) provides ingestion, timeline and metric comparisons. ProcessLens instead groups activity histories into case paths and produces path edits and interval explanations. These bounded findings do not establish that all concepts or primitives are new. ## Evolution and optional sequence package The project continues the local repository previously named MoonCEP. Its detector is now in `sequence`; it is retained and identified rather than presented as new process-mining code. Root analysis does not import it. Renaming does not erase Git history or previously published artifacts. Package/account changes require the owner's later instructions. The original detector's source/reference notes follow. MoonCEP's MoonBit implementation and synthetic tests were created for this project with AI assistance. It is not a line-by-line port and does not include Apache Flink or SASE source code, diagrams or external data fixtures. Conceptual references: - [Apache Flink CEP documentation](https://nightlies.apache.org/flink/flink-docs-release-1.20/docs/libs/cep/): vocabulary for contiguity, repetition and match selection. Flink is Apache-2.0 licensed; MoonCEP defines independent boundary behavior. - [Wu, Diao and Rizvi, SASE, SIGMOD 2006](https://www.cs.columbia.edu/~ewu/files/papers/sase-sigmod06.pdf): automata-based event recognition, windows and delayed absence confirmation. - [Agrawal et al., SIGMOD 2008](https://people.cs.umass.edu/~yanlei/publications/sase-sigmod08.pdf): take/ignore/proceed choices and capture evidence for event-stream matching. - [Foundations of Complex Event Processing](https://arxiv.org/abs/1709.05369): explicit semantics and the importance of stating the supported model before claiming complexity guarantees. Dependencies: the standard library is supplied by the MoonBit toolchain. The file CLI uses [moonbitlang/x](https://github.com/moonbitlang/x) 0.4.49 under Apache-2.0. Dependency source is fetched by Moon, not vendored or relicensed. The original project code is MIT licensed, as stated in the root LICENSE. The small C#/PowerShell Windows argument adapter is reused from the author's local MoonBit tooling and adapted to this project. It changes only extended drive-path arguments passed to installed official tools. It is auxiliary build code and excluded from the MoonBit source-size count. ## Ecosystem position MoonCEP's independent contribution is temporal sequence recognition with named captures, repetition, exclusion/absence and explicit selection semantics. [Moon Stream Quality](https://github.com/zmjknn/moon-stream-quality) evaluates data-quality rules; [TideWeave](https://github.com/LMK-ai-nb/TideWeave) maintains correctable aggregate windows; [MoonUp](https://github.com/mw1696803-hub/moonup) implements Rete working-memory rules. Their event and time primitives are adjacent capabilities; this project does not claim those primitives are new to MoonBit. Directory and source checks on 2026-10-22 did not identify an equivalent published sequence detector. This is a dated, scoped investigation, not proof that no hidden or subsequently released implementation exists. Detailed process reports and competition materials are maintained outside this repository.