prism

Analytics and ML suggestors for Converge formations.

prism is a Converge extension. It keeps feature extraction, analytic packs, training, inference, model registry, monitoring, and deployment-decision suggestors outside the Converge foundation while using Converge contracts for in-loop behavior.

Cargo package: converge-prism-analytics. Rust library name remains prism.

Why It Exists

Converge should not become an analytics framework. Prism gives formations a place to ask data-driven agents for proposals while Converge keeps authority over promotion.

What Prism Owns

• Polars-based ingestion and feature extraction.
• Burn-based inference examples.
• Reusable fuzzy inference capability through prism::fuzzy.
• Analytic pack solvers and typed inputs/outputs.
• Training pipeline agents: dataset, validation, feature engineering, hyperparameter search, model training, evaluation, registry, monitoring, deployment decision, and sample inference.
• Typed proposal provenance through ProvenanceSource / PRISM_PROVENANCE.
• Suggestor-boundary tracing through prism.suggestor.execute spans.
• Compile-fail tests that enforce Converge authority boundaries.

Packs

Pack	Algorithm family
AnomalyDetectionPack	Z-score anomaly detection
ClassificationPack	Logistic classification
DescriptiveStatsPack	Mean, median, variance, percentiles
ForecastingPack	Exponential smoothing
FuzzyInferencePack	Membership functions and explainable fuzzy rules
RankingPack	Weighted multi-criteria ranking
RegressionPack	Linear regression
SegmentationPack	K-means clustering
SimilarityPack	Pairwise vector similarity
TrendDetectionPack	Moving-average trend detection

Boundary

Layer	Responsibility
Converge	Suggestor contract, proposal promotion, and shared context.
Prism	Analytics packs, feature agents, training agents, and ML pipeline behavior.
Products	Domain datasets, model rollout policy, credentials, and deployment topology.

Repository Layout

crates/prism/
  src/engine.rs     FeatureAgent and feature vectors
  src/fuzzy/        Reusable fuzzy membership and inference capability
  src/ingest.rs     CSV, TSV, Parquet, and optional Excel ingestion
  src/model.rs      Burn inference example
  src/packs/        Analytics packs and solvers
  src/training.rs   Training and monitoring suggestors
  tests/            Integration, property, negative, and compile-fail tests

Usage

use prism::FeatureAgent;

let agent = FeatureAgent::new(None);
engine.register_suggestor(agent);

Feature Flags

• Default: none.
• storage: enables optional converge-storage support.
• excel: enables Excel ingestion through calamine.

Development

just check
just check-all
just test
just lint
just doc

Converge platform dependencies resolve from crates.io.

Project Files

• AGENTS.md - agent entrypoint and boundary rules.
• CHANGELOG.md - release notes.
• CONTRIBUTING.md - contribution guide.
• SECURITY.md - vulnerability reporting and operator notes.
• CODE_OF_CONDUCT.md - community expectations.

Status

Extracted from converge/crates/analytics on 2026-05-05 as part of the v3.8 foundation extraction.

License

MIT - see LICENSE.