pictl — Process Intelligence Control

21 process mining algorithms compiled to WebAssembly. Discover process models from event logs in browsers and Node.js — no Python, no services, just npm install.

Quick Start

CLI (60 seconds)

npm install -g @pictl/cli
pictl run process-log.xes

Node.js (30 seconds)

const pictl = require('@pictl/engine');
await pictl.init();

const log = pictl.load_eventlog_from_xes(xesContent);
const dfg = JSON.parse(pictl.discover_dfg(log, 'concept:name'));
console.log(`${dfg.nodes.length} activities, ${dfg.edges.length} flows`);

Browser (30 seconds)

<script type="module">
  import pictl from '@pictl/engine';
  await pictl.init();
  const log = pictl.load_eventlog_from_xes(xesContent);
  const dfg = JSON.parse(pictl.discover_dfg(log, 'concept:name'));
</script>

Streaming (IoT / infinite event streams)

const pictl = require('@pictl/engine');
await pictl.init();

// Open session — no full log held in memory
const handle = pictl.streaming_dfg_begin();

// Feed events as they arrive
pictl.streaming_dfg_add_event(handle, 'case-1', 'Register');
pictl.streaming_dfg_add_event(handle, 'case-1', 'Approve');
pictl.streaming_dfg_close_trace(handle, 'case-1');

// Live snapshot (non-destructive)
const dfg = JSON.parse(pictl.streaming_dfg_snapshot(handle));

// Finalize: flush open traces, return DFG
const result = JSON.parse(pictl.streaming_dfg_finalize(handle));
console.log(`${result.nodes} nodes, ${result.edges} edges`);

What It Does

pictl discovers process models from event logs. Give it a log of activities (who did what, when), and it finds the underlying process structure.

| Capability | What It Gives You | | ---------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | | 21 discovery algorithms | DFG, Alpha++, ILP, Genetic, PSO, A*, DECLARE, Heuristic Miner, Inductive Miner, Hill Climbing, ACO, Simulated Annealing, Process Skeleton, Optimized DFG | | 6 ML analysis algorithms | Classify, cluster, forecast, anomaly detection, regression, PCA | | Streaming API | Process infinite event streams with bounded memory | | Conformance checking | Token-based replay with fitness, precision, generalization | | 20+ analytics functions | Bottlenecks, variants, concept drift, dependencies, rework detection | | Predictions | Next activity, remaining time, outcome, drift detection |

Input: XES or JSON event logs. Output: Petri nets, DFGs, process trees, DECLARE models, JSON analytics.

CLI Reference

pictl run <log.xes>              # Process discovery
pictl compare <algos> -i <log>   # Side-by-side algorithm comparison
pictl diff <log1> <log2>         # Compare two event logs
pictl predict <task> -i <log>    # Predictive mining (next-activity, remaining-time, outcome)
pictl drift-watch -i <log>       # Real-time EWMA drift monitoring
pictl ml <task> -i <log>         # ML-powered analysis (classify, cluster, forecast, anomaly)
pictl watch                      # Config file watcher — re-run on change
pictl status                     # WASM engine health + system info
pictl doctor                     # 6-check environment diagnostic
pictl explain                    # Human/academic algorithm explanations
pictl init                       # Scaffold pictl.toml + .env.example
pictl results                    # Browse saved results in .pictl/results/
pictl powl <subcommand>          # POWL model analysis

Output formats: --format human (colored terminal) or --format json (structured output)

Performance Benchmarks

Version: v26.4.7 | Hardware: Apple M3 Max (16P/4E, 36GB unified memory) | Methodology: Median of 7 runs

At a Glance (10K cases)

| Tier | Algorithms | Time | Use Case | | --------------- | -------------------------------------- | ---------- | ------------------------------------------ | | ⚡ Ultra-fast | DFG, Process Skeleton | 2.7–3.0 ms | Real-time dashboards, high-throughput APIs | | ⚡ Fast | Heuristic Miner, Inductive Miner | 14–25 ms | Interactive discovery, UI-driven analysis | | 🚀 Evolutionary | Genetic, ACO, PSO, Simulated Annealing | 21–25 ms | Quality optimization via population search | | 🔍 Optimal | ILP, A* Search | 77–87 ms | Best possible model, provable quality | | 📊 Streaming | Streaming DFG, Noise-Filtered DFG | 69–135 ms | Infinite streams, IoT, memory-constrained |

Key metrics:

• All 21 algorithms tested on real data (BPI 2020: 10,500 traces, 141K events)
• Linear scaling from 100 to 50,000+ cases
• Streaming: 1.4–23x overhead for bounded memory on infinite event streams
• Browser benchmarks: ~40–60% slower than Node.js (expected WASM sandbox overhead)

📖 Full benchmark report: docs/BENCHMARKS.md 📖 Benchmark documentation: docs/benchmarks/ — tutorials, methodology, reference

Documentation

Getting Started

• Quick Start — 5-minute setup
• Build Guide — Build from source
• Development — Development workflow

Performance & Benchmarks

• Benchmark Results — Full performance tables, all algorithms, all sizes
• Benchmark Docs — Tutorials, how-to guides, methodology, reference
  • Your First Benchmark
  • Understanding Results
  • Streaming vs Batch Tradeoffs
  • Methodology

API & Algorithms

• API Reference — Complete function reference
• Algorithm Reference — All 21 algorithms explained
• Configuration — Config files, ENV vars, profiles

Integration

• MCP Integration — Claude AI / MCP server
• Architecture — System design
• Contributing — How to contribute

Architecture

pictl/
├── wasm4pm/                # Rust core — 21 algorithms compiled to WASM
│   └── src/                #   Discovery, conformance, analytics, streaming
├── packages/               # TypeScript monorepo (9 packages)
│   ├── @pictl/kernel       #   WASM facade — run(algorithm, handle, params)
│   ├── @pictl/engine       #   Lifecycle state machine
│   ├── @pictl/config       #   Zod-validated config, 5-layer precedence
│   ├── @pictl/planner      #   Execution plan generation
│   ├── @pictl/observability#   CLI output, JSONL, OTEL spans
│   ├── @pictl/contracts    #   Receipts, errors, plans, hashing
│   ├── @pictl/testing      #   Parity/determinism/CLI test harnesses
│   ├── @pictl/ml           #   Micro-ML: classify, cluster, forecast, anomaly
│   └── @pictl/swarm        #   Multi-worker coordinator
├── apps/pmctl/             # CLI tool — pictl command
├── benchmarks/             # Node.js + browser benchmark suite
└── docs/                   # Documentation (Diátaxis)

Installation

# CLI
npm install -g @pictl/cli

# Library
npm install @pictl/engine

# From source
git clone https://github.com/seanchatmangpt/pictl.git
cd pictl/wasm4pm
npm install && npm run build

Status

Production Ready — v26.4.7

• All 21 algorithms tested and operational
• All algorithms benchmarked with real Criterion results
• Validated on BPI 2020 (10,500 traces, 141K events)
• Linear scaling from 100 to 50,000+ cases
• Ready for npm publish

License

MIT OR Apache-2.0