QECTOR Decoder Workbench

QECTOR Decoder Workbench v3.5.0

Production Release

Quantum Error Correction Analysis Suite

Download v3.5.0

• QectorWorkbench-V3.5.0-Portable.exe

Key Features

6 GUI Tabs

• Code Explorer — Build codes from 6 families (repetition, ring, rotated_surface, unrotated_surface, toric, heavy_hex) with configurable distance. View code properties (qubits, checks, distance, rate).

• Decoder Lab — Interactive single-syndrome decoding with 5 decoders (union_find, fast_union_find, blossom, sparse_blossom, bp_osd). Configurable error rate and seed. blossom is weight-optimal exact MWPM; sparse_blossom is near-optimal (not exact); the union-find family is fast/approximate with a higher logical error rate.

• Benchmark Suite — Configurable decode benchmarks with throughput and latency stats (mean/p50/p99/min/max). Export results to JSON.

• Batch & Streaming — Batch decode with explicit CPU/CUDA/OpenCL routing (no silent fallback — unavailable backends fail loudly). Streaming decode with sliding-window commit semantics.

• Hardware & System — Auto-detect CUDA, OpenCL, and CPU backends via qector_decoder_v3 APIs (cuda_is_available, opencl_is_available). Note: the standard qector-decoder-v3 wheel ships a CUDA path but no OpenCL kernels, so opencl_is_available() is False unless the backend was built from source with the opencl feature. System info and hardware-aware decoder recommendations via hardware_routing.detect_hardware() and hardware_routing.recommend().

• Documentation Studio — Multi-format doc export: Markdown, JSON, HTML, LaTeX, PDF (matplotlib PdfPages), SVG (matplotlib). Full provenance metadata and decoder recommendation tables.

29-Tool MCP Server

• Real stdio JSON-RPC 2.0 transport (protocol version 2024-11-05)
• stdin/stdout newline-delimited JSON-RPC — no HTTP bridge, no port 8765
• Implements initialize, notifications/initialized, ping, tools/list, tools/call
• All 29 tools wired to the real backend.py API

Backend

• Thin Python wrapper over qector_decoder_v3 (compiled Rust core + public Python layer), developed and verified against the installed v0.6.6 wheel (minimum supported 0.6.2)
• 6 code families × 5 decoders, all cross-compatible
• Batch decode: CPU/CUDA/OpenCL via CPUBatchDecoder/CUDABatchDecoder/OpenCLBatchDecoder (CUDA is shipped in the wheel; OpenCL requires a source build)
• Streaming decode: sliding-window FIFO with configurable window size, committed corrections, logical error rate tracking
• Spring-layout Tanner graph computation for all families

Honest performance posture

Following the upstream QECTOR Decoder v3 manual: all logical-error-rate, throughput, and latency figures are hardware-, driver-, seed-, and workload-dependent simulation results — regenerate them on your own target before quoting. PyMatching remains the speed leader on standard surface-code MWPM; QECTOR's exact BlossomDecoder reaches PyMatching's logical error rate but is not faster. The real strengths are batch throughput via approximate Union-Find (higher LER), qLDPC coverage via BP-OSD, and correctness that always satisfies the GF(2) syndrome equation H·c == s (mod 2). This is a research/evaluation platform, not a real-time or fault-tolerant hardware decoding stack.

Other

• Auto-updater checking PyPI for newer qector_decoder_v3 on boot
• Fonts: Consolas (mono), Segoe UI (body/heading)

License

Workbench (this app): source-available under EULA.txt. The EULA grants a royalty-free, worldwide license to use, execute, copy, and distribute the software for any purpose, including commercial, academic, and personal use — provided the embedded "QECTOR" notices and watermarks are retained (EULA §2).

Backend (qector-decoder-v3): a separately licensed, source-available Rust/Python platform by the same author. Its own license is free for personal, academic, educational, and non-commercial research, but commercial use of qector-decoder-v3 (company R&D, SaaS, hosted API, OEM, redistribution) requires a paid commercial license — see www.qector.store. This workbench depends on it at runtime; honor the backend's license for any commercial deployment.

Author: Guillaume Lessard / iD01t Productions © 2026

· ORCID 0009-0000-3465-3753

 · www.qector.store 

· admin@qector.store

Built on qector_decoder_v3 (v0.6.6).