Feature Reference

The 5-Step Guided Workflow

Every Vizalyze project follows the same ribbon-based flow. Each step is accessible at any time - click the ribbon button to jump directly. The app keeps state across steps and highlights the active phase.

Load & Import

Vizalyze's import engine handles 31+ file formats with intelligent preamble detection, unit extraction, CAN bus decoding, locale-aware decimal/date parsing, and lazy loading for fast previews on large files.

Import Template Builder

Create reusable import configurations from Tools → Data → Import Template Builder. Templates capture:

• Preamble parsing rules (rows, regex, delimiter modes)
• Header-row detection and unit extraction (metadata vs row-based)
• Channel cleanup rules (regex replacements, per format)
• File-grouping rules and CAN/LIN database assignments

Templates are saved as JSON in Import_Templates/ and can be applied per-file, per extension, or as a global default. Share templates across machines by exporting the JSON.

Advanced Import Utility

Open from Tools → Data → Advanced Import Utility for per-file import tuning without saving a template - preamble modes, delimiter detection, column cleanup, and live preview on messy lab exports.

Preamble Parsing

Many test data files contain header metadata before the actual data table. Vizalyze auto-detects the header row and offers 5 preamble modes:

• Auto - automatically detects format (recommended)
• Key-value - parses Key: Value pairs
• Delimited - fixed delimiter (comma, semicolon, tab)
• Regex - custom pattern per line
• Raw - skip N rows, take the rest as-is

Vendor Format Auto-Detection

Vizalyze uses a two-pass content sniffer to identify file formats from their actual bytes, independent of file extension. This handles mislabelled files, non-standard extensions, and vendor exports that look like plain CSV or binary blobs.

• Pass 1 - Binary magic bytes: HDF5, Apache Parquet, Arrow/Feather, SQLite, PDF, Excel OLE2/OOXML, MATLAB workspace, ASAM MDF, Vector BLF, NI TDMS
• Pass 2 - Text header fingerprints: Gamry Framework DTA (EXPLAIN header), Zahner Zennium / IM6 EIS data, Bronkhorst FlowView, Siemens WinCC / SIMATIC, Mettler-Toledo LabX, Metrohm MagIC Net, Dewesoft ASCII export, NI LabVIEW LVM

When the sniffer detects a format with high or medium confidence, the loader routes to the correct parser and the Advanced Import wizard pre-fills delimiter, mode, and vendor hints automatically.

Excel & PDF Import

For Excel files: configure sheet policy (first / all / prompt), and use the range suffix syntax file.xlsx:::SheetName:::B5:Z5000 for precise cell ranges.

For PDF files: Vizalyze uses pdfplumber to extract tables. A region picker dialog lets you select the exact table area on the page.

Lazy Loading & Performance

Large CSV files are previewed using Polars scan_csv() (lazy evaluation) - only the first 2,000 rows load for the preview. Full load happens on demand. Loaded files are automatically cached as Parquet for near-instant re-open.

Data Health Panel

The Health panel runs an automatic quality audit on every loaded dataset and produces a compact, actionable summary. It's the first thing to check after loading new data.

What it checks

• Data completeness - percentage of non-null values per channel
• Constant channels - channels with zero variance flagged as potential sensor faults
• Timestamp conflicts - monotonicity checks, duplicate timestamps, irregular intervals
• Unit consistency - detects channels in different unit families mixed together
• Sampling rate - estimated from time column, checked for consistency

Status badges

Each channel gets a badge. The overall dataset badge is the worst of all channels. Clicking a badge jumps to the recommended fix in the Compare or Analysis panel.

Recommended actions

The panel maps each issue to a concrete next step - e.g., "3 channels have >5% missing data → open Compare Runs and enable interpolation" or "Timestamp duplicates detected → use step alignment mode".

Compare Runs

Overlay any combination of loaded datasets on a single chart, automatically aligned to a common time base. Works across different file formats and sampling rates.

Alignment modes

Delta table

Switch to Delta Table view to see the numerical difference between runs for every channel. Shows mean, max deviation, and a pass/fail indicator against user-set tolerances.

Y-offset tools

Per-run Y-offset: zero at min, max, mean, start, or end of each channel. Useful for removing DC offset before shape comparison.

Missing channel warnings

If a channel exists in some runs but not others, the panel shows a warning with options to skip, fill with NaN, or use a default value.

Analysis Panel

Go beyond charts with computed statistics, correlation matrices, domain KPIs, and limit evaluation - all driven by 8 selectable domain profiles.

Statistics

Per-channel: min, max, mean, std, median, P5, P95, skewness, kurtosis. Supports up to 240 channels and 200k rows in a single analysis. Export to CSV or directly into the report template.

Correlation analysis

Pearson correlation matrix for up to 16 channels at once. Highlights strong positive/negative correlations above a configurable threshold. Useful for identifying redundant sensors or causal relationships.

Domain profiles

Select a domain profile to get pre-loaded KPI recommendations, suggested analysis sections, and appropriate limit defaults:

• Generic - universal statistics, no domain assumptions
• Fuel Cell - stack voltage, current density, power density, efficiency KPIs
• Battery - SOC, capacity, C-rate, temperature gradient KPIs
• Thermal - heat flux, temperature uniformity, thermal resistance
• Vehicle - acceleration, braking, energy consumption per km
• Emissions - NOx, CO, HC, PM cycle-averaged values
• Pharma HPLC - retention time, peak area, resolution, tailing factor
• Pharma GC / MS - TIC, BPI, spectral purity, fragment ion ratios

Limit evaluation

Define upper/lower limits per channel. The analysis panel evaluates each channel against its limits and assigns Pass / Warn / Fail badges. Limit definitions can be saved to recipes for reuse.

Unit detection & conversion

The Units tab detects engineering units from channel names and metadata, then suggests affine conversions so channels can be compared fairly. The full Unit Library (Tools → Units → Engineering Unit Library) covers 30+ families and 200+ built-in units.

Batch Reporting

Generate professional PPTX, Word, or PDF reports from reusable templates. Batch-process entire folders. Save pipeline recipes for one-click re-runs on new data.

Power Mode

Sandboxed Python against the active canvas dataset - plot scripting, dataframe transforms, and inline stats from the Analysis Dashboard.

Report Template Builder

Accessed from the Report panel or Tools → Report → Template Builder. Design templates with:

• Text blocks with dynamic placeholders (channel values, statistics, localized date/time)
• Auto-populated data tables (stats, limits, delta)
• Chart references (any chart in the current session)
• KPI cards with conditional pass/warn/fail colouring
• ISO 17025-style metadata fields (operator, instrument, calibration date)
• Computed columns and formulas

Templates are saved as .dpt files. Vizalyze ships with three ready-to-use templates for fuel cell stack analysis.

Pipeline DAG & Recipes

The pipeline view lets you wire up processing steps in dependency order:

health_check → compare_runs → limit_eval → analyze → auto_summary → export_pptx

Steps run in topological order. Save the full pipeline as a recipe JSON in Recipes/. Load any recipe and re-run it on a new batch of files - ideal for repeated test campaigns.

Workflow Library

Save AI-assisted or manual analysis sequences as named workflows. Browse, duplicate, export as .vzwf, and replay on new data from Tools → Workflow → Workflow Library.

Auto-summary

The "Auto Summary" feature merges findings from the Data Health and Compare Runs panels into a single narrative section that goes into the report automatically.

7 built-in presets

• Durability report
• Run comparison report
• Sensor drift report
• Step test report
• Limit check report
• Transient response report
• Efficiency analysis report

Chart Types

17 chart types powered by Pyqtgraph with OpenGL acceleration. Every chart is interactive: zoom, pan, hover tooltips, cursor measurement, right-click styling, and legend toggling.

Chart interaction features

• Hover highlighting - active series brightens on mouse-over, others dim
• Click selection - click a series to select it; stats appear in the status bar
• Inspect tool (A/B cursors) - shaded measuring range with A/B handles; live ΔY / ΔX / 1/ΔX readout and min/avg/max/P2P/std for every plotted channel
• Export - PNG, PDF, SVG from right-click context menu; Copy Chart Image (Ctrl+Shift+C) for PowerPoint / Teams
• Legend - draggable; click series name to toggle visibility
• Per-series styling - colour, line width, dash pattern, marker shape
• Alert bands - define horizontal limit zones with colour fill

Performance

Vizalyze renders 1M+ data points instantly using LTTB (Largest Triangle Three Buckets) peak-preserving downsampling. In Auto mode, downsampling activates automatically for dense signals. In Quality mode, full resolution is kept. See Performance settings below.

Performance & Cache

Vizalyze is engineered for speed on large datasets. Three performance modes let you choose the balance between quality and responsiveness.

Performance modes

• Auto (recommended) - activates optimisations adaptively based on data size
• Quality - preserves full resolution; may be slower on 1M+ point datasets
• Maximum Speed - aggressive optimisations: heavier downsampling, smaller 3D point caps, disabled antialiasing

Auto plot optimisations

• LTTB peak-preserving downsampling (configurable on/off)
• Clip-to-view: only renders points in the current viewport
• Adaptive antialiasing: disables on dense plots, re-enables after zoom
• Marker & label caps: limits number of decorations rendered at once
• 3D scatter point caps: 500k (Quality) / 150k (Auto) / 50k (Max Speed)
• Hover tracking throttle on dense plots

Parquet cache

On first load, CSV and other text formats are automatically converted to Parquet and stored in the Sessions/ folder. Re-opening a cached file is near-instant (Polars lazy scan vs full text parse). Cache is LRU-evicted to manage disk usage.

Performance benchmarks

Diagnostics

Export a performance log from Help → Diagnostics → Export Performance Log. The log captures timing for startup, file loads, chart redraws, and catalog searches - useful for diagnosing slowdowns on specific hardware.

Enable verbose tracing with the environment variable DPT_PERF_TRACE=1.

Signal Processing

100+ signal processing functions available from the Analysis panel, Code panel, and right-click chart menu. Every result is a computed column with full lineage tracking.

Frequency analysis

• FFT - Fast Fourier Transform with Hann, Hamming, or Blackman windowing
• PSD - Power Spectral Density using Welch's method
• Spectrogram (STFT) - Short-Time Fourier Transform with log-scaled power display

Digital filtering

• Filter families: Butterworth, Chebyshev I, Chebyshev II, Bessel, Elliptic
• Filter types: Low-pass, High-pass, Band-pass, Band-stop
• Phase mode: Zero-phase forward-backward filtering (no phase delay)
• Order: 1 to 10, configurable

Smoothing

• Moving average (simple, exponential)
• Median filter
• Savitzky-Golay (polynomial smoothing)

Transforms

• Resampling - interpolate to a uniform time grid at any target rate
• Differentiation - numerical first and second derivatives
• Integration - cumulative numerical integration (trapezoid rule)
• Unit conversion - vectorised affine transforms (~300× faster than row-by-row)

Event detection

• Peak detection - configurable minimum height and distance
• Anomaly detection - statistical outlier flagging (z-score, IQR)
• Step detection - identify step changes and segment by state

NVH & durability

• 1/3-Octave bands - IEC 61260 fractional-octave levels with A/C/Z weighting and optional 20 µPa SPL reference
• Rainflow counting - ASTM E1049 cycle ranges and histogram for fatigue / S-N damage models
• Order analysis - computed order tracking via angular resampling against an RPM channel

Pivot / Group-By

Summarise any loaded dataset like an Excel pivot table: group by mode/step columns, aggregate channels (mean/min/max/sum/std/…), then copy as TSV or export. Open from Tools → Data → Pivot / Group-By Summary or the command palette.

Code Panel

Write Python-syntax expressions to add computed columns or filter rows - directly in the app, with AI autocomplete and inline error messages.

Computed columns

# Example: compute power from voltage and current
Power_kW = (HV_Voltage * HV_Current) / 1000

# Example: efficiency ratio
Efficiency = Motor_Power_out / Motor_Power_in

Column names auto-complete as you type. Available functions include Polars expressions, numpy operations, and all custom transforms.

Filter conditions

# Keep only rows where vehicle is moving above 5 km/h
Vehicle_Speed > 5

# Keep rows in a specific temperature range
(Battery_Temp >= 15) & (Battery_Temp <= 40)

Filter conditions create a new derived dataset - the original data is untouched (immutable lineage). Toggle the filter on/off in the dataset tree.

AI assistance

Click the AI button (or type /ai) in the code panel to describe what you want in plain English. The AI generates the expression using your actual column names and explains its reasoning.

Metadata Catalog

An FTS5 SQL-backed searchable index of all files Vizalyze has ever seen. Stays fast with 50,000+ indexed files.

What gets indexed

• File path, format, size, modification date
• Row count, channel count, time column name
• Column names (all, searchable as full-text)
• Health status (Pass/Warn/Fail)
• Test ID and run ID (parsed from preamble)

Search filters

• Full-text search across column names and file paths
• Format filter (by extension)
• Date range picker
• Health status filter
• "Discovered only" - files seen but not loaded

Double-click any result to load it into the active session. The catalog updates in the background when the working folder changes. Files whose mtime hasn't changed reuse cached metadata (no re-read needed).

Data Lineage & Provenance

Every load, filter, merge, transform, and computed column is tracked in an immutable dependency graph.

Data Lineage Panel

Visualise the full transformation DAG - see exactly how each dataset was derived. Each node shows the operation, timestamp, and operator (on Team tier).

Rollback

Revert to any prior state in the workflow by clicking a node in the lineage graph. The original file is never modified.

Audit log (Team+)

Append-only, immutable audit log records every action: who loaded which file, when, what transforms were applied, and what was exported. Meets requirements for regulated environments.

Automotive & CAN/LIN

Native support for ASAM MDF files and CAN/LIN bus decoding. The Automotive & EV extension pack includes:

Supported MDF formats

• .mf4 - ASAM MDF4 (all versions, current standard)
• .mf3 - ASAM MDF3 (legacy)
• .mf2 - ASAM MDF2 (legacy)
• .atfx - ASAM ODS Transfer Format

CAN Database Manager

Open from Tools → Data → CAN Database Manager. Features:

• Persistent library of DBC, ARXML, KCD, SYM database files
• Browse message/signal trees with IDs, units, min/max, factor, offset, comments
• Free-text search across all signal names and descriptions
• Assign a database to a specific file, or set as the global default
• 6 built-in EV signal presets (see below)

EV signal presets

Automatic CAN decoding

When a database is assigned (via the CAN Database Manager or an import template), CAN/LIN signals are decoded automatically on load. MDF/MF4 files use asammdf.MDF.extract_bus_logging(); BLF files use python-can + cantools. Physical signals (with scaling applied) appear directly as columns in both cases.

BLF (Vector Binary Log) support - new in v1.0.7

Vector .blf captures are now a first-class format. Drop a BLF file into the Files panel like any other source. Two modes:

• Without DBC assigned - raw frame inspection: timestamp, arbitration_id, dlc, data_hex. Useful to verify what message IDs are present before assigning a database.
• With DBC/ARXML assigned - full signal decode: open the CAN Database Manager (Tools - Data - CAN Database Manager), pick your BLF file under Bus Log Assignments, assign the matching database, then reload. Every signal appears as its own named column with physical units.

The same DBC library and assignment workflow used for MDF/MF4 files applies unchanged to BLF - no extra setup required if you already have databases configured.

Export Dataset As - new in v1.0.7

Any loaded or processed dataset can be exported to a portable format via Tools - Data - Export Dataset As.... This works on all ingested formats: BLF decoded signals, MDF/MF4, TDMS, HDF5, NetCDF, and all others.

• CSV / TSV - universally readable, works in Excel, Python, R
• Excel (.xlsx) - formatted workbook for sharing with stakeholders
• Parquet - compressed columnar format for large datasets (zstd compression)
• Feather / Arrow IPC - zero-copy interchange with pandas, polars, R Arrow
• JSON - row-oriented JSON for web and API consumers

Pharma & Analytical Chemistry

The Pharma extension pack provides native support for the most common analytical chemistry data formats, plus domain-specific analysis profiles.

Supported formats

• .cdf - AIA/ANDI CDF chromatography data (netCDF-based). Columns: retention_time, signal. Supports HPLC, GC, IC.
• .mzml - mzML mass spectrometry. Extracts TIC (Total Ion Current), BPI (Base Peak Intensity), and per-scan summaries. Requires pymzml.
• .jdx / .dx - JCAMP-DX spectra and chromatograms. Supports NMR, IR, UV-Vis, Raman, mass spectra. Columns: numeric X/Y.
• .raw - Thermo Fisher RAW files. Requires external ThermoRawFileParser converter (free, open-source).
• .wiff - Sciex WIFF files. Requires external ProteoWizard msconvert (free).

Pharma domain profiles

• Pharma HPLC - retention time, peak area, resolution, symmetry factor, tailing factor, theoretical plates
• Pharma GC - peak area %, relative retention time, split ratio diagnostics
• Pharma MS - TIC, BPI, spectral purity, fragment ion ratios, m/z ranges

ISO 17025-style metadata and traceability fields

Report templates include ISO 17025-style metadata fields: analyst name, instrument serial, calibration date, method reference, and sample ID. These populate automatically from preamble metadata when detected.

File-based live tailing & SPC

Tail measurement files as they grow on disk (not a hardware DAQ stream). Mark any supported file as a live source — Vizalyze polls for appends and updates the chart automatically.

Setup

1. Right-click a dataset in the file tree → Mark as live source
2. Set the poll interval in Settings → Performance & Cache (default: 2.0s; range 250ms–10s)
3. The chart and statistics refresh when new rows are appended to the file

Live SPC panel

• Control-limit style monitoring (UCL, LCL, warning limits)
• Real-time plot & statistics refresh
• Drift detection and out-of-control rule alerts
• Alarm annotations on the chart

AI Integration

Vizalyze has a built-in AI chat panel and AI-assisted features throughout the app. All AI calls go directly to your chosen provider - Vizalyze is never the intermediary.

Built-in AI chat panel

Ask questions about your loaded data in plain English. The AI has access to your column names, statistics, and health report. Example queries:

• "Which channel shows the highest correlation with Battery_Temp?"
• "Write a Butterworth low-pass filter at 5 Hz for the Vibration_Z channel"
• "Summarise the key findings from the data health report"
• "Generate a report recipe that compares Run_001 and Run_002 on speed and SOC"

AI-assisted features

• Code panel - describe a computed column or filter condition in English; AI generates the Polars expression
• Chart styling - "make the speed line red and the SOC line dashed green"
• Channel recommendations - AI suggests which channels to plot based on data health findings
• Formula autocomplete - intelligent suggestion of channel names and operators as you type

Supported providers

• Claude (Anthropic) - recommended for analytical reasoning
• GPT-4 (OpenAI) - broad capability
• Gemini (Google) - fast responses
• Ollama - fully local, no data leaves the machine
• LM Studio - fully local with a GUI model manager

Configure your API key in Control Settings → AI. For local models, set the server URL (default http://localhost:11434 for Ollama).

AI Environments (Pro+)

Save and restore AI conversation context to the cloud. Switch machines and pick up exactly where you left off - the AI remembers your project context, preferred channel names, and analysis conventions.

Sessions

Sessions save the complete state of a Vizalyze project: loaded files, charts, filters, computed columns, compare settings, analysis results, and AI conversation context.

Save & load

Sessions are stored in Sessions/ inside your working folder. Use File → Save Session or Ctrl+S. On startup, Vizalyze can auto-restore the last session (configurable: Always / Ask / Never).

Session limits

• Free - local sessions only, unlimited
• Pro - 25 cloud-synced sessions
• Team - 200 cloud-synced sessions

Cloud Features (Pro+)

Desktop vs website: Analysis always runs in the installed app. Cloud features sync settings from the desktop and expose a few read-only experiences on vizalyze.app (account dashboard, shared report viewer, live monitor). Raw measurement files are not uploaded for analysis in the browser.

Encrypted cloud sync

Workflows, sessions, templates, and channel configs are encrypted with AES-256-GCM before leaving your machine. Your encryption key is derived from your credentials - Vizalyze's servers never have access to your content.

Secure report sharing

From the desktop app, publish a report and generate a share link. Recipients open an interactive read-only view in their browser at vizalyze.app - no desktop install required. Links can be set to expire.

Live monitor dashboard (web)

Pro+ users can open a browser dashboard on vizalyze.app to watch live statistics and sparklines for files the desktop app is monitoring - a companion view, not a substitute for the desktop Visual Workspace.

Org workspace (Team+)

Shared workspace with org-scoped template library, channel config library, and CAN database library. All members see the same shared resources.

Settings & Configuration

Access all settings from the Control Settings step in the ribbon, or from the ⚙ icon in the top-right.

Machine-level settings

Stored in ~/.Vizalyze/app_settings.json. This includes your working folder path, performance mode, and API keys (encrypted).

Environment variables

• DPT_PERF_TRACE=1 - verbose performance tracing to stderr
• VIZALYZE_DEV=1 - enable developer mode (unlocks all tiers locally)
• QT_QPA_PLATFORM=offscreen - headless mode for CI/automated testing