Settings

Settings & Preferences¶

Access the Settings menu via the gear icon in the main toolbar (or press Ctrl + Shift + S).

The Settings Panel is organized into two tabs:

General — fitting behavior and file locations: Fit Parameters, Global Peak Limits, and the SPECTROview Working Folder (sections 1–3 below).
AI — API keys and the chat-history folder for the AI Chat Agent (section 4 below).

1. Fit Parameters (General tab)¶

These parameters control the behavior of the Vectorized Batch Fit (VBF) engine — the mathematical optimizer that performs curve fitting on your spectra. Adjusting these values directly affects fitting speed, precision, and convergence.

Parameter	Setting	Default	Range	Description & Behavior
Maximum Iterations	`max_ite`	200	1 – 10,000	The maximum number of Levenberg-Marquardt iterations per spectrum. • Lower (50–100): Faster fitting for previews. • Higher (200+): Better convergence for complex spectra.
x-tolerance	`xtol`	1×10⁻⁴	1×10⁻⁶ – 1×10⁻¹	Relative tolerance for the parameter step size. Convergence requires: mean(\|δp\| / \|p\|) < `xtol`
f-tolerance	`ftol`	1×10⁻⁴	1×10⁻⁶ – 1×10⁻¹	Relative tolerance for the cost function. Convergence requires: \|cost_n - cost_n-1\| / \|cost_n\| < `ftol`
Noise Threshold Coefficient	`coef_noise`	1.0	0 – 100	Multiplier for the auto-estimated noise level (`coef_noise` × estimated noise amplitude). Masks noisy data and suppresses ghost peaks. • 0: Disabled • 0.5–1.0: Conservative • 1.0–2.0: Moderate (Default) • 3.0+: Aggressive Note: You can refer to the noise level displayed in the toolbar or activate the "Show noise level" checkbox within the `View Options` of the `SpectraViewer` to see the noise level directly on the spectra plot.
Fit Negative Values	-	Unchecked	-	When checked, negative intensity values are included in the fit. When unchecked, they are assigned zero weight.

Note: A spectrum is considered converged only when both xtol and ftol criteria are simultaneously satisfied.

2. Global Peak Limits (General tab)¶

These parameters define absolute boundary conditions that are applied globally to all peaks across the application. When you add a new peak to a fit model, these limits are used to set the initial parameter bounds automatically.

Note: You can check the "Limits" checkbox under the PeakTable to display the actual bounds of each peak.

Limit	Default	Range	Description & Constraint
Max Peak Shift	20	0 – 100	Maximum allowed displacement of a peak center (`x0`) from its initial value. x₀ - maxshift ≤ x₀^fitted ≤ x₀ + maxshift
Min Peak FWHM	0.1	0.01 – 500	Minimum allowed Full Width at Half Maximum (`FWHM`). Prevents peaks from collapsing into delta-function artifacts.
Max Peak FWHM	200	1 – 1,000	Maximum allowed `FWHM`. Prevents peaks from broadening into flat baselines.
Max Peak Intensity	100,000	1 – 1×10⁹	Upper bound for peak amplitude. Prevents unrealistically high intensity values.

3. SPECTROview Working Folder (General tab)¶

Specify a single root folder for all of SPECTROview's saved templates. When you set it, three subfolders are created automatically inside it:

Subfolder	Contents
`fit_model/`	Custom fit models (JSON). The `Fit Model Builder` scans this folder and lists every model here for one-click reuse.
`plot_recipe/`	Plot Recipes — saved sets of full plot configurations (data bindings + style) for the `Graphs` workspace.
`plot_style/`	Style Templates — saved appearance-only styles that can be applied to any existing graph.

Click Browse to select a folder, or type the path directly.

Note: This single Working Folder replaces the older separate "Fit model folder" / "Plot template folder" settings. If you had one of those configured previously, it is migrated automatically the first time the Working Folder is read.

4. AI Settings (AI tab)¶

The AI tab configures the AI Chat Agent. Nothing here is required to use the rest of SPECTROview.

API Keys¶

The fields for a Custom (OpenAI-compatible) endpoint are shown by default:

Field	Description
Custom API Key	The API key for your custom / internal endpoint.
Base URL	The endpoint's OpenAI-compatible root URL (e.g. `https://host/v1`).
Model Name	A comma-separated list of model names (e.g. `model-a, model-b`). These populate the model dropdown in the chat panel — handy for endpoints that don't expose a model-listing API.

Keys for the built-in cloud providers are tucked under the collapsible ▸ Other cloud providers section (click to expand): OpenAI, Anthropic, Gemini, DeepSeek, and Mistral. All keys are stored masked and remembered between sessions.

Where your keys are stored. API keys go into your operating system's per-user settings store — the registry (HKEY_CURRENT_USER\Software\CEA-Leti\SPECTROview) on Windows, ~/Library/Preferences on macOS. They are never written to a file inside the SPECTROview project folder, are not included in an exported workspace, and cannot be committed to a Git repository by accident. They are stored in plain text there, so treat them like any other saved credential: they are readable by anything running under your user account.

Chat History¶

Field	Description
History Folder	Where AI chat conversations are saved (as JSON). Click Browse to choose a folder.