Skip to main content
Scan Configuration controls how and when Earnie scans: which files a scan reads, how Earnie matches them, and what it writes on pull requests. It never changes a verdict, a gate, or a finding Earnie has already recorded. There are two Scan Configuration pages:
  • A project’s own page, Scan configuration in the project’s sidebar, for that project.
  • The organisation’s page, the Scan defaults tab in Settings, which applies to every project. It was called Scan Configuration before.
This page explains how the settings combine, and covers three sections in detail: Skip patterns, which keeps files out of a scan, Snippet tuning, which controls how Earnie fingerprints files, which weak matches it keeps, and how many components it reports per match, and Resolve likely non-matches, which settles snippet matches that are very likely wrong. Before you start: You need the Operator or Admin role to change Scan Configuration, except Resolve likely non-matches, which only an Admin can change. Viewers can read it. See Team & roles.

The Sections

Every setting is on one page, in the order Earnie acts on a change. A list down the left side names each section, follows you as you scroll, and jumps to a section when you pick it. On a project’s page, a strip at the top names the connected repository, with the option to disconnect it; disconnecting keeps scans, findings, and history. Each section shows when a change takes effect, for example Applies on the next scan. Component overrides, Snippet tuning, and Resolve likely non-matches appear only where open-source scanning is enabled, and AI provenance only where AI provenance is enabled.

Three Layers

Skip patterns, size rules, component overrides, and snippet tuning can each come from three layers:
  • The project’s Scan Configuration.
  • The repository’s own scanoss.json file, when it has one.
  • The organisation’s Scan Configuration.
How the layers combine depends on the setting: On a project’s page, the rules the repository’s scanoss.json and the organisation add sit under the project’s own, each labelled with where it comes from: Also skipped under the skip patterns and size rules, and Also applied under the component overrides. Use them to find where a rule came from when it isn’t in the project’s own list. They’re read-only here: Earnie never writes to scanoss.json, so to change those rules, edit the file in the repository, and change the organisation’s on its own Scan Configuration page. A rule in scanoss.json appears after the next scan reads the file.

Seeing Which Projects Differ

The last section of the organisation’s page, Projects, lists every active project in one table, so you can see which ones don’t follow the organisation without opening each. A value the project set itself is in normal text; an inherited one is muted. Under each project’s name, Earnie says why it differs. A project differs when it has its own pull-request feedback values, any rule of its own (skip patterns, size rules, component overrides or snippet tuning), a pre-commit hook set to Block the commit, automatic scanning paused while a repository is connected, or AI models fetched from Git LFS. A project with no repository doesn’t differ for that alone. Choose Only projects that differ to hide the rest; each choice shows how many projects it holds. The section is read-only: open a project to change its settings. Anyone who can read Scan Configuration, Viewers included, can see it.

Keeping Files Out of a Scan

Before fingerprinting, Earnie applies keep rules: path patterns and size limits that decide which files it scans. See How your source reaches Earnie for where this step fits in a scan.

Skip Patterns

A skip pattern excludes files or directories by path, one gitignore-style pattern per line, for example vendor/** or **/*.min.js. Earnie also follows the repository’s root .gitignore. In scanoss.json, the same patterns live under settings.skip.patterns.scanning.

Size Rules

A size rule excludes files by size. Use size rules for very large files that the built-in exclusions don’t already cover, because a large file can raise a scan’s memory use far above what a source file needs. Common AI model weight formats are already excluded. The Size rules table is under Skip patterns, on both pages. To add one:
  1. Select Add size rule, or open an existing rule.
  2. Under Patterns, enter one or more gitignore-style patterns the rule applies to, one per line.
  3. Set Skip files smaller than, Skip files larger than, or both, with a number and a unit: B, KB, MB, or GB. Sizes are binary, so 1 MB is 1024 KB.
Earnie skips a matching file when it’s smaller than the minimum or larger than the maximum. An empty bound means no limit on that side. A rule needs at least one pattern and at least one bound, and the minimum can’t be larger than the maximum. In scanoss.json, size rules live under settings.skip.sizes. Your project and organisation size rules are added to the repository’s own.

Built-In Exclusions

Even with no skip patterns of your own, the scanner skips a built-in list of paths on every scan. The list includes:
  • directories such as node_modules, vendor, venv, and __pycache__
  • documentation, data, and lock files, by extension, such as .md, .html, .csv, .yaml, .json, and .lock
  • files whose names end in words such as readme, license, changelog, or notice
  • build wrappers such as gradlew and mvnw
  • AI model weights, such as .safetensors, .gguf, and .onnx files
The scanner also skips hidden files, empty files, and symlinks. To see the full list, open Built-in exclusions, always skipped under Skip patterns. It lists the directories, file names, extensions, and name endings exactly as the scanner applies them. The list is read-only, and none of your patterns can bring one of these files back. The built-in list is why a scan’s Files scanned count can be lower than the number of files in your repository. The count includes only the files the scanner read after these exclusions.

Files Earnie Keeps Anyway

Some files are kept even when a skip pattern or a built-in exclusion matches them, because a scanner needs them:
  • Dependency manifests, such as package.json or pom.xml, so Earnie can read your declared dependencies.
  • Files in example and examples directories. The open-source scan skips them, but cryptography still inventories them.
  • Files a scanner reads for its own analysis, unless they’re inside a skipped directory such as node_modules.
Changing skip patterns or size rules triggers a full rescan. Earnie fingerprints your effective configuration, so that it doesn’t mistake a change in which files it keeps for a change in your code. After you change a skip pattern or size rule, Earnie plans the project’s next scan as full, not incremental.

Dependency Resolution

Dependency resolution decides how much of each manifest’s dependency tree a scan records. Leave a setting empty to inherit it. The project wins, then the organisation, then the default, and a project can enter 0 to lift a limit its organisation set. On a project’s page, each setting says where its value comes from, for example Inherited from the organization. Every package a scan records is checked for vulnerabilities and licences and evaluated by your policies, so a limit also limits those. Packages beyond the limit are left out of the scan’s dependency tree, findings, and SBOM. On the Dependencies page, the manifests the limit cut short read Partial, and the coverage line names the limit. Findings an earlier scan raised for those packages stay open: the scan didn’t look, so it can’t say they’re gone.
Changing dependency resolution makes the next scan run again, even for a commit or folder already scanned. Every change is recorded in the audit log.

Snippet Tuning

Snippet tuning controls how Earnie fingerprints your files, which weaker open-source matches it keeps, and how many components it reports per match. Apart from Components per match, it’s the same setting as settings.file_snippet in a repository’s scanoss.json, so you can set it once for many repositories instead of editing each file. Leave a setting empty, or on Inherit, to take it from the next layer. As in Three Layers, each setting is resolved on its own: the project wins, then the repository’s scanoss.json, then the organisation, then the SCANOSS default. Components per match is the exception: it has no key in scanoss.json, so only the project and the organisation decide it, in that order. On a project’s page, each setting shows its Effective value for the next scan and the layer it comes from, for example Effective: Keep matches ranked 3 or better · from scanoss.json.

Components Per Match

The same file is often vendored in more than one package, so SCANOSS can name several components for one match. Components per match sets how many it reports, best first.
  • A number or all. Enter a whole number of 1 or more, or all for no limit. The field refuses 0, a fraction, and any other word before you can save.
  • Whole-file matches. The number counts the strongest candidates. SCANOSS adds up to three more on top, called context picks, each labelled in Review with why it was picked: Best match in this folder, Best match for this file alone, or Hosted at this path. A label is a hint for the reviewer. It never changes a finding’s identity.
  • Snippet matches. The number is how many candidates SCANOSS reports for each matched block. The default is one. Snippet matches carry no labels.
  • Inherit. Leave the field empty to inherit. When no layer sets it, Earnie sends no limit, SCANOSS uses its own default, and your results don’t change. The project page shows SCANOSS default until a layer sets it.
  • The ranking threshold can hide context picks. A context pick isn’t one of the top-ranked candidates, so a ranking threshold can hide it.
Unlike the ranking threshold, this setting changes what Earnie finds, because SCANOSS answers with more or fewer candidates. Lowering the limit makes candidates that SCANOSS no longer returns absent on the next scan, including ones you already decided. Earnie keeps your decision, and the candidate returns with it if you raise the limit again. The field reminds you of this while you type a number below the limit in force.
Changing snippet tuning triggers a full rescan, as skip patterns and size rules do. Earnie compares the resolved values, so setting a value the last scan already used doesn’t.

Resolving Likely Non-Matches

After each scan, Earnie scores every open-source snippet match with its snippet classifier, which checks whether the lines reported in your file and in the open-source file correspond. Resolve likely non-matches lets Earnie act on that score: it resolves a snippet match as Not a match when the classifier reads it as not a match with at least the probability you set. Only an Admin can change this section. An Operator sees it read-only, with the note Only an admin can change this. A project can leave either value on inherit to take the organisation’s, and each value is resolved on its own: the project wins, then the organisation, then the default. There is no scanoss.json key for it.
  • Who decided it. Each match Earnie resolves this way shows Snippet classifier as who decided it.
  • Your decisions come first. Earnie never touches a match a person has already decided. Anyone with permission can reopen a match Earnie resolved, and a reopened match is never resolved this way again.
  • When it applies. A change to either value applies from the next scan, which runs again even for a commit or folder already scanned.
  • What it needs. The classifier needs the scanned files, so a scan that imports an SBOM, or that turns off the Snippet classification layer, resolves nothing this way.
A value saved during the preview now applies. If your organisation or a project turned this section on while it was a preview feature, and the preview was later turned off, the saved value stayed in place without effect. Now that the section is generally available, the next completed scan resolves likely non-matches. Before upgrading, check Settings → Scan defaults and each project’s Scan configuration, and turn the section off where nobody wants it.

What’s Next

To see how these rules fit into a run, go back to Your first scan. If the files you’re keeping out are model weights, see AI provenance.