Struct WorkspaceScanner 

pub struct WorkspaceScanner {
    pub concurrency: usize,
}

Scans downstream workspaces for installed catalog assets.

Construct with WorkspaceScanner::new. The scanner is Clone + Send and safe to share across tokio tasks via Arc.

Fields

concurrency: usize

Maximum number of workspaces scanned concurrently.

Implementations

impl WorkspaceScanner

pub fn new(concurrency: usize) -> Self

Create a scanner with the given concurrency limit.

concurrency controls how many workspace scans run simultaneously via tokio tasks. The default recommended value is 8 (Req 23.1).

pub fn parse_export_metadata(content: &str) -> Option<ExportMeta>

Parse the # VFA-EXPORT: {json} metadata comment format (Req 7.7).

Searches content line-by-line (up to 20 lines) for a line whose trimmed form starts with # VFA-EXPORT: (case-sensitive). The JSON payload after the colon is parsed with serde_json.

Returns None for any malformed or missing line without panicking.

pub fn match_content_signature( first_50_lines: &str, canonical_first_50: &str, ) -> bool

Conservative content-similarity heuristic for the first 50 lines.

Heuristic: normalise both sides (trim whitespace, lowercase), split into a set of non-empty lines, and compute the Jaccard-like overlap:

overlap = |file_lines ∩ template_lines| / |template_lines|

Returns true when overlap >= 0.40 AND both sides have at least 3 non-empty lines. This threshold is deliberately conservative to avoid false positives on short boilerplate. Callers should treat this as a supporting signal only; confirmation requires a second agreeing strategy.

pub fn scan_harness_dir( &self, _workspace_path: &Path, harness: &HarnessDir, dir: &Path, index: &CatalogIndex, ) -> Vec<InstalledAsset>

Scan a single harness directory and return detected assets.

Per-harness file-extension filters (Req 7.8):

Harness	Extensions scanned
Claude	*.md
Cursor	*.json
Kiro	*.md
Codex	plugin.json only
Opencode	*.toml, *.yaml, *.yml

Validates the layout before walking (Req 7.6); returns empty vec with a warn! if the layout does not match any known pattern.

pub fn scan_workspace( &self, ws: &ResolvedWorkspace, index: &CatalogIndex, ) -> Vec<InstalledAsset>

Scan a single resolved workspace, returning all detected assets sorted by asset_id (Req 27.2).

Harness directories that do not match any known layout are warned and skipped (Req 7.6).

Bug #2: Recursively checks for agents/ and skills/ subdirectories within each harness root (e.g., .claude/agents/, .cursor/agents/, etc.).

pub async fn scan_all( &self, workspaces: &[ResolvedWorkspace], index: &CatalogIndex, ) -> HashMap<PathBuf, Vec<InstalledAsset>>

Scan all workspaces in parallel (up to self.concurrency simultaneous tasks) and return a map from workspace canonical path → installed assets.

Uses a Semaphore to bound concurrency (Req 23.1). Each workspace scan runs in its own tokio::spawn task so I/O does not block the executor.

Offline/unavailable workspaces are skipped silently here; callers should pre-filter or mark them based on ResolvedWorkspace::status.

Trait Implementations

impl Clone for WorkspaceScanner

fn clone(&self) -> WorkspaceScanner

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for WorkspaceScanner

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.