* Build resolution picker HTML for a given graph bounds. * Disables options that exceed browser canvas limits.

Build prompt for metadata-only analysis. Args: files: List of FileStructureContext dicts with: - filePath: str -

Build the dynamic user prompt with metadata and code. Args: code: The code to analyze metadata: Source location metadata for node

* Extract and cache call graphs for all files in a content map.

* Calculate estimated cost for selected files

* Calculate text scale factor for large graphs * Scales down text when there are many nodes to keep export readable

* Close the file picker immediately (no animation)

* Collect initially selected paths and token counts from tree

* Compute workflows using LLM-assigned workflow names per file, * with hub detection for end-to-end shared service inclusion. * Falls back t

* Fallback: compute workflows from graph connectivity (original algorithm). * Used when no LLM workflow assignments are stored in cache.

* Condense raw repo structure into workflow-relevant summary. * Uses LLM to filter out irrelevant files and identify workflow entry points.

Condense raw repo structure into workflow-relevant summary. Takes tree-sitter extracted structure and returns a condensed version con

* Count total files in directory

* Count selected files in directory

* Helper to create a single node element with all its structure. * Extracted from renderNodes for reuse in incremental updates.

* Create a structural hash for change detection. * Only includes function names and call relationships, not line numbers.

* Create edges from title node to entry nodes

* Create a synthetic title node for a workflow

Detect if code contains LLM workflow patterns

Clean up all resources

* Expand detection to files that import base LLM files (cross-file analysis) * This catches files that call LLM wrappers indirectly (e.g., main.p

* Extract text with explicit hyphens from a rendered element. * Detects where the browser hyphenated and inserts actual hyphen characters. * Returns

* Filter tree based on search query

* Find connected components within a set of node IDs * Returns array of arrays, each inner array is a connected component

* Find entry nodes for a component (nodes with no incoming edges from within the component)

* Find node by path

* Find primary file for a workflow (file with most nodes)

* Find files related through imports/exports

Find connected components in a subgraph (undirected).

* Force immediate save of any pending changes. * Call this on extension deactivation to ensure cache is persisted.

* Generate minimap-scaled SVG path from ELK edge route

Generate metadata using a simple prompt (no workflow analysis). Used for incremental updates where we just need labels/descriptions.

* Get current batch stats for completion message.

* Synchronously get the last known merged graph (for instant feedback). * Returns null if no graph has been computed yet. * Use this for imm

* Get cached file data

Get the Language object for running queries

* Get the current workspace/codebase name from state