Crabviz is a LSP-based call graph generator. It leverages the Language Server Protocol to generate interactive call graphs, helps you visually explore source code.

• Workable for various programming languages

• Highlight on click

• Two kinds of graphs

  You can generate a call graph for selected files to get an overview, or for a selected function to track the call hierarchy.

• Collapse Files to view file relationships

• Save call graphs as HTML or SVG

• Go to definition

• Search symbols

Since Crabviz utilizes the capabilities of language servers, it is better suited as an IDE/editor extension than a standalone command line tool.

It is currently available on VS Code, and PRs for other editors are welcome.

The repository contains an experimental CLI (@crabviz/cli) that can emit DOT/HTML/SVG graphs and a simplified interactive export similar to the VS Code extension.

Basic usage:

crabviz --roots path/to/project --out graph.html --renderer export --format html

Key flags:

• --call-depth N – call hierarchy depth (0=file-level only, 1=direct calls, 2+=multi-hop)
• --symbol-depth N – symbol-level depth control for detailed graphs.
• --ui-file – produce a file-level interactive HTML (extension-like look).
• --rankdir LR|TB – layout direction (Left-to-Right or Top-to-Bottom).
• --symbol-layout table|split – choose between table layout or split layout for symbols.
• --show-internal-file-calls – include internal calls within files in the graph.

To generate an interactive HTML graph with specific depth and layout:

$rootA='path/to/rootA'; $rootB='path/to/rootB'; $out='path/to/output.html';
node dist/cli.js --roots "$rootA" "$rootB" --out "$out" --renderer export --format html --ui-file --call-depth 2 --symbol-depth 3 --rankdir TB --symbol-layout table --show-internal-file-calls

# Static analyzer (default)
crabviz --roots myproj --out out.html --python-engine static

# Force legacy LSP behavior (pyright)
crabviz --roots myproj --out out.html --python-engine lsp

Set CRV_DEBUG=py to see analyzer/log debugging details.

1. Static Python import + intra-file call capture (done)
2. Integrate static call edges into symbol graph (in progress)
3. Cross-module call resolution & provenance tagging
4. Performance tuning & large project resilience

Limitations (current): Python function-to-function edges are partial; imported targets unresolved until Stage 3.

The following additional flags are available for advanced usage:

• --impl – Include interface implementation edges (default: true).
• --trim-last-depth – Trim the deepest collected call depth (one level before leaves) (default: false).
• --quiet – Suppress log/debug output (default: false).
• --files-per-row N – When --rankdir=TB, pack up to N file nodes per horizontal row within a folder (default: 0).
• --root-grid CxR – Arrange roots in a grid (e.g., 2x1 for 2 columns and 1 row) (default: undefined).
• --hide-imports – Hide import edges (show only call edges) (default: false).
• --dot-out path/to/file.dot – Write raw DOT graph to a specified file (useful for debugging).

These flags provide additional control over the graph generation process and are useful for debugging or customizing the output.

Crabviz is inspired by graphql-voyager and go-callvis.