CodeGraph: v1.5.0 (latest, installed via npm i -g @colbymchenry/codegraph@latest)
Platform: [Windows]
I'm working with a large-scale multi-module codebase containing over 20,000 files (multiple languages, deeply nested directory structure). When running codegraph init (or codegraph init -i), the process experiences severe performance issues:
Extremely slow initialization – it takes an exceptionally long time to complete.
Frequent timeouts / freezing – the process often hangs indefinitely and eventually times out, especially when called via MCP clients (e.g., Claude Code) that have a default handshake timeout (~30s).
Unstable subsequent commands – even if init eventually finishes after a very long wait, subsequent ref (references) commands sometimes fail or return incomplete results, presumably due to incomplete/corrupted indexing.
Expected Behavior
For a 20k+ file repository, initialization should ideally:
Complete within a reasonable timeframe (e.g., < 60 minutes).
Avoid blocking the main MCP handshake, or support a "background indexing" mode that returns a partial graph immediately and updates it progressively.
Provide clear progress feedback (per-module or per-directory) so users know it's still alive.
CodeGraph: v1.5.0 (latest, installed via npm i -g @colbymchenry/codegraph@latest)
Platform: [Windows]
I'm working with a large-scale multi-module codebase containing over 20,000 files (multiple languages, deeply nested directory structure). When running codegraph init (or codegraph init -i), the process experiences severe performance issues:
Extremely slow initialization – it takes an exceptionally long time to complete.
Frequent timeouts / freezing – the process often hangs indefinitely and eventually times out, especially when called via MCP clients (e.g., Claude Code) that have a default handshake timeout (~30s).
Unstable subsequent commands – even if init eventually finishes after a very long wait, subsequent ref (references) commands sometimes fail or return incomplete results, presumably due to incomplete/corrupted indexing.
Expected Behavior
For a 20k+ file repository, initialization should ideally:
Complete within a reasonable timeframe (e.g., < 60 minutes).
Avoid blocking the main MCP handshake, or support a "background indexing" mode that returns a partial graph immediately and updates it progressively.
Provide clear progress feedback (per-module or per-directory) so users know it's still alive.