Running graphify . on a large mixed code/docs repo on Windows (PowerShell), almost every semantic extraction chunk logs:
[graphify] incremental cache checkpoint failed: argument should be a str or an os.PathLike object where fspath returns a str, not 'FileSlice'
This appears right after each [graphify extract] chunk N/66 done line, suggesting the per-chunk incremental checkpoint (introduced in #1715) is being passed an internal FileSlice object instead of a str/Path.
Environment:
graphify 0.9.15 (latest, confirmed via graphify --version)
Installed via uv tool install graphifyy (confirmed via uv tool list → graphifyy v0.9.15 exposing graphify and graphify-mcp)
Windows, PowerShell
Repo: ~705 code files, 1579 docs, 5 images (~2300 files total), run as graphify .
Repro:
powershellgraphify .
on a repo of this size/mix. Error fires on ~every chunk of the semantic extraction phase (chunks 1-17, 60-66 in my run all hit it).
Impact:
Final graph.json write still completes, so the run isn't fully blocked.
Since the checkpoint write fails every time, the incremental-resume mechanism from #1715 doesn't appear functional — a rerun after an interruption (e.g. API credits running out mid-run, which also happened here) likely can't resume from partial progress and may re-process already-completed chunks.
Workaround: setting GRAPHIFY_NO_INCREMENTAL_CACHE=1 should bypass the checkpoint (untested against this specific error, but matches the documented opt-out).
Running graphify . on a large mixed code/docs repo on Windows (PowerShell), almost every semantic extraction chunk logs:
[graphify] incremental cache checkpoint failed: argument should be a str or an os.PathLike object where fspath returns a str, not 'FileSlice'
This appears right after each [graphify extract] chunk N/66 done line, suggesting the per-chunk incremental checkpoint (introduced in #1715) is being passed an internal FileSlice object instead of a str/Path.
Environment:
graphify 0.9.15 (latest, confirmed via graphify --version)
Installed via uv tool install graphifyy (confirmed via uv tool list → graphifyy v0.9.15 exposing graphify and graphify-mcp)
Windows, PowerShell
Repo: ~705 code files, 1579 docs, 5 images (~2300 files total), run as graphify .
Repro:
powershellgraphify .
on a repo of this size/mix. Error fires on ~every chunk of the semantic extraction phase (chunks 1-17, 60-66 in my run all hit it).
Impact:
Final graph.json write still completes, so the run isn't fully blocked.
Since the checkpoint write fails every time, the incremental-resume mechanism from #1715 doesn't appear functional — a rerun after an interruption (e.g. API credits running out mid-run, which also happened here) likely can't resume from partial progress and may re-process already-completed chunks.
Workaround: setting GRAPHIFY_NO_INCREMENTAL_CACHE=1 should bypass the checkpoint (untested against this specific error, but matches the documented opt-out).