feat: add graphify dry-run command

[nuthalapativarun]

Summary

Adds a graphify dry-run [path] CLI command that scans the corpus and prints a file-count/health summary without writing any output files or building the graph.

This is a safe preview step — useful for validating what graphify sees before committing to a full extraction run that may consume LLM tokens.

Usage

$ graphify dry-run ./my-project
Corpus scan: /abs/path/my-project

  Code files          23
  Documents            7
  Total               30  (~84,200 words)

Corpus looks healthy — no warnings.

No files were written. Run without dry-run to build the graph.

With a large corpus:

warning: Large corpus: 312 files · ~620,000 words. Semantic extraction
will be expensive (many Claude tokens). Consider running on a subfolder,
or use --no-semantic to run AST-only.

Implementation

• graphify/__main__.py — new elif cmd == "dry-run" branch + help text entry
• Reuses detect.detect() entirely — no new detection logic
• graphify-out/ is never created or touched

Test plan

• test_dry_run_prints_summary — file-count table appears in output
• test_dry_run_no_files_written — graphify-out/ is not created
• test_dry_run_default_path — defaults to current directory when path omitted
• test_dry_run_missing_path — exits non-zero for a missing path
• test_dry_run_no_graphify_out_written — "No files were written" in output

[nuthalapativarun]

Hey @safishamsi — just checking in on this one. Happy to rebase or make any adjustments if needed. Let me know!

[qodo-ai-reviewer]

Hi, graphify dry-run calls graphify.detect.detect(), but detect() can create graphify-out/converted/*.md sidecar files when it encounters .docx/.xlsx files. This violates the dry-run promise and can cause unexpected filesystem writes during what is advertised as a no-write preview step.

Severity: action required | Category: correctness

How to fix: Make detect side-effect-free

Agent prompt to fix - you can give this to your LLM of choice:

Issue description

graphify dry-run must not write any files, but it currently calls graphify.detect.detect() which may write office conversion sidecars into graphify-out/converted/.

Issue Context

• graphify/__main__.py dry-run branch calls _detect(root) and prints “No files were written”.
• graphify/detect.py converts .docx/.xlsx by writing markdown sidecars.

Fix Focus Areas

• Add a dry_run/write_sidecars/convert_office boolean parameter to graphify.detect.detect() (default preserving current behavior).
• Ensure that when the flag is disabled, detect() does not create directories or write any files (skip conversion, and optionally count words directly from the office file or as 0).
• Call detect(..., write_sidecars=False) (or equivalent) from the dry-run CLI branch.

References

• graphify/main.py[794-823]
• graphify/detect.py[347-376]
• graphify/detect.py[187-213]

---

Found by Qodo code review

[nuthalapativarun]

Good catch @qodo-ai-reviewer — fixed in 0b3e6eb.

detect() now accepts a write_sidecars=False keyword argument. When disabled, office files (.docx/.xlsx) are counted directly without calling convert_office_file() or touching graphify-out/converted/. The dry-run CLI branch passes this flag, so the no-write promise holds even for corpora containing office files.

Added test_dry_run_office_no_sidecar_written which mocks convert_office_file and asserts it is never called during a dry-run invocation.