Show files tracked by standalone .dvc pointer files in directory listings and figures pages

[Copilot]

Files tracked via dvc add (standalone .dvc pointer files) were invisible in directory listings and figures auto-detection — only files in dvc.lock (pipeline outputs) were surfaced.

Changes

backend/app/projects.py — get_contents_from_tree

When building a directory listing, scan for .dvc pointer files among the git-tracked children. For each one (e.g., figures/plot.png.dvc), parse the pointer, derive the actual tracked path (figures/plot.png), and add it to the listing with storage="dvc" and size/type from the pointer metadata. Paths already covered by dvc_lock_outs are skipped to avoid duplicates.

# figures/plot.png.dvc is in git; figures/plot.png was previously invisible
dvc_pointer_outs: dict[str, dict] = {}
for p in paths:
    if not p.endswith(".dvc"):
        continue
    actual_path = p[:-4]
    if not actual_path or actual_path in dvc_lock_outs:
        continue
    dvc_file_data = yaml.safe_load(tree.read_text(p))
    ...
    dvc_pointer_outs[actual_path] = out

backend/app/api/routes/projects/core.py — get_project_figures

When traversing git blobs for figure auto-detection, also call _maybe_add_figure on the path derived from any .dvc blob (i.e., strip the .dvc suffix), so figures stored via dvc add are picked up alongside pipeline outputs from dvc.lock.

Tests

• Directory listing with a .dvc pointer for a file: verifies both the pointer (git) and the actual entry (dvc) appear with correct storage, size, and type.
• Directory listing with a .dvc pointer for a directory (md5 ending in .dir): verifies type="dir".
• Figures auto-detection via .dvc pointer blobs, including deduplication against dvc_lock_outs.

[Copilot]

Pull request overview

This PR makes files tracked via standalone DVC pointer files (*.dvc created by dvc add) visible to the backend’s directory listing logic and figure auto-detection, aligning behavior with existing support for pipeline outputs from dvc.lock.

Changes:

• Extend directory listings to detect *.dvc pointer files and surface the corresponding tracked paths as storage="dvc".
• Extend figure auto-detection to consider paths derived from *.dvc blobs.
• Add tests for directory listings and figure auto-detection involving *.dvc pointer files.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
backend/app/projects.py	Adds scanning/parsing of *.dvc pointer files during directory listing to surface derived DVC-tracked entries.
backend/app/api/routes/projects/core.py	Updates figure auto-detection to also consider paths derived from *.dvc blobs.
backend/app/tests/test_projects.py	Adds tests ensuring directory listings include both the pointer (git) and derived tracked path (dvc).
backend/app/tests/api/routes/projects/test_core.py	Adds tests for figure auto-detection via *.dvc pointer blobs and deduplication behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

out = outs[0] ... is stored into dvc_pointer_outs without validating it is a dict. If a malformed .dvc file has outs: [<non-mapping>], the later dvc_out.get(...) call will raise an AttributeError and break directory listing. Guard with isinstance(out, dict) (or coerce to {}) before storing.

Suggested change

	dvc_pointer_outs[actual_path] = out
	dvc_pointer_outs[actual_path] = out if isinstance(out, dict) else {}

[Copilot]

Figures auto-detection derives the “actual” path by stripping the .dvc suffix from the blob path, but the authoritative tracked path is inside the pointer YAML (outs[0].path). If a pointer file is renamed or its outs[0].path differs, this logic will miss (or mislabel) figures. Consider parsing the .dvc blob and building the derived path from outs[0].path relative to the pointer file’s directory (and skipping if it can’t be parsed).

[Copilot]

Current tests only cover the common case where the pointer filename matches the tracked output (e.g. plot.png.dvc -> plot.png). Since the implementation should ideally use outs[0].path, add a test case where the .dvc pointer filename is different from the tracked outs[0].path (renamed pointer) to prevent regressions.

[Copilot]

actual_path = p[:-4] assumes the pointer filename always matches the tracked output path. DVC’s canonical tracked path is outs[0].path; if the .dvc file is renamed or tracks a different path, this will add an incorrect entry to the listing. Consider deriving actual_path from the pointer’s outs[0].path joined to the pointer file’s directory, and only falling back to stripping the suffix when that field is missing.

Suggested change

	actual_path = p[:-4]
	if not actual_path or actual_path in dvc_lock_outs:
	continue
	try:
	dvc_file_data = yaml.safe_load(tree.read_text(p))
	if not isinstance(dvc_file_data, dict):
	continue
	outs = dvc_file_data.get("outs")
	out = outs[0] if isinstance(outs, list) and outs else {}
	try:
	dvc_file_data = yaml.safe_load(tree.read_text(p))
	if not isinstance(dvc_file_data, dict):
	continue
	outs = dvc_file_data.get("outs")
	out = outs[0] if isinstance(outs, list) and outs else {}
	out_path = out.get("path") if isinstance(out, dict) else None
	if isinstance(out_path, str) and out_path:
	actual_path = os.path.normpath(
	os.path.join(os.path.dirname(p), out_path)
	)
	else:
	actual_path = p[:-4]
	if not actual_path or actual_path in dvc_lock_outs:
	continue