docs: document frame metadata access in output converters

[bwsw]

Documents the optional metadata argument for model output converters, added in #1172 ("Support metadata passing to converters").

Change

docs/source/savant_101/30_dm.rst — adds an "Accessing Frame Metadata In A Converter" subsection to the detector-model chapter's Converter section, covering:

• How to opt in — declare a metadata: Optional[NvDsFrameMeta] = None parameter on the converter's __call__ (with a code snippet).
• What it exposes — source_id, pts, the underlying video_frame, the frame's objects, and tags.
• That it is opt-in and backward compatible: Savant inspects the converter's __call__ signature once (cached per converter class, no per-frame overhead) and passes metadata only when declared, so existing/built-in converters keep working unchanged. Also notes the nested convert-output telemetry span.
• A pointer to the output_converter_metadata sample and a cross-link to the Etcd dynamic-reconfiguration page.

Notes

• The API reference (reference/api/converter.rst) is auto-generated from docstrings, which Support metadata passing to converters #1172 already updated, so no manual change is needed there — the metadata parameter flows through automatically.
• RST validated: section underlines are well-formed, the :doc: Etcd target resolves, and the NvDsFrameMeta cross-reference uses the fully-qualified savant.deepstream.meta.frame path.

🤖 Generated with Claude Code

---

Generated by Claude Code

---

Note

Low Risk
Documentation-only change with no runtime or API behavior modifications.

Overview
Adds "Accessing Frame Metadata In A Converter" under the detector-model Converter section in 30_dm.rst, documenting the optional metadata argument on converter __call__ (from #1172).

The new text explains how to opt in with metadata: Optional[NvDsFrameMeta] = None, what fields are available (source_id, pts, video_frame, objects, tags), and that callers must handle None. It also notes opt-in / backward compatibility (signature inspected once per class, cached), nested convert-output telemetry when metadata is used, and links to the output_converter_metadata sample plus the Etcd dynamic-reconfiguration doc for per-source thresholds.

^{Reviewed by Cursor Bugbot for commit b9cf7b0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

Pull request overview

This PR extends the Savant 101 detector-model documentation to explain how output converters can optionally receive per-frame context via the new metadata argument, enabling context-aware post-processing while staying backward compatible.

Changes:

• Adds an “Accessing Frame Metadata In A Converter” subsection under the detector-model “Converter” docs.
• Documents how to opt in via a metadata: Optional[NvDsFrameMeta] = None parameter on __call__.
• Links to the output_converter_metadata sample and cross-references the Etcd dynamic-reconfiguration page.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.