Skip to content

docs(api): document GraphIngestor and ServiceIngestor SDK contract (NVBug 6628849) - #2577

Draft
kheiss-uwzoo wants to merge 1 commit into
NVIDIA:mainfrom
kheiss-uwzoo:docs/6628849-api-reference
Draft

docs(api): document GraphIngestor and ServiceIngestor SDK contract (NVBug 6628849)#2577
kheiss-uwzoo wants to merge 1 commit into
NVIDIA:mainfrom
kheiss-uwzoo:docs/6628849-api-reference

Conversation

@kheiss-uwzoo

@kheiss-uwzoo kheiss-uwzoo commented Aug 24, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Document create_ingestor() return types by run mode: GraphIngestor for inprocess and batch, ServiceIngestor for service.
  • Generate API entries for the concrete classes and 26.08 generation and answer types instead of the stub Ingestor interface.
  • Align the document-ingestion workflow page with the same factory contract.

Fixes NVBug 6628849.

Test plan

  • mkdocs build --strict on this diff (local leftover untracked pages are not in this change)
  • Confirm published API Reference headings for GraphIngestor, ServiceIngestor, Retriever.answer() types, and generation operators
  • Confirm create_ingestor() is no longer described as graph-only on the published page

pre-draft: leakage, mkdocs --strict, ::a, ::p, ::r on the diff vs main

Base: upstream/main
Files: docs/docs/extraction/nemo-retriever-api-reference.md, docs/docs/extraction/workflow-document-ingestion.md

Check Result
Leakage (page roles + see [ CTAs) PASS — no see [ CTAs; faq/overview/multimodal-extraction nimOperator/nvcr.io/nim/installFfmpeg unchanged
Allowed paths PASS — 2 documentation files
mkdocs --strict PASS — exit 0 for this diff; local leftover untracked pages fail strict and are not in this change
::a audit PASS — factory return types and listed methods match RC6 source (ingestor/core.py, graph_ingestor.py, service_ingestor.py, models/llm, operators/generation); generated create_ingestor summary hidden because source docstring is still graph-only
::p polish Applied — factory table, import paths, and generated class sections; backticked GraphIngestor in method list
::r style 95% — no blocking issues; remaining source docstring is code drift, not this docs PR

Code drift (not in this docs PR): nemo_retriever/src/nemo_retriever/ingestor/core.py create_ingestor() docstring still says Graph-only ingestion factory, and the annotation still returns Ingestor.

PR: #2577

…VBug 6628849)

Point the API Reference at the concrete objects create_ingestor() returns and the public generation types.
@kheiss-uwzoo kheiss-uwzoo self-assigned this Aug 24, 2026
@kheiss-uwzoo kheiss-uwzoo added 26.08 doc Improvements or additions to documentation 26.08.1 and removed 26.08 labels Aug 24, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

26.08.1 doc Improvements or additions to documentation

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant