Open
Conversation
…-ingest into kheiss/readme-quickstart
kheiss-uwzoo
commented
Mar 5, 2026
Collaborator
Author
kheiss-uwzoo
left a comment
There was a problem hiding this comment.
changed title and opening paragraph
Changed opening paragraph to be more specific
kheiss-uwzoo
commented
Mar 5, 2026
Collaborator
Author
kheiss-uwzoo
left a comment
There was a problem hiding this comment.
updated opening
sosahi
reviewed
Mar 5, 2026
| ## Prerequisites | ||
| This quick start guide shows how to run NeMo Retriever in **library mode**, directly from your application, without Docker. In library mode, NeMo Retriever Library supports two deployment options: | ||
| - Load Hugging Face models locally on your GPU. | ||
| - Use locally deployed NeMo Retriever NIM endpoints for embedding and OCR. |
Collaborator
nkmcalli
requested changes
Mar 6, 2026
nemo_retriever/README.md
Outdated
| This installs the retriever in editable mode and its in-repo dependencies. Core dependencies (see `nemo_retriever/pyproject.toml`) include Ray, pypdfium2, pandas, LanceDB, PyYAML, torch, transformers, and the Nemotron packages (page-elements, graphic-elements, table-structure). The retriever also depends on the sibling packages `nv-ingest`, `nv-ingest-api`, and `nv-ingest-client` in this repo. | ||
|
|
||
| ### OCR and CUDA 13 runtime | ||
| ## 2. Create and activate the NeMo Retriever Library environment |
Collaborator
There was a problem hiding this comment.
Numbered steps don't go in headings. Remove the number from the heading, or remove the heading.
Collaborator
Author
There was a problem hiding this comment.
Updated document based on your comments
nemo_retriever/README.md
Outdated
Comment on lines
+215
to
+217
| - `run.runtime.summary.json`: run totals (input files, pages, elapsed seconds) | ||
| - `run.ray.timeline.json`: detailed Ray execution timeline | ||
| - `run.rd_dataset.stats.txt`: Ray dataset stats dump |
Collaborator
There was a problem hiding this comment.
Correct format: Change code formatting to bold, change colons to dash (Alt+0150)
nemo_retriever/README.md
Outdated
Comment on lines
+225
to
+226
| - `bo20`: ~9.0 MiB total, ~8.6 MiB LanceDB | ||
| - `jp20`: ~36.8 MiB total, ~36.2 MiB LanceDB |
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
Co-authored-by: nkmcalli <nkmcalli@yahoo.com>
updated per Nicole's review
Fixed formatting per Nicole's review
Updated formatting per Nicole's review
changed from step to procedure
kheiss-uwzoo
commented
Mar 6, 2026
Collaborator
Author
kheiss-uwzoo
left a comment
There was a problem hiding this comment.
updated per Nicole review comments
nemo_retriever/README.md
Outdated
| This installs the retriever in editable mode and its in-repo dependencies. Core dependencies (see `nemo_retriever/pyproject.toml`) include Ray, pypdfium2, pandas, LanceDB, PyYAML, torch, transformers, and the Nemotron packages (page-elements, graphic-elements, table-structure). The retriever also depends on the sibling packages `nv-ingest`, `nv-ingest-api`, and `nv-ingest-client` in this repo. | ||
|
|
||
| ### OCR and CUDA 13 runtime | ||
| ## 2. Create and activate the NeMo Retriever Library environment |
Collaborator
Author
There was a problem hiding this comment.
Updated document based on your comments
removed Library Mode
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pull Request Summary: NeMo Retriever README — NVIDIA Style Guide and PRD Alignment
Overview
Updates to
nemo_retriever/README.mdto align with the NVIDIA Writing Style Guide and the Ingest 2.0 PRD for NeMo Retriever Library, covering voice and tone, formatting, links, acronyms, structure, and naming/positioning.Changes
Voice and Tone (PACE)
Acronyms and First Use
Links
Replaced generic or bare link text with descriptive text that matches the destination (avoiding "here," "read more," and raw URLs):
[docs.nvidia](...benchmarking/)→ NeMo Retriever extraction benchmarking documentation[docs.nvidia](...extraction/audio/)→ NeMo Retriever audio extraction documentation[docs.nvidia](...25.6.3/extraction/audio/)→ NeMo Retriever audio extraction documentation (25.6.3)[docs.nvidia](...ray.html)→ NeMo Ray run guide[huggingface](...parakeet-ctc-1.1b)→ Parakeet CTC 1.1B model on Hugging Face[discuss.ray](...)→ Connecting to a remote Ray cluster on Kubernetes[cohesity](...)→ How Cohesity uses NVIDIA NeMo Retriever microservices to improve RAG AI retrieval recall (Cohesity blog)Formatting and Structure
LD_LIBRARY_PATHexample in a fenced code block and introduced it with a full sentence and colon.## 8heading, explanatory paragraph, and extra horizontal rule).PRD Conformance
Conformance: Yes (with 2 fixes applied)
The README was checked against the Ingest 2.0 PRD (NeMo Retriever Library). It already matched the PRD’s naming and positioning; two small fixes were made so it fully conforms.
Fixes Applied
nemo_retriever):nemoretrieverPython package”nemo_retrieverPython package”nemo_retriever(lowercase, underscore)..txtingestion” (with a trailing space).txtingestion”What Already Matched the PRD
NVIDIA/NeMo-Retrieverreferenced correctly.nemo-retriever(lowercase, hyphenated) in install commands.nemo_retrieverin paths and module references (aside from the onenemoretrieverfix).nv-ingest,nv_ingest, or abbreviations likenr/nemo-ret.nemo-retriever); Python/internal use underscores (nemo_retriever).Optional Follow-up
nemo-retriever-bench, while the README currently documentsretriever harness(for example,retriever harness run).nemo-retriever-bench, add a short note or adjustment in the benchmark section.retriever harnessis the correct, shipped interface, the README is fine as‑is.A short conformance report was saved at
docs/README_PRD_Conformance.mdfor reference and for inclusion in the PR if desired.Files Changed
nemo_retriever/README.mdReference