You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
docs: lead README with the problem and benchmark proof - #151
README-only change that combines two efforts: the problem statement and benchmark proof table from this PR, and the model-context diagram and notebook walkthrough from #153 (merged in with its history, so its author's commits are preserved).
Tagline now frames Kumo Engine as "the dataloader for structured-data models" instead of "accelerator for context generation".
Adds a short problem statement and a three-row proof table. All numbers are already in the Benchmarks section; ratios are derived from it (5.3 min / 19 s ≈ 17x; 240 s / 5 s = 48x and 390 s / 39 s = 10x; 41:14 vs 64:13 ≈ 36% less time). The accuracy difference stays next to the training claim.
If this lands, #153 would be superseded. Not closing it until its author agrees.
Open review suggestions (not yet applied): tighten the NVMe wording, narrow "Databases are slow" to the DuckDB benchmark, and scope the requirements line to the published wheel and training workflows.
Change map: see the comment below.
Testing Done
Docs only; git diff --check is clean and all referenced asset paths exist. Not yet reviewed rendered on GitHub.
Checklist
I listed all the breaking changes introduced by the PR (none)
I made sure to update examples and documentation with my changes (docs-only)
I have added and modified unit tests required to cover changes made by this PR. (n/a)
Replace the "accelerator for context generation" tagline with a plain
dataloader framing, add a proof table drawn from the existing benchmark
numbers, and state hardware requirements above the fold.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
No actionable comments were generated in the recent review.
Walkthrough
The README describes Kumo Engine as a dataloader for structured-data models and as an open dataloader and training engine. It adds benchmark comparisons and qualifications, supported use cases, and Linux x86-64, NVIDIA GPU, and local SSD requirements.
Merge Risk: Low · up to 92b8c
The README is mergeable with bounded documentation corrections, but its headline claims should be narrowed to match the documented evidence and supported workflows.
Pre-merge checks: 2 · 1
Failed checks: 1 inconclusive
Check name
Status
Explanation
Resolution
Change Map Design Review
The required <!-- kumo-pr-change-map --> conversation comment is not present in the supplied review context, and it is not stored in the repository. The authoritative PR diff shows only README.md …
Provide the exact author- or authorized-bot-authored conversation comment beginning with <!-- kumo-pr-change-map -->, including its complete Change map and required design rationale. Re-run the review after that comment is available.
Passed checks: 2 passed
Check name
Status
Explanation
Pr Size
The pull request changes 1 file: README.md. The count does not exceed 40.
Public Resource Access
The only external reference whose text changed is the benchmark report link in README.md line 27. Anonymous GET access returned HTTP 200 from docs.nvidia.com. No added package source or other changed …
Full details: Public Resource Access
Explanation
The only external reference whose text changed is the benchmark report link in README.md line 27. Anonymous GET access returned HTTP 200 from docs.nvidia.com. No added package source or other changed external resource reference is present.
Full details: Change Map Design Review
Explanation
The required <!-- kumo-pr-change-map --> conversation comment is not present in the supplied review context, and it is not stored in the repository. The authoritative PR diff shows only README.md changed (13 additions, 5 deletions); the new benchmark claims and referenced local assets are present, and git diff --check passes. However, the required map-specific evidence—observed failure or release requirement, smallest sufficient change, focused verification, stopping condition, and optional refinements—is unavailable, so each proposed change cannot be assessed.
Resolution
Provide the exact author- or authorized-bot-authored conversation comment beginning with <!-- kumo-pr-change-map -->, including its complete Change map and required design rationale. Re-run the review after that comment is available.
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Tip
🎯 Customize your risk assessments
Guide how CodeRabbit assesses merge risk and security architecture risk. Highlight sensitive changes and explain your security boundaries using Merge Risk Instructions and Security Risk Instructions in your review settings. Security architecture risk assessments require the Advanced plan or above.
✅ Automatic approvals, with your rules
Let CodeRabbit approve pull requests once required threads are resolved, pre-merge checks pass, and your risk limits are met. Enable Automatic approvals in review settings, with overrides for teams or file paths.
CodeRabbit Verify
Run CodeRabbit Verify to test the changes in this pull request.
Reviewing files that changed from the base of the PR and between 43f1c1c and 92b8c0c.
Files selected for processing: 1
README.md
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.
The reason will be displayed to describe this comment to others. Learn more.
Note
Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.
Other comments: 3
README.md: 3
17-17: Minor · Functional Correctness · Quick win
Clarify the NVMe-backed index claim.
“Served from NVMe instead of RAM” can imply that every read bypasses RAM. The detailed README wording supports an NVMe-backed index without keeping the full index in RAM. Use that qualification directly.
Suggested wording
-**The dataloader for structured-data models: connected, time-correct context for every prediction, served from NVMe instead of RAM.**+**The dataloader for structured-data models: connected, time-correct context for every prediction, served from an NVMe-backed index without keeping the full index in RAM.**
19-19: Minor · Functional Correctness · Quick win
Narrow the database performance claim to the benchmarked DuckDB workload.
The benchmark below compares Kumo Engine only with prepared DuckDB, on one synthetic 50-billion-node graph with warm caches. The sentence "Databases are slow" applies that result to databases in general.
Proposed wording
-Models that predict from tables, relationships, and event histories need connected context for every prediction. Databases are slow at producing it, and holding a large graph in memory is expensive. Kumo Engine converts your tables into a compressed, NVMe-optimized graph (Rust-based DiskGraph) and streams time-correct neighborhoods to the GPU for training and batch prediction.+Models that predict from tables, relationships, and event histories need connected context for every prediction. Producing it with prepared DuckDB queries can be slow, and holding a large graph in memory is expensive. Kumo Engine converts your tables into a compressed, NVMe-optimized graph (Rust-based DiskGraph) and streams time-correct neighborhoods to the GPU for training and batch prediction.
31-31: Minor · Functional Correctness · Quick win
Scope the headline requirements to the published wheel and training workflows.
The headline lists Linux x86-64, an NVIDIA GPU, and local SSD as universal requirements. The Installation section limits them. The published 1.0.0 wheel targets Linux x86-64. The GPU and local SSD apply to the documented training workflows. The README also documents an ARM64 source build. The context-generation benchmark runs on one i4i.32xlarge, which has no GPU. Qualify the headline so it does not exclude supported source-build or CPU-only context-generation use.
Proposed wording
-**Requirements:** Linux x86-64, an NVIDIA GPU, and local SSD working storage. Refer to [Installation](#installation) for details.+**Published wheel requirements:** Linux x86-64 with glibc 2.34+ and CPython 3.11+. Training workflows also require an NVIDIA GPU and local SSD working storage. ARM64 source builds are documented separately.
Reviewing files that changed from the base of the PR and between fa46bed and 3f35d41.
Files selected for processing: 1
README.md
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.
Disha, Joseph is also updating the README in #153 (@9Olive), adding visuals and a shorter Quick Tour. Could you two align on the final layout and structure? I’d suggest combining his visuals with your dataloader framing and benchmark table.
Proposed order:
Tagline + intro: Explain what Kumo Engine does and how developers use it.
Installation and Quick Tour links: Give readers a direct path to getting started.
Customer/order example and Joseph’s diagram: Show how connected records become model inputs.
Your benchmark table: Support the performance claims, with the relevant test context.
Key Capabilities and Joseph’s shortened Quick Tour: Follow with the notebook demo and links to complete examples.
Would this combined direction work for both of you?
Merges the model-context diagram and notebook walkthrough from #153 with
the problem statement and benchmark proof table from #151.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Keep the problem statement, proof table and requirements line from this PR
alongside the model-context diagram and shorter tour from #153. Drop the
now-redundant 'Benchmark highlight' line and the old accelerator intro.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
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
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.
GitHub Actions
Error
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Summary
README-only change that combines two efforts: the problem statement and benchmark proof table from this PR, and the model-context diagram and notebook walkthrough from #153 (merged in with its history, so its author's commits are preserved).
If this lands, #153 would be superseded. Not closing it until its author agrees.
Open review suggestions (not yet applied): tighten the NVMe wording, narrow "Databases are slow" to the DuckDB benchmark, and scope the requirements line to the published wheel and training workflows.
Change map: see the comment below.
Testing Done
Docs only;
git diff --checkis clean and all referenced asset paths exist. Not yet reviewed rendered on GitHub.Checklist