[DOC] Update single node performance benchmarks

[jairad26]

Description of changes

Updated perf benchmarks for single node chroma on docs, also adds theme-image, which can be reused in the future to have different images for light vs dark theme

Test plan

How are these changes tested?

• Tests pass locally with pytest for python, yarn test for js, cargo test for rust

Documentation Changes

Are all docstrings for user-facing APIs updated if required? Do we need to make documentation changes in the docs section?

[jairad26]

• [DOC] Update single node performance benchmarks #4695 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

Reviewer Checklist

Please leverage this checklist to ensure your code review is thorough before approving

Testing, Bugs, Errors, Logs, Documentation

• Can you think of any use case in which the code does not behave as intended? Have they been tested?
• Can you think of any inputs or external events that could break the code? Is user input validated and safe? Have they been tested?
• If appropriate, are there adequate property based tests?
• If appropriate, are there adequate unit tests?
• Should any logging, debugging, tracing information be added or removed?
• Are error messages user-friendly?
• Have all documentation changes needed been made?
• Have all non-obvious changes been commented?

System Compatibility

• Are there any potential impacts on other parts of the system or backward compatibility?
• Does this change intersect with any items on our roadmap, and if so, is there a plan for fitting them together?

Quality

• Is this code of a unexpectedly high quality (Readability, Modularity, Intuitiveness)

[propel-code-bot]

Update Single-Node Performance Benchmarks and Add Theme-Aware Images to Docs

This PR updates the single-node Chroma performance benchmark documentation with new empirical data and significantly expands the discussion of system behavior, latency, and scaling. It introduces a reusable, theme-aware MarkdocImage React component to support light and dark mode documentation imagery, adjusts Markdoc configuration to integrate this component, and replaces static inline images with the new multi-theme approach and updated graphics. The benchmarks table and descriptive sections have been refactored for improved accuracy and clarity, and corresponding image assets have been added.

Key Changes:
• Introduced MarkdocImage React component to enable light/dark theme image support in documentation.
• Extended Markdoc configuration (config.ts) to register the MarkdocImage tag and its attributes.
• Rewrote large sections of the performance.md page for clearer, empirically updated benchmarking data, adding detailed descriptions of performance limitations, latency characteristics, concurrency, and scalability implications.
• Swapped inline images for MarkdocImage tags and added new/updated image assets optimized for both light and dark themes.
• Revised data tables and narrative to reflect recent performance test results.

Affected Areas:
• Documentation - performance benchmarks and explanatory text
• Markdoc documentation configuration and components
• Image assets for documentation

Potential Impact:

Functionality: No effect on core library functionality, but documentation experience is improved, especially for users in different display modes. Benchmarks reflect more recent real-world hardware/software performance.

Performance: No runtime system impact; documentation site performance may slightly increase due to additional conditional rendering logic in MarkdocImage but not significant.

Security: No material impact on security.

Scalability: No direct application impact; documentation now provides clearer guidance on scaling limits and system behaviors under load.

Review Focus:
• Ensure all new/existing images display as expected in both light and dark mode, and fallbacks are correct.
• Look for copy-paste or content errors in the updated benchmarking tables or narrative (fact checking).
• Check that MarkdocImage is correctly integrated and not breaking any previous use-cases.
• Assess whether any remaining in-line image references should use MarkdocImage for theme support.

Testing Needed

• Manually inspect documentation pages in both light and dark themes to verify correct image rendering and overall presentation.
• Spot-check MarkdocImage usage across updated docs to ensure alt and src attributes are set and fallback logic works if theme info is missing or not loaded.
• Review new benchmark tables and accompanying narrative for consistency and accuracy.

Code Quality Assessment

docs/docs.trychroma.com/components/markdoc/markdoc-image.tsx: Good: Well-structured, leverages React state and next-themes for dynamic theme detection, uses prop typing, sets sensible defaults. No obvious performance or code-style issues.

docs/docs.trychroma.com/markdoc/content/production/administration/performance.md: Well-organized, factually dense, uses Markdoc tags for modularity. Some long sentences, but style matches technical documentation norms.

docs/docs.trychroma.com/markdoc/config.ts: Clean integration of new tag, maintains strong typing, logical grouping of component registrations.

Best Practices

Documentation:
• Provides real, current, empirically gathered data for benchmarks.
• Improves accessibility and UX with alt text and alt theme images.
• Uses modular custom components for maintainability and extensibility.

React:
• Uses functional components, hooks, prop typing, and proper resource management.

Potential Issues

• Theme detection is client-only; images may briefly flash with incorrect theme on SSR/hydration edge cases.
• Any old direct image references remaining will not benefit from theme support (may not be a problem, but worth a check).
• If MarkdocImage is misused (missing a required src or alt), renders may break or degrade accessibility.
• The documentation's technical accuracy depends on the correctness of the new benchmark data (cannot be programmatically verified).

This summary was automatically generated by @propel-code-bot

[HammadB]

This could be odd if you don't have context

[HammadB]

Were these in the same region?feels a bit high

[jairad26]

yeah these are in the same region. i think the reason p999 for t3.small is this high is bc this was measured it's near max collection size (these p999's are for collection size of 200k). got that number based on the last run of tests. I'll drop the ones way out of st dev, because theyre not representative, and have what we can consider outside factors (OOM)

[HammadB]

Btw I think these will look odd in light mode, they are optimized for dark mode

[jairad26]

Btw I think these will look odd in light mode, they are optimized for dark mode

yes, here's an image in light mode

and here's another page from the docs

i think it looks comparable to the rest of the docs, so didn't look into adding alt images for light/dark mode (i couldnt find any examples of that either). can look into that if necessary?

[jairad26]

[HammadB]

Discussed a few changes offline