fix: Fix prefix synthesis error with text_input trace

[lkomali]

Synthesizer is adding input_length for any trace if not provided.
This causes an error if text_input is provided in the trace because only one of the two fields should be provided in mooncake format.

Summary by CodeRabbit

Release Notes

• Bug Fixes

  • Fixed an issue where synthesis operations could unintentionally modify input content length for traces using text_input fields.

• Tests

  • Added integration test for text_input trace handling with synthesis speedup, validating request counts and output integrity.
  • Added unit test ensuring text_input content is preserved during synthesis operations.

[github-actions]

Try out this PR

Quick install:

pip install --upgrade --force-reinstall git+https://github.com/ai-dynamo/aiperf.git@79e5bdde4ea6b87dfb2485830ad81c096f961538

Recommended with virtual environment (using uv):

uv venv --python 3.12 && source .venv/bin/activate
uv pip install --upgrade --force-reinstall git+https://github.com/ai-dynamo/aiperf.git@79e5bdde4ea6b87dfb2485830ad81c096f961538

---

Last updated for commit: 79e5bdd • Browse code

[coderabbitai]

Warning

Rate limit exceeded

@lkomali has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 17 minutes and 49 seconds before requesting another review.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

Walkthrough

The trace synthesis logic now conditionally assigns input_length only for traces without text_input, preventing Mooncake validation conflicts. Supporting tests verify this behavior through integration testing with text_input traces and speedup, plus unit test confirmation of text_input preservation.

Changes

Cohort / File(s)	Summary
Synthesizer Logic Update src/aiperf/dataset/synthesis/synthesizer.py	Conditional guard added to assign input_length only when text_input is absent, preventing validation conflicts. Guard avoids modifying traces that already use text_input.
Test Coverage tests/integration/test_mooncake_trace.py, tests/unit/dataset/synthesis/test_synthesizer.py	Integration test validates Mooncake traces with text_input and synthesis speedup ratio applied. Unit test confirms text_input preservation and input_length exclusion in synthesis output.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Poem

🐰 A trace with text_input came hopping by,
No need for length—it's standing quite high!
The synthesizer knows when to play,
Leave text_input alone, let speedup have its day!
With tests to confirm, all fuzzy and bright,
The validation conflict is fixed just right! ✨

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely summarizes the main fix: preventing synthesis errors when processing traces with text_input fields.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[matthewkotila]

🚀

[ajcasagrande]

looks good, but please add tests

[debermudez]

agreed. 1 test to cover us for later and this is good to go