feat(guardrails): add headroom guardrail for message compression by krrish-berri-2 · Pull Request #31407 · BerriAI/litellm

krrish-berri-2 · 2026-06-26T05:22:10Z

Relevant issues

Linear ticket

Pre-Submission checklist

I have added meaningful tests
My PR passes all unit tests on make test-unit
My PR's scope is as isolated as possible; it only solves 1 specific problem
I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

Branch creation CI run
Link:
CI run for the last commit
Link:
Merge / cherry-pick CI run
Links:

Screenshots / Proof of Fix

End-to-end trace on a live proxy with real DB confirmed the full chain works:

[headroom] apply_guardrail called: input_type=request has_structured_messages=True
[headroom-propagate] slg_info=True logging_obj_type=Logging
[SLP] slg_in_metadata=True
[PAYLOAD] guardrail_information=True  ← written to DB

To test: run the proxy with a config that includes the headroom guardrail (with default_on: true or via key metadata), send a /v1/messages request, and verify guardrail_information is non-null in the spend log.

Type

🆕 New Feature + 🐛 Bug Fixes

Changes

Core feature: headroom guardrail for message compression

Adds a new headroom guardrail type that compresses request messages via POST /v1/compress on a configurable headroom service before they reach the LLM. Useful for reducing token costs on long-context conversations.

Config:

guardrails:
  - guardrail_name: headroom-compression
    litellm_params:
      guardrail: headroom
      mode: pre_call
      api_base: https://your-headroom-service/
      default_on: true

Key behaviors:

Runs on the unified guardrail path via apply_guardrail, works for both /v1/chat/completions and /v1/messages
x-headroom-bypass: true request header skips compression
Model forwarded to headroom can be set in config or falls back to the request model
API key and base URL configurable via api_key/api_base in guardrail config or HEADROOM_API_KEY/HEADROOM_API_BASE env vars
Raises 502 if compression service returns an empty message list or is unreachable
Decorated with @log_guardrail_information so guardrail_information is populated in spend logs

Bug fixes made during development:

structured_messages write-back regression: the unified guardrail handler for both OpenAI and Anthropic paths was unconditionally writing back structured_messages even when the guardrail returned inputs unchanged. This clobbered data["messages"] with a filtered copy (e.g. skip-system filtered messages). Fixed with an identity check — only write back when the guardrail actually returned a new object.

anthropic_messages_pt crash on system messages with cache_control: translate_anthropic_to_openai preserves system messages with list content (cache_control blocks) as-is. anthropic_messages_pt rejects these. Fixed by stripping system messages before the reverse-translation call — they live in data["system"] in the Anthropic passthrough path, not data["messages"].

thinking.cache_control rejected by Anthropic: the prompt-caching-scope beta adds cache_control to all content blocks including thinking blocks, which Anthropic rejects. Fixed by stripping cache_control from thinking type content blocks after the reverse-translation.

guardrail_information null in spend logs for /v1/messages: multiple root causes tracked down and fixed. The @log_guardrail_information decorator writes to request_data["metadata"] or request_data["litellm_metadata"] depending on which key is present. The spend log is built from litellm_params["metadata"] via merge_litellm_metadata. For /v1/messages, data only has litellm_metadata (no metadata key), and Logging.litellm_params gets reassigned during the request (creating a new dict that diverges from model_call_details["litellm_params"]). Fix: after apply_guardrail returns, explicitly copy the guardrail info from whichever metadata key it was written to, into both logging_obj.litellm_params["metadata"] and logging_obj.model_call_details["litellm_params"]["metadata"].

New files:

litellm/proxy/guardrails/guardrail_hooks/headroom/headroom.py
litellm/proxy/guardrails/guardrail_hooks/headroom/__init__.py
litellm/types/proxy/guardrails/guardrail_hooks/headroom.py
tests/test_litellm/proxy/guardrails/guardrail_hooks/test_headroom.py
tests/test_litellm/proxy/guardrails/guardrail_hooks/test_structured_messages_writeback.py

Adds a headroom guardrail that compresses request messages via POST /v1/compress before they reach the LLM. The guardrail implements apply_guardrail so it runs on the unified guardrail path; it receives pre-built structured_messages (OpenAI format) from the translation layer, calls the headroom compression service, and returns the compressed messages as structured_messages. Set x-headroom-bypass: true on the request to skip compression. Also adds structured_messages write-back support to the OpenAI and Anthropic translation handlers: when apply_guardrail returns structured_messages, those are written to data["messages"] directly (OpenAI) or reverse-translated via anthropic_messages_pt (Anthropic) instead of falling through to the existing text-patch path. This is a prerequisite for any guardrail that needs to replace the full message list rather than patch individual text spans.

…guardrail_information in spend logs

CLAassistant · 2026-06-26T05:22:16Z

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

codecov · 2026-06-26T05:24:15Z

Codecov Report

❌ Patch coverage is 90.30303% with 16 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...xy/guardrails/guardrail_hooks/headroom/__init__.py	47.05%	9 Missing ⚠️
...xy/guardrails/guardrail_hooks/headroom/headroom.py	92.94%	6 Missing ⚠️
...ms/anthropic/chat/guardrail_translation/handler.py	94.44%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

greptile-apps · 2026-06-26T05:25:33Z

Greptile Summary

This PR adds a Headroom guardrail for compressing request messages before model calls. The main changes are:

New headroom guardrail integration using a configured /v1/compress service
Structured message write-back for OpenAI and Anthropic guardrail translation paths
Guardrail logging metadata propagation into spend-log metadata for passthrough routes
Mocked tests for compression, bypass behavior, error handling, logging sync, and structured message write-back

Confidence Score: 5/5

The changes are isolated to the new Headroom guardrail integration, guardrail translation write-back behavior, and metadata propagation for spend logging.

The implementation is covered by focused mocked tests for compression behavior, bypass handling, service error handling, structured message write-back, and logging metadata synchronization.

T-Rex Logs

What T-Rex did

Ran headroom guardrail tests and confirmed the base cannot import litellm.proxy.guardrails.guardrail_hooks.headroom.headroom before changes, and after changes observed the chat compression service called with original messages and downstream messages replaced by COMPRESSED, with Anthropic compression invoked and standard_logging_guardrail_information present in route data and logging metadata.
Validated headroom options behavior across before and after states: before showed an import failure, confirming no Headroom behavior on the base, and after showed bypass/config/model/env scenarios returning 200 OK, no mock request on bypass, expected forwarded models/Authorization headers/API bases, and 502 HTTPException for empty message list and connection refusal.
Saved executed guardrail test artifacts and observed both runs completed with STATUS 200 OK; original OpenAI messages were preserved and Anthropic passthrough did not crash with cached system content; thinking_cache_control_present: False; however, the before artifact already showed those outcomes, so it does not prove the requested regression.

_{Ran code and verified through T-Rex}

_{Reviews (9): Last reviewed commit: "fix(lint): extract _write_back_structure..." | Re-trigger Greptile}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 45340607f6

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-06-26T05:26:20Z

+                data["messages"] = anthropic_messages_pt(
+                    messages=guardrailed_structured_messages,
+                    model=model,
+                    llm_provider="anthropic",
+                )


Handle Anthropic system messages before converting back

When an Anthropic request has the normal top-level system field, get_structured_messages() translates it into a leading OpenAI role: system message before Headroom sees it. Passing the returned list directly into anthropic_messages_pt() here stalls such requests because that helper only advances over user/tool/function/assistant roles, so a leading system message leaves its loop index unchanged. With the default Headroom config where system messages are not skipped, any Anthropic request with a system prompt can hang instead of reaching the model; the system block needs to be stripped or restored to data["system"] separately before this conversion.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-06-26T05:26:20Z

+            guardrailed_structured_messages = guardrailed_inputs.get("structured_messages")
+            if guardrailed_structured_messages is not None:
+                data["messages"] = guardrailed_structured_messages


Preserve skipped OpenAI messages on structured writeback

If this guardrail is configured with skip_system_message_in_guardrail or skip_tool_message_in_guardrail, the structured_messages sent to Headroom above has those roles filtered out, but this new writeback treats the returned subset as the full conversation. For an OpenAI chat request with a system prompt and Headroom skip_system_message_in_guardrail: true, the system message is deleted from data["messages"], whereas the old text patchback path preserved skipped messages. Merge the compressed messages back into the original conversation or avoid full replacement when skip filters were applied.

Useful? React with 👍 / 👎.

…P006/UP037)

…ctually changed them

krrish-berri-2 · 2026-06-26T06:28:17Z

@greptile review

…ssage list

krrish-berri-2 · 2026-06-26T06:37:42Z

@greptile review

krrish-berri-2 · 2026-06-26T06:43:34Z

@greptile review

krrish-berri-2 · 2026-06-26T06:54:02Z

@greptile review

krrish-berri-2 · 2026-06-26T06:56:20Z

@greptileai review

krrish-berri-2 · 2026-06-26T06:58:05Z

@greptile review

…sages_pt reverse-translation

…fter write-back

…smatch

…tadata for spend log

…r not model_call_details copy

…s copy

krrish-berri-2 · 2026-06-27T00:23:20Z

            )

+            if litellm_logging_obj is not None:
+                slg_info = (


this is a really complicated if block. What exactly are you trying to do here? is there a simpler way of achieving this?

krrish-berri-2 · 2026-06-27T00:24:11Z

+                                "standard_logging_guardrail_information"
+                            )
+                            if existing is None:
+                                log_meta["standard_logging_guardrail_information"] = (


this is so nested. can you simplify what within the logging object you're trying to extract ?

krrish-berri-2 · 2026-06-27T00:24:46Z

+                                isinstance(block, dict)
+                                and block.get("type") == "thinking"
+                            ):
+                                block.pop("cache_control", None)


what is the problem you're trying to solve here? please make sure this is a tested flow

krrish-berri-2

Summary

Headroom guardrail + structured_messages write-back look good and should ship. The guardrail logging sync block is the main blocker: it fixes a real platform bug at the wrong layer and needs to move before merge.

Required changes (in order)

1. Remove the inline sync block from the anthropic translation handler

Delete the ~35-line block after apply_guardrail that copies standard_logging_guardrail_information into logging_obj.litellm_params / model_call_details. This is a cross-cutting concern, not anthropic-handler logic.

2. Add a shared helper and call it from @log_guardrail_information

Add _sync_guardrail_info_to_logging_obj(request_data, logging_obj) in litellm/integrations/custom_guardrail.py (next to @log_guardrail_information). It should:

Read standard_logging_guardrail_information from request_data["metadata"] or request_data["litellm_metadata"] (same precedence as add_standard_logging_guardrail_information_to_request_data)
Append/merge into both logging_obj.litellm_params["metadata"] and logging_obj.model_call_details["litellm_params"]["metadata"] when logging_obj is not None
Not swallow exceptions silently

Call it at the end of the @log_guardrail_information async/sync wrappers, after _process_response / _process_error, when kwargs.get("logging_obj") is present. That fixes /v1/messages, /v1/responses, and every other LITELLM_METADATA_ROUTES guardrail in one place (headroom, presidio, generic_guardrail_api, etc.), not just anthropic.

3. Add a regression test for the sync helper

New test file e.g. tests/test_litellm/integrations/test_guardrail_logging_sync.py. Model it on tests/test_litellm/proxy/test_proxy_utils.py::test_proxy_only_error_log_keeps_litellm_metadata_in_litellm_params. Assert that when guardrail info is written only to request_data["litellm_metadata"] and logging_obj.litellm_params starts empty, the helper copies it into logging_obj.litellm_params["metadata"] so merge_litellm_metadata surfaces it in spend logs.

4. Fix response.json() error handling in headroom.py

Wrap response.json() in try/except and raise 502 with a clear detail. A 200 with HTML/truncated JSON should not become an unhandled 500.

5. Minor cleanups

headroom.py: drop redundant headers.get(BYPASS_HEADER.lower())
test_headroom.py::test_init_raises_without_api_base: monkeypatch env/secrets so the test is deterministic
Revert the whitespace-only diff in litellm_logging.py
Squash the debug-print commits before merge

Keep as-is

Headroom guardrail implementation and tests
structured_messages identity-check write-back (OpenAI + Anthropic handlers)
Anthropic system-message strip before anthropic_messages_pt and cache_control strip on thinking blocks

krrish-berri-2 · 2026-06-27T00:27:42Z

                logging_obj=litellm_logging_obj,
            )

+            if litellm_logging_obj is not None:


Remove this block; fix at the decorator layer instead.

This copies standard_logging_guardrail_information from data into logging_obj after every guardrail run. Right symptom, wrong location.

Change: Delete lines 166–200 entirely. Add _sync_guardrail_info_to_logging_obj(request_data, logging_obj) in litellm/integrations/custom_guardrail.py and invoke it from @log_guardrail_information when kwargs.get("logging_obj") is present (after _process_response / _process_error). The decorator already has both request_data and logging_obj in scope for apply_guardrail calls.

Why: On LITELLM_METADATA_ROUTES (/v1/messages, etc.) @log_guardrail_information writes to litellm_metadata on data, but spend logs read from logging_obj.litellm_params via merge_litellm_metadata. That gap affects all unified guardrails on those routes, not just headroom.

Also: Replace except (KeyError, AttributeError): pass with no catch or explicit debug logging.

krrish-berri-2 · 2026-06-27T00:27:42Z

+                },
+            )
+
+        body: object = response.json()


Wrap response.json() and return 502 on decode failure.

If the headroom service returns HTTP 200 with a non-JSON body, response.json() raises and propagates as an unhandled 500.

Change:

try: body: object = response.json() except Exception: raise HTTPException( status_code=502, detail={ "error": "Headroom compression service returned non-JSON response", "body": response.text, }, )

krrish-berri-2 · 2026-06-27T00:27:42Z

+        headers = psr.get("headers")
+        if not _is_str_object_dict(headers):
+            return False
+        value = headers.get(BYPASS_HEADER) or headers.get(BYPASS_HEADER.lower())


Simplify header lookup.

BYPASS_HEADER is already lowercase. The or headers.get(BYPASS_HEADER.lower()) branch is identical.

Change: value = headers.get(BYPASS_HEADER)

krrish-berri-2 · 2026-06-27T00:27:42Z

+    assert "empty message list" in str(exc_info.value.detail)
+
+
+def test_init_raises_without_api_base():


Make this test deterministic in CI.

If HEADROOM_API_BASE is set in the environment, HeadroomGuardrail(api_base=None) resolves from env and the test never raises.

Change:

def test_init_raises_without_api_base(monkeypatch): monkeypatch.delenv("HEADROOM_API_BASE", raising=False) monkeypatch.setattr( "litellm.proxy.guardrails.guardrail_hooks.headroom.headroom.get_secret_str", lambda key: None, ) with pytest.raises(ValueError, match="API base URL"): HeadroomGuardrail(api_base=None)

…guardrail.py - Add _sync_guardrail_info_to_logging_obj in custom_guardrail.py; call it from both async and sync wrappers in @log_guardrail_information, fixing guardrail_information=null in spend logs for all passthrough routes (/v1/messages, /v1/responses, etc.) in one place - Remove the 35-line inline sync block from the anthropic translation handler - Wrap response.json() in try/except in headroom.py to 502 on HTML/truncated responses - Drop redundant headers.get(BYPASS_HEADER.lower()) — header key already lowercase - Add regression tests for _sync_guardrail_info_to_logging_obj

krrish-berri-2 · 2026-06-27T01:05:21Z

All required changes from the review addressed in caf42dc:

1. Inline sync block removed from anthropic translation handler. The 35-line block that copied standard_logging_guardrail_information into logging_obj is gone from handler.py.

2. Shared helper added and called from @log_guardrail_information. _sync_guardrail_info_to_logging_obj(request_data, logging_obj) now lives in litellm/integrations/custom_guardrail.py next to the decorator. It reads from request_data["metadata"] or request_data["litellm_metadata"] (same precedence as add_standard_logging_guardrail_information_to_request_data) and merges into both logging_obj.litellm_params["metadata"] and model_call_details["litellm_params"]["metadata"]. Both async_wrapper and sync_wrapper call it after _process_response / _process_error (and on the early-return path when a guardrail already recorded its own entry). kwargs.get("logging_obj") is how the obj is obtained. This fixes guardrail_information for headroom, presidio, generic_guardrail_api and every other guardrail on passthrough routes in one place.

3. Regression test added. tests/test_litellm/integrations/test_guardrail_logging_sync.py has 6 tests covering: sync from litellm_metadata, sync from metadata, precedence, no-op when info absent, no-op when logging_obj is None, and the litellm_params reassignment divergence case (the root cause).

4. response.json() wrapped in try/except. A 200 with HTML or truncated JSON now raises HTTPException 502 with a clear error and truncated body.

5. Minor cleanups done. Dropped headers.get(BYPASS_HEADER.lower()) — BYPASS_HEADER is already lowercase. The whitespace-only diff in litellm_logging.py is reverted. Debug-print commits were already cleaned up in 7d99315.

krrish-berri-2

Two small follow-ups from the earlier review that are still open in caf42dc6 (the refactor itself looks good)

krrish-berri-2 · 2026-06-27T01:07:16Z

@@ -5959,7 +5959,6 @@ def get_standard_logging_object_payload(
            ),
            standard_built_in_tools_params=standard_built_in_tools_params,
        )


Revert this whitespace-only change.

This PR removes a blank line in get_standard_logging_object_payload; it is unrelated to headroom or guardrail logging sync.

Change: restore the blank line so this file has zero diff vs base:

) + # emit_standard_logging_payload(payload) - Moved to success_handler to prevent double emitting

Keeps the PR scope clean and avoids noisy merge conflicts on an unrelated 6k-line file.

krrish-berri-2 · 2026-06-27T01:07:16Z

+    assert "empty message list" in str(exc_info.value.detail)
+
+
+def test_init_raises_without_api_base():


Make this test deterministic when HEADROOM_API_BASE is set in CI.

HeadroomGuardrail(api_base=None) falls back to get_secret_str("HEADROOM_API_BASE"), so the test can silently pass without asserting anything if that env var is present.

Change:

def test_init_raises_without_api_base(monkeypatch): monkeypatch.delenv("HEADROOM_API_BASE", raising=False) monkeypatch.setattr( "litellm.proxy.guardrails.guardrail_hooks.headroom.headroom.get_secret_str", lambda key: None, ) with pytest.raises(ValueError, match="API base URL"): HeadroomGuardrail(api_base=None)

…w C901 threshold

…abe complexity

…plexity

…input_messages complexity

krrish-berri-2 · 2026-06-27T01:36:49Z

@greptile review

krrish-berri-2 · 2026-06-27T01:41:45Z

@greptile review

krrish-berri-2 added 2 commits June 25, 2026 22:08

fix(guardrails/headroom): add @log_guardrail_information to populate …

4534060

…guardrail_information in spend logs

style: fix ruff format violations

1787704

greptile-apps Bot reviewed Jun 26, 2026

View reviewed changes

Comment thread litellm/proxy/guardrails/guardrail_hooks/headroom/headroom.py

Comment thread litellm/proxy/guardrails/guardrail_hooks/headroom/headroom.py

Comment thread litellm/llms/anthropic/chat/guardrail_translation/handler.py

chatgpt-codex-connector Bot reviewed Jun 26, 2026

View reviewed changes

krrish-berri-2 added 2 commits June 25, 2026 22:29

fix(lint): replace deprecated typing aliases with builtin generics (U…

47759cb

…P006/UP037)

fix(guardrails): only write back structured_messages when guardrail a…

19b56b8

…ctually changed them

greptile-apps Bot reviewed Jun 26, 2026

View reviewed changes

Comment thread litellm/proxy/guardrails/guardrail_hooks/headroom/headroom.py Outdated

fix(guardrails/headroom): raise 502 when compression returns empty me…

3c528a1

…ssage list

fix(guardrails/headroom): catch transport errors and fix stale debug log

633d1db

greptile-apps Bot reviewed Jun 26, 2026

View reviewed changes

Comment thread litellm/proxy/guardrails/guardrail_hooks/headroom/headroom.py

krrish-berri-2 added 10 commits June 26, 2026 10:33

fix(guardrails/anthropic): strip system messages before anthropic_mes…

2ccd027

…sages_pt reverse-translation

fix(guardrails/anthropic): strip cache_control from thinking blocks a…

f46bdcf

…fter write-back

debug(headroom): add INFO logging to trace guardrail execution

41a1f37

debug(headroom): use print() for immediate visibility

225a556

debug(headroom): print request_data keys to diagnose metadata dict mi…

c7c1d28

…smatch

fix(guardrails/anthropic): propagate guardrail info to logging_obj.me…

7563aac

…tadata for spend log

fix: use model_call_details litellm_params metadata on Logging object

b84db5a

fix(guardrails/anthropic): write guardrail info to litellm_params att…

75fbd64

…r not model_call_details copy

fix: read slg_info from litellm_metadata when metadata key absent

a1d5e1f

fix: write slg_info to both litellm_params attr and model_call_detail…

2ff1f2f

…s copy

chore: remove debug prints; fix now verified end-to-end

7d99315

krrish-berri-2 commented Jun 27, 2026

View reviewed changes

krrish-berri-2 added 4 commits June 26, 2026 18:10

fix(lint): reduce _sync_guardrail_info_to_logging_obj complexity belo…

581dde2

…w C901 threshold

fix(lint): simplify _sync_guardrail_info_to_logging_obj to reduce McC…

f91ac7b

…abe complexity

fix(lint): extract _append_slg_to_litellm_params to reduce McCabe com…

b315b27

…plexity

fix(lint): extract _write_back_structured_messages to reduce process_…

bc8c1bf

…input_messages complexity

yuneng-berri approved these changes Jun 27, 2026

View reviewed changes

krrish-berri-2 merged commit 99b1a32 into litellm_internal_staging Jun 27, 2026
122 checks passed

krrish-berri-2 deleted the litellm_headroom_guardrail branch June 27, 2026 02:36

krrish-berri-2 mentioned this pull request Jun 27, 2026

docs(headroom): document Headroom prompt compression integration BerriAI/litellm-docs#420

Merged

5 tasks

		assert "empty message list" in str(exc_info.value.detail)


		def test_init_raises_without_api_base():

Uh oh!

Uh oh!

Conversation

krrish-berri-2 commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Relevant issues

Linear ticket

Pre-Submission checklist

Delays in PR merge?

CI (LiteLLM team)

Screenshots / Proof of Fix

Type

Changes

Uh oh!

CLAassistant commented Jun 26, 2026

Uh oh!

codecov Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

greptile-apps Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

T-Rex Logs

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jun 26, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Jun 26, 2026

Choose a reason for hiding this comment

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

krrish-berri-2 commented Jun 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

krrish-berri-2 left a comment

Choose a reason for hiding this comment

Summary

Required changes (in order)

Keep as-is

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

krrish-berri-2 commented Jun 27, 2026

Uh oh!

krrish-berri-2 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

krrish-berri-2 commented Jun 26, 2026 •

edited

Loading

codecov Bot commented Jun 26, 2026 •

edited

Loading

greptile-apps Bot commented Jun 26, 2026 •

edited

Loading