7 views
-/https://github.com/berriai/litellm/issues/30724
GitHub · issue

#30724 [Bug]: Bedrock pass-through endpoint does not log request `messages` payload

  • State: open
  • Author: @aashishrbhandari
  • Labels: bug, proxy, llm translation

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

Greetings LiteLLM Team,

## [Quick Summary]

When i use Bedrock Passthrough Endpoint (Link: https://docs.litellm.ai/docs/pass_through/bedrock#1-using-configyaml-recommended-for-model-endpoints) I have configured `s3` and `Custom Callback API` and on both i do not have the actual `message` payload showing (system, tools and user) rather in the S3 object i see the data as ``` "messages": [ { "role": "user", "content": "default-message-value" } ], ``` and on the API endpoint.

## My LiteLLM Setup

### .env

```bash #========[Docker & Docker Compose]======= COMPOSE_PROJECT_NAME=litellm-stack STACK_PREFIX=litellm-stack

# =====[LiteLLM]============

# LiteLLM Container & Volume Names LITELLM_APP_CONTAINER_NAME=litellm-app LITELLM_DB_CONTAINER_NAME=litellm-db LITELLM_POSTGRES_VOLUME_NAME=litellm_postgres_data

# Postgres superuser (bootstrap admin — do not use in app connection strings) POSTGRES_ADMIN_USER=postgres_admin_user POSTGRES_ADMIN_PASSWORD=p0ostresA@dmin123

# LiteLLM app user (least-privileg…

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (3 events)
#0 of 0 · 31d18h39m55s ago — entered · #import:https:::github.com:berriai:litellm post #2216
#30724 is harder because it likely requires tracing request data through the pass-through pipeline and multiple observability backends, while preserving callback contracts, redaction behavior, and coverage across provider-specific paths. #27454 appears more localized to request-schema translation and validation compatibility, with a narrower implementation and test surface.
Issue #30724 is harder because it likely requires tracing and correcting data propagation across the proxy, provider pass-through, and multiple callback/storage integrations, with broader regression and privacy-sensitive testing. Issue #33036 is comparatively localized to response normalization in one provider transformation path.
#0 of 0 · 31d17h48m17s ago — current · #import:https:::github.com:berriai:litellm post #3080
The right issue is harder because it requires tracing request data through the proxy, callback, and object-storage serialization paths, while preserving payload fidelity across integrations and adding broader regression coverage. The left issue is comparatively localized to provider-specific parameter construction with a focused compatibility fix and targeted tests.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search