9 views
-/https://github.com/berriai/litellm/issues/33211
GitHub · issue

#33211 [Bug]: Incorrect Bedrock geo cross-region inference ID for Amazon Nova 2 Lite (apac. doesn't exist, jp. is missing)

  • State: open
  • Author: @romanbashk
  • Labels: bug, llm translation, SDK

### Check for existing issues

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

### What happened?

model_prices_and_context_window.json currently contains a pricing entry for apac.amazon.nova-2-lite-v1:0, but this inference profile ID does not exist according to AWS's own documentation for this model.

Per the AWS Bedrock model card for Amazon Nova 2 Lite (https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-2-lite.html), the only documented Geo Inference IDs are: - us.amazon.nova-2-lite-v1:0 (Geo: US) - eu.amazon.nova-2-lite-v1:0 (Geo: EU) - jp.amazon.nova-2-lite-v1:0 (Geo: JP) - global.amazon.nova-2-lite-v1:0 (Global)

There is no APAC-wide geo tier for this model — Tokyo (ap-northeast-1)/Osaka (ap-northeast-3) traffic uses the jp. prefix specifically, per the "Geo: JP" section of the model card.

Calling Bedrock with apac.amazon.nova-2-lite-v1:0 (which currently exists in model_cost) returns: litellm.BadRequestError: BedrockException - {"message":"The provided model identifier is invalid."}

Request: 1. Remove the incorrect apac.amazon.nova-2-lite-v1:0 entry from model_prices_and_context_window.json 2. Add the corre…

GitHub resolver

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

Refresh page
vote history (4 events)
#0 of 0 · 31d18h37m21s ago — entered · #import:https:::github.com:berriai:litellm post #2261
The right item spans request orchestration, rate-limit handling, fallback semantics, and cross-component integration testing. The left item is a narrowly scoped catalog correction with straightforward validation.
Issue 32862 is substantially harder because it requires secure request-path changes, careful SSRF threat analysis, regression testing, and validation across synchronous/asynchronous guardrail behavior. Issue 33211 is primarily a targeted model-metadata correction with focused SDK/configuration tests.
The left issue is harder because it requires tracing and correcting cross-provider request/response translation, handling a model-specific edge case without breaking normal token-limit semantics, and validating compatibility with the proxy client flow. The right issue is primarily a targeted Bedrock model-metadata correction with focused validation.
#0 of 0 · 31d17h40m2s ago — current · #import:https:::github.com:berriai:litellm post #3226
The left issue requires tracing a proxy endpoint through provider failures, normalizing multiple error paths, preserving response-schema compatibility, and adding regression coverage. The right issue is primarily a targeted provider metadata correction with comparatively limited code and test impact.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search