litellm - ✅(Solved) Fix [Bug]: Azure GPT-5.2 chat rejects leaked vector_store_ids from extra_body passthrough [1 pull requests, 4 comments, 4 participants]

xykong · 2026-03-05T12:23:58Z

[litellm] When routing Azure GPT-5.2 chat requests through LiteLLM, vector store ids can leak into provider payload via extra body , causing Azure to reject th… When routing Azure GPT-5.2 chat requests through LiteLLM, `vector_store_ids` can leak into provider payload via `extra_body`, causing Azure to reject the request with: `litellm.BadRequestError: AzureException BadRequestError - Unknown parameter: 'vector_store_ids'` # PR #22899: fix(azure): prevent vector_store_ids leak into extra_body passthrough - Repository: BerriAI/litellm - Author: xykong - State: closed | merged: True - Link: https://github.com/BerriAI/litellm/pull/22899 ## Description (problem / solution / changelog) ## Summary - prevent LiteLLM internal `vector_store_id(s)` params from being forwarded into `extra_body` for openai/azure/openai-compatible providers - keep non-internal unknown params pass-through behavior unchanged - add regression tests for Azure GPT-5.2 optional param construction path ## Why Azure GPT-5.2 chat rejects `vector_store_ids` as unknown top-level input when it leaks via `extra_body` passthrough. Fixes #22898. ## Test - `poetry run pytest tests/test_litellm/test_utils.py -k "vector_store_ids or get_optional_params_azure_gpt5" -vv` - `poetry run pytest tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py -vv` ## Changed files - `litellm/utils.py` (modified, +9/-1) - `tests/test_litellm/test_utils.py` (modified, +37/-0) ## Fixed - Fixed by PR: fix(azure): prevent vector_store_ids leak into extra_body passthrough (https://github.com/BerriAI/litellm/pull/22899) ## Summary When routing Azure GPT-5.2 chat requests through LiteLLM, `vector_store_ids` can leak into provider payload via `extra_body`, causing Azure to reject the request with: `litellm.BadRequestError: AzureException BadRequestError - Unknown parameter: 'vector_store_ids'` ## Repro ```python import litellm resp = litellm.completion( model="gpt-5.2", custom_llm_provider="azure", messages=[{"role": "user", "content": "hi"}], temperature=1, vector_store_ids=["vs_123"], ) ``` Observed: request fails with unknown parameter `vector_store_ids`. ## Root cause In `litellm/utils.py` (`add_provider_specific_params_to_optional_params`), non-openai params are moved into `extra_body` for openai/azure providers. `vector_store_ids` / `vector_store_id` are LiteLLM internal orchestration params, not valid Azure/OpenAI chat completion body params, but they get forwarded. ## Proposed fix Block internal vector store orchestration keys from `extra_body` pass-through: - `vector_store_id` - `vector_store_ids` ## Validation Added tests (local branch): 1. openai-compatible passthrough should not forward vector store keys. 2. `get_optional_params(model=\"gpt-5.2\", custom_llm_provider=\"azure\", vector_store_ids=[...])` should not include `vector_store_ids` in `extra_body`. Happy to open PR with code + tests.

litellm2026-03-05 12:23:58

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

BerriAI/litellm#22898•Fetched 2026-04-08 00:39:27

View on GitHub

Comments

Participants

Timeline

Reactions

Author

Participants

Timeline (top)

commented ×4subscribed ×3mentioned ×2closed ×1

When routing Azure GPT-5.2 chat requests through LiteLLM, vector_store_ids can leak into provider payload via extra_body, causing Azure to reject the request with:

litellm.BadRequestError: AzureException BadRequestError - Unknown parameter: 'vector_store_ids'

Root Cause

In litellm/utils.py (add_provider_specific_params_to_optional_params), non-openai params are moved into extra_body for openai/azure providers. vector_store_ids / vector_store_id are LiteLLM internal orchestration params, not valid Azure/OpenAI chat completion body params, but they get forwarded.

Fix Action

Fixed

Fixed by PR: fix(azure): prevent vector_store_ids leak into extra_body passthrough (https://github.com/BerriAI/litellm/pull/22899)

PR fix notes

PR #22899: fix(azure): prevent vector_store_ids leak into extra_body passthrough

Repository: BerriAI/litellm
Author: xykong
State: closed | merged: True
Link: https://github.com/BerriAI/litellm/pull/22899

Description (problem / solution / changelog)

Summary

prevent LiteLLM internal vector_store_id(s) params from being forwarded into extra_body for openai/azure/openai-compatible providers
keep non-internal unknown params pass-through behavior unchanged
add regression tests for Azure GPT-5.2 optional param construction path

Why

Azure GPT-5.2 chat rejects vector_store_ids as unknown top-level input when it leaks via extra_body passthrough.

Fixes #22898.

Test

poetry run pytest tests/test_litellm/test_utils.py -k "vector_store_ids or get_optional_params_azure_gpt5" -vv
poetry run pytest tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py -vv

Changed files

litellm/utils.py (modified, +9/-1)
tests/test_litellm/test_utils.py (modified, +37/-0)

Code Example

import litellm

resp = litellm.completion(
    model="gpt-5.2",
    custom_llm_provider="azure",
    messages=[{"role": "user", "content": "hi"}],
    temperature=1,
    vector_store_ids=["vs_123"],
)

RAW_BUFFERClick to expand / collapse

Summary

When routing Azure GPT-5.2 chat requests through LiteLLM, vector_store_ids can leak into provider payload via extra_body, causing Azure to reject the request with:

litellm.BadRequestError: AzureException BadRequestError - Unknown parameter: 'vector_store_ids'

Repro

import litellm

resp = litellm.completion(
    model="gpt-5.2",
    custom_llm_provider="azure",
    messages=[{"role": "user", "content": "hi"}],
    temperature=1,
    vector_store_ids=["vs_123"],
)

Observed: request fails with unknown parameter vector_store_ids.

Root cause

Proposed fix

Block internal vector store orchestration keys from extra_body pass-through:

vector_store_id
vector_store_ids

Validation

Added tests (local branch):

openai-compatible passthrough should not forward vector store keys.
get_optional_params(model=\"gpt-5.2\", custom_llm_provider=\"azure\", vector_store_ids=[...]) should not include vector_store_ids in extra_body.

Happy to open PR with code + tests.

extent analysis

Fix Plan

To fix the issue, we need to modify the add_provider_specific_params_to_optional_params function in litellm/utils.py to exclude internal vector store orchestration keys from the extra_body pass-through.

Code Changes

We can achieve this by adding a simple check to filter out the unwanted keys:

def add_provider_specific_params_to_optional_params(params, provider):
    # ... existing code ...
    extra_body = {}
    for key, value in params.items():
        if key in ["vector_store_id", "vector_store_ids"] or key.startswith("_"):
            continue
        extra_body[key] = value
    # ... existing code ...

Alternatively, we can use a more explicit approach by defining a set of allowed keys:

ALLOWED_KEYS = set(["temperature", "max_tokens", ...])  # add all allowed keys here

def add_provider_specific_params_to_optional_params(params, provider):
    # ... existing code ...
    extra_body = {}
    for key, value in params.items():
        if key in ALLOWED_KEYS:
            extra_body[key] = value
    # ... existing code ...

Verification

To verify the fix, we can run the added tests:

Test that openai-compatible passthrough does not forward vector store keys.
Test that get_optional_params does not include vector_store_ids in extra_body for the gpt-5.2 model with azure provider.

Example Use Case

After applying the fix, the following code should work as expected:

resp = litellm.completion(
    model="gpt-5.2",
    custom_llm_provider="azure",
    messages=[{"role": "user", "content": "hi"}],
    temperature=1,
    vector_store_ids=["vs_123"],
)

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

#api #ssr #installation #tensor shape #vector store #GPU setup #container setup #orchestration issue #cache issue

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Data

Security

Network

Code

UI/UX

Text

System

Multimedia

Protocol

API

Engineering

litellm - ✅(Solved) Fix [Bug]: Azure GPT-5.2 chat rejects leaked vector_store_ids from extra_body passthrough [1 pull requests, 4 comments, 4 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Fix Action

Fixed

PR fix notes

PR #22899: fix(azure): prevent vector_store_ids leak into extra_body passthrough

Description (problem / solution / changelog)

Summary

Why

Test

Changed files

Code Example

Summary

Repro

Root cause

Proposed fix

Validation

extent analysis

Fix Plan

Code Changes

Verification

Example Use Case

Still need to ship something?

TRENDING

litellm - ✅(Solved) Fix [Bug]: Azure GPT-5.2 chat rejects leaked vector_store_ids from extra_body passthrough [1 pull requests, 4 comments, 4 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Fix Action

Fixed

PR fix notes

PR #22899: fix(azure): prevent vector_store_ids leak into extra_body passthrough

Description (problem / solution / changelog)

Summary

Why

Test

Changed files

Code Example

Summary

Repro

Root cause

Proposed fix

Validation

extent analysis

Fix Plan

Code Changes

Verification

Example Use Case

Still need to ship something?

RELATED_DISCOVERY

TRENDING