Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

fix(vlm): forward extra request body to Volcengine (#5432)

Qin Haojie1f4f7039fc39
diff --git a/openviking/models/vlm/backends/volcengine_vlm.py b/openviking/models/vlm/backends/volcengine_vlm.py
index fcab417a2..a82339084 100644
--- a/openviking/models/vlm/backends/volcengine_vlm.py
+++ b/openviking/models/vlm/backends/volcengine_vlm.py
@@ -207,6 +207,8 @@ class VolcEngineVLM(OpenAIVLM):
"thinking": {"type": "disabled" if not effective_thinking else "enabled"},
"extra_headers": build_volcengine_request_headers(self.extra_headers),
}
+ if self.extra_request_body:
+ kwargs["extra_body"] = dict(self.extra_request_body)
if self.max_tokens is not None:
kwargs["max_tokens"] = self.max_tokens
if tools:
@@ -249,6 +251,8 @@ class VolcEngineVLM(OpenAIVLM):
"thinking": {"type": "disabled" if not effective_thinking else "enabled"},
"extra_headers": build_volcengine_request_headers(self.extra_headers),
}
+ if self.extra_request_body:
+ kwargs["extra_body"] = dict(self.extra_request_body)
if effective_max_tokens is not None:
kwargs["max_tokens"] = effective_max_tokens
if tools:
@@ -432,6 +436,8 @@ class VolcEngineVLM(OpenAIVLM):
"thinking": {"type": "disabled" if not effective_thinking else "enabled"},
"extra_headers": build_volcengine_request_headers(self.extra_headers),
}
+ if self.extra_request_body:
+ kwargs["extra_body"] = dict(self.extra_request_body)
if self.max_tokens is not None:
kwargs["max_tokens"] = self.max_tokens
if tools:
@@ -480,6 +486,8 @@ class VolcEngineVLM(OpenAIVLM):
"thinking": {"type": "disabled" if not effective_thinking else "enabled"},
"extra_headers": build_volcengine_request_headers(self.extra_headers),
}
+ if self.extra_request_body:
+ kwargs["extra_body"] = dict(self.extra_request_body)
if self.max_tokens is not None:
kwargs["max_tokens"] = self.max_tokens
if tools:
diff --git a/tests/unit/test_extra_headers_vlm.py b/tests/unit/test_extra_headers_vlm.py
index 4f7251bde..d5bc26700 100644
--- a/tests/unit/test_extra_headers_vlm.py
+++ b/tests/unit/test_extra_headers_vlm.py
@@ -10,7 +10,8 @@ from unittest.mock import AsyncMock, MagicMock, patch
import httpx
import pytest
from litellm.llms.ollama.chat.transformation import OllamaChatConfig
-from openai import OpenAI
+from openai import AsyncOpenAI, OpenAI
+from volcenginesdkarkruntime import Ark, AsyncArk
from volcenginesdkarkruntime._exceptions import ArkAPIStatusError
from openviking.models.vlm.backends.litellm_vlm import (
@@ -487,35 +488,68 @@ class TestVLMBaseExtraHeaders:
class TestVLMExtraRequestBody:
"""Test provider-specific VLM request body passthrough."""
- @patch("openviking.models.vlm.backends.openai_vlm.openai.OpenAI")
- def test_completion_body_overrides_configured_reasoning_effort(self, mock_openai_class):
+ @pytest.mark.parametrize(
+ ("provider", "model", "sync_client_cls", "async_client_cls", "extra_body"),
+ [
+ (
+ "glm",
+ "glm-5.3-flash",
+ OpenAI,
+ AsyncOpenAI,
+ {"reasoning_effort": "high", "seed": 7},
+ ),
+ (
+ "volcengine",
+ "ep-test",
+ Ark,
+ AsyncArk,
+ {"service_tier": "flex", "thinking": {"type": "disabled"}},
+ ),
+ ],
+ )
+ async def test_completion_body_overrides_configured_defaults(
+ self, monkeypatch, provider, model, sync_client_cls, async_client_cls, extra_body
+ ):
send = MagicMock(
- return_value=httpx.Response(
+ side_effect=lambda request: httpx.Response(
200,
json={
"choices": [{"message": {"content": "ok"}, "finish_reason": "stop"}],
},
)
)
- with OpenAI(
+ vlm = VLMConfig(
+ provider=provider,
+ model=model,
+ api_key="sk-test",
+ reasoning_effort="low",
+ thinking=True,
+ extra_request_body=extra_body,
+ ).get_vlm_instance()
+
+ with sync_client_cls(
api_key="sk-test",
http_client=httpx.Client(transport=httpx.MockTransport(send)),
) as client:
- mock_openai_class.return_value = client
- vlm = VLMConfig(
- provider="glm",
- model="glm-5.3-flash",
- api_key="sk-test",
- reasoning_effort="low",
- extra_request_body={"reasoning_effort": "high", "seed": 7},
- ).get_vlm_instance()
-
+ monkeypatch.setattr(vlm, "get_client", lambda: client)
assert vlm.get_completion("hello") == "ok"
assert vlm.get_vision_completion("describe", images=[b"\x89PNG\r\n\x1a\n"]) == "ok"
+ async with async_client_cls(
+ api_key="sk-test",
+ http_client=httpx.AsyncClient(transport=httpx.MockTransport(send)),
+ ) as async_client:
+ monkeypatch.setattr(vlm, "get_async_client", lambda: async_client)
+ assert await vlm.get_completion_async("hello") == "ok"
+ assert (
+ await vlm.get_vision_completion_async("describe", images=[b"\x89PNG\r\n\x1a\n"])
+ == "ok"
+ )
+
bodies = [json.loads(call.args[0].content) for call in send.call_args_list]
- assert [body["reasoning_effort"] for body in bodies] == ["high", "high"]
- assert [body["seed"] for body in bodies] == [7, 7]
+ assert len(bodies) == 4
+ for body in bodies:
+ assert {key: body[key] for key in extra_body} == extra_body
@patch("openviking.models.vlm.backends.openai_vlm.openai.OpenAI")
def test_dashscope_thinking_merges_with_extra_request_body(self, mock_openai_class):