hermes-admin/hermes-agent
1
0
Fork 0
Code Issues Pull Requests Actions 30 Packages Projects Releases Wiki Activity

test(custom): pin think=false to Ollama URLs, omit it for Mistral

Browse Source 
Cover the Mistral extra_forbidden case and keep the Ollama dual-emission
contract (think=false + reasoning_effort=none) on port 11434 / ollama hosts.
This commit is contained in:
xxxigm
2026-08-29 22:30:44 +07:00
committed by kshitij
parent 31f0336da7
commit 6ba8308309
3 changed files with 86 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
tests/agent/transports/test_chat_completions.py
View File

@@ -297,9 +297,24 @@ class TestChatCompletionsBuildKwargs:
model="qwen3", messages=msgs,
provider_profile=profile,
reasoning_config={"effort": "none"},
base_url="http://127.0.0.1:11434/v1",
)
assert kw["extra_body"]["think"] is False
def test_custom_omits_think_on_mistral(self, transport):
from providers import get_provider_profile
profile = get_provider_profile("custom")
msgs = [{"role": "user", "content": "Hi"}]
kw = transport.build_kwargs(
model="mistral-small-latest",
messages=msgs,
provider_profile=profile,
reasoning_config={"effort": "none"},
base_url="https://api.mistral.ai/v1",
)
assert kw.get("extra_body", {}).get("think") is None
assert kw.get("reasoning_effort") == "none"
def test_gemini_openai_compat_flash_reasoning_maps_to_nested_google_thinking_config(self, transport):

67
tests/plugins/model_providers/test_custom_profile.py
View File

@@ -7,12 +7,13 @@ nothing when reasoning was *enabled*, so a configured ``reasoning_effort``
was silently dropped for every custom endpoint.
These tests pin the wire-shape contract:
- disabled → extra_body.think = False
- enabled + effort → top-level reasoning_effort (native OpenAI-compat
- disabled on Ollama → extra_body.think = False + reasoning_effort=none
- disabled elsewhere → reasoning_effort=none, no think (strict APIs 422)
- enabled + effort → top-level reasoning_effort (native OpenAI-compat
format GLM/ARK expect), passed through verbatim
including ``max``/``xhigh``
- enabled + no effort → nothing emitted (endpoint's server default applies)
- ollama_num_ctx → extra_body.options.num_ctx, orthogonal to reasoning
- enabled + no effort → nothing emitted (endpoint's server default applies)
- ollama_num_ctx → extra_body.options.num_ctx, orthogonal to reasoning
"""
from __future__ import annotations
@@ -49,23 +50,71 @@ class TestCustomReasoningWireShape:
assert tl == {}
def test_disabled_sends_think_false(self, custom_profile):
"""enabled=False → reasoning_effort='none' top-level + think=False.
"""enabled=False on an Ollama URL → reasoning_effort='none' + think=False.
Both fields are required: Ollama's /v1/chat/completions silently
Both fields are required on Ollama: /v1/chat/completions silently
ignores extra_body.think (only /api/chat honours it — ollama#14820)
but respects top-level reasoning_effort (#25758). think=False stays
for proxies and the native /api/chat path.
"""
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"enabled": False}, model="glm-5.2"
reasoning_config={"enabled": False},
model="qwen3",
base_url="http://127.0.0.1:11434/v1",
)
assert eb == {"think": False}
assert tl == {"reasoning_effort": "none"}
def test_effort_none_sends_think_false(self, custom_profile):
"""effort='none' is the disable alias → same dual emission."""
"""effort='none' is the disable alias → same dual emission on Ollama."""
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"enabled": True, "effort": "none"}, model="glm-5.2"
reasoning_config={"enabled": True, "effort": "none"},
model="qwen3",
base_url="http://localhost:11434/v1",
)
assert eb == {"think": False}
assert tl == {"reasoning_effort": "none"}
def test_disabled_omits_think_on_mistral(self, custom_profile):
"""Strict OpenAI-compat hosts forbid extra ``think`` (HTTP 422)."""
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"enabled": True, "effort": "none"},
model="mistral-small-latest",
base_url="https://api.mistral.ai/v1",
)
assert "think" not in eb
assert tl == {"reasoning_effort": "none"}
def test_disabled_omits_think_without_base_url(self, custom_profile):
"""Unknown custom endpoint — do not send the Ollama-only flag."""
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"enabled": False}, model="glm-5.2"
)
assert "think" not in eb
assert tl == {"reasoning_effort": "none"}
@pytest.mark.parametrize(
"base_url",
[
"http://127.0.0.1:8080/v1",
"http://localhost:1234/v1",
"https://api.groq.com/openai/v1",
],
)
def test_disabled_omits_think_on_non_ollama_relays(self, custom_profile, base_url):
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"effort": "none"},
model="llama3",
base_url=base_url,
)
assert "think" not in eb
assert tl == {"reasoning_effort": "none"}
def test_disabled_sends_think_false_on_ollama_cloud_host(self, custom_profile):
eb, tl = custom_profile.build_api_kwargs_extras(
reasoning_config={"enabled": False},
model="qwen3",
base_url="https://ollama.com/v1",
)
assert eb == {"think": False}
assert tl == {"reasoning_effort": "none"}

13
tests/providers/test_transport_parity.py
View File

@@ -168,5 +168,18 @@ class TestCustomOllamaParity:
tools=None,
provider_profile=get_provider_profile("custom"),
reasoning_config={"enabled": False, "effort": "none"},
base_url="http://127.0.0.1:11434/v1",
)
assert kw["extra_body"]["think"] is False
def test_think_omitted_for_mistral_custom(self, transport):
kw = transport.build_kwargs(
model="mistral-small-latest",
messages=_simple_messages(),
tools=None,
provider_profile=get_provider_profile("custom"),
reasoning_config={"enabled": False, "effort": "none"},
base_url="https://api.mistral.ai/v1",
)
assert kw.get("extra_body", {}).get("think") is None
assert kw.get("reasoning_effort") == "none"