mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-07-28 18:19:28 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 6

fix(deepseek): drop retired models from picker and provider defaults

Browse source 
Stop offering deepseek-chat/reasoner in the static catalog and point
fallback/aux defaults at the permanent v4 IDs. Keep retired aliases in
a detection-only map so /model deepseek-chat still resolves to deepseek.
This commit is contained in:
HexLab98 2026-07-25 12:45:59 +07:00 • committed by Teknium
parent cc7c418b33
commit aff48958d3
4 changed files with 60 additions and 36 deletions
Download patch file Download diff file Expand all files Collapse all files

32
plugins/model-providers/deepseek/__init__.py
View file

@ -1,11 +1,11 @@
"""DeepSeek provider profile.
DeepSeek's V4 family (and the legacy ``deepseek-reasoner``) defaults to
thinking-mode ON when ``extra_body.thinking`` is unset. The API then returns
``reasoning_content`` and starts enforcing the contract that subsequent turns
echo it back; combined with how Hermes replays history this lands on the
notorious HTTP 400 ``reasoning_content must be passed back`` error after the
first tool call (#15700, #17212, #17825).
DeepSeek's V4 family defaults to thinking-mode ON when ``extra_body.thinking``
is unset. The API then returns ``reasoning_content`` and starts enforcing
the contract that subsequent turns echo it back; combined with how Hermes
replays history this lands on the notorious HTTP 400
``reasoning_content must be passed back`` error after the first tool call
(#15700, #17212, #17825).
This profile overrides :meth:`build_api_kwargs_extras` to mirror the Kimi /
Moonshot wire shape that DeepSeek's OpenAI-compat endpoint expects:
@ -13,8 +13,12 @@ Moonshot wire shape that DeepSeek's OpenAI-compat endpoint expects:
{"reasoning_effort": "<low|medium|high|max>",
"extra_body": {"thinking": {"type": "enabled" | "disabled"}}}
Non-thinking models (only ``deepseek-chat`` today, which is V3) are left as
no-ops so we don't perturb the V3 wire format.
Non-thinking models (``deepseek-v3-*`` variants) are left as no-ops so we
don't perturb the V3 wire format.
The legacy aliases ``deepseek-chat`` / ``deepseek-reasoner`` were retired on
2026-07-24. Use ``deepseek-v4-flash`` or ``deepseek-v4-pro``; Hermes remaps
the retired IDs in ``hermes_cli.model_normalize``.
"""
from __future__ import annotations
@ -29,8 +33,8 @@ def _model_supports_thinking(model: str | None) -> bool:
"""DeepSeek thinking-capable model families.
Currently covers the V4 family (``deepseek-v4-pro``, ``deepseek-v4-flash``,
and any future ``deepseek-v4-*`` variants) and the legacy
``deepseek-reasoner`` (R1). ``deepseek-chat`` is V3 with no thinking mode.
and any future ``deepseek-v4-*`` variants). Retired aliases are remapped
before requests leave Hermes, so they are not listed here.
"""
m = (model or "").strip().lower()
if not m:
@ -39,8 +43,6 @@ def _model_supports_thinking(model: str | None) -> bool:
# deepseek-v4-*, deepseek-v5-*, etc. — every V4+ generation has
# thinking. v3 explicitly excluded.
return True
if m == "deepseek-reasoner":
return True
return False
@ -90,11 +92,11 @@ deepseek = DeepSeekProfile(
description="DeepSeek — native DeepSeek API",
signup_url="https://platform.deepseek.com/",
fallback_models=(
"deepseek-chat",
"deepseek-reasoner",
"deepseek-v4-pro",
"deepseek-v4-flash",
),
base_url="https://api.deepseek.com/v1",
default_aux_model="deepseek-chat",
default_aux_model="deepseek-v4-flash",
)
register_provider(deepseek)