From 19e5eee11b35e7a90e95507cb8f659ebf67ee3fe Mon Sep 17 00:00:00 2001 From: Christian Munley Date: Fri, 26 Jun 2026 23:09:27 -0400 Subject: [PATCH 1/2] fix Signed-off-by: Christian Munley --- responses_api_agents/hermes_agent/app.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/responses_api_agents/hermes_agent/app.py b/responses_api_agents/hermes_agent/app.py index b8980bcda..09e8371e0 100644 --- a/responses_api_agents/hermes_agent/app.py +++ b/responses_api_agents/hermes_agent/app.py @@ -164,6 +164,7 @@ class HermesAgentConfig(BaseResponsesAPIAgentConfig): system_prompt: Optional[str] = None compression_enabled: bool = True compression_threshold: float = 0.85 + chat_template_kwargs_enabled: bool = True delegation_max_iterations: int = 50 checkpoints_enabled: bool = False @@ -280,7 +281,7 @@ async def responses( agent = AIAgent( base_url=base_url, - api_key="gym", # pragma: allowlist secret + api_key=os.environ.get("OPENAI_API_KEY", "gym"), # pragma: allowlist secret model=model_name, use_streaming=False, temperature=self.config.temperature, @@ -298,6 +299,8 @@ async def responses( def _patched_build_api_kwargs(api_messages): kw = _original_build_api_kwargs(api_messages) + if not self.config.chat_template_kwargs_enabled: + return kw ctk = kw.setdefault("extra_body", {}).setdefault("chat_template_kwargs", {}) ctk.setdefault("enable_thinking", True) ctk["truncate_history_thinking"] = False From a8c9b70279538c7c19545df1694701807b3aeea5 Mon Sep 17 00:00:00 2001 From: Christian Munley Date: Fri, 26 Jun 2026 23:12:31 -0400 Subject: [PATCH 2/2] cfg Signed-off-by: Christian Munley --- responses_api_agents/hermes_agent/app.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/responses_api_agents/hermes_agent/app.py b/responses_api_agents/hermes_agent/app.py index 09e8371e0..3ed561f9d 100644 --- a/responses_api_agents/hermes_agent/app.py +++ b/responses_api_agents/hermes_agent/app.py @@ -165,6 +165,7 @@ class HermesAgentConfig(BaseResponsesAPIAgentConfig): compression_enabled: bool = True compression_threshold: float = 0.85 chat_template_kwargs_enabled: bool = True + api_key: Optional[str] = None delegation_max_iterations: int = 50 checkpoints_enabled: bool = False @@ -281,7 +282,7 @@ async def responses( agent = AIAgent( base_url=base_url, - api_key=os.environ.get("OPENAI_API_KEY", "gym"), # pragma: allowlist secret + api_key=self.config.api_key or os.environ.get("OPENAI_API_KEY", "gym"), # pragma: allowlist secret model=model_name, use_streaming=False, temperature=self.config.temperature,