Skip to content

armgit / letta

Public Imported from GitHub
0

Platform for stateful agents: AI with advanced memory that can learn and self-improve over time.

feat: consolidate reasoning model checks (#3862)

cthomas639641137dbd
diff --git a/.github/scripts/model-sweep/model_sweep.py b/.github/scripts/model-sweep/model_sweep.py
index c92b93643..322b427b4 100644
--- a/.github/scripts/model-sweep/model_sweep.py
+++ b/.github/scripts/model-sweep/model_sweep.py
@@ -96,7 +96,7 @@ all_configs = [
"openai-gpt-4o-mini.json",
# "azure-gpt-4o-mini.json", # TODO: Re-enable on new agent loop
"claude-3-5-sonnet.json",
- "claude-3-7-sonnet.json",
+ "claude-4-sonnet-extended.json",
"claude-3-7-sonnet-extended.json",
"gemini-1.5-pro.json",
"gemini-2.5-flash-vertex.json",
diff --git a/.github/workflows/send-message-integration-tests.yaml b/.github/workflows/send-message-integration-tests.yaml
index eee96265d..795614a95 100644
--- a/.github/workflows/send-message-integration-tests.yaml
+++ b/.github/workflows/send-message-integration-tests.yaml
@@ -19,7 +19,7 @@ jobs:
- "openai-gpt-4o-mini.json"
- "azure-gpt-4o-mini.json"
- "claude-3-5-sonnet.json"
- - "claude-3-7-sonnet.json"
+ - "claude-4-sonnet-extended.json"
- "claude-3-7-sonnet-extended.json"
- "gemini-pro.json"
- "gemini-vertex.json"
diff --git a/letta/llm_api/anthropic_client.py b/letta/llm_api/anthropic_client.py
index 7b886dfca..7c6cc744e 100644
--- a/letta/llm_api/anthropic_client.py
+++ b/letta/llm_api/anthropic_client.py
@@ -182,7 +182,7 @@ class AnthropicClient(LLMClientBase):
}
# Extended Thinking
- if llm_config.enable_reasoner:
+ if self.is_reasoning_model(llm_config) and llm_config.enable_reasoner:
data["thinking"] = {
"type": "enabled",
"budget_tokens": llm_config.max_reasoning_tokens,
@@ -200,7 +200,7 @@ class AnthropicClient(LLMClientBase):
# Special case for summarization path
tools_for_request = None
tool_choice = None
- elif llm_config.enable_reasoner:
+ elif self.is_reasoning_model(llm_config) and llm_config.enable_reasoner:
# NOTE: reasoning models currently do not allow for `any`
tool_choice = {"type": "auto", "disable_parallel_tool_use": True}
tools_for_request = [OpenAITool(function=f) for f in tools]
@@ -296,6 +296,13 @@ class AnthropicClient(LLMClientBase):
token_count -= 8
return token_count
+ def is_reasoning_model(self, llm_config: LLMConfig) -> bool:
+ return (
+ llm_config.model.startswith("claude-3-7-sonnet")
+ or llm_config.model.startswith("claude-sonnet-4")
+ or llm_config.model.startswith("claude-opus-4")
+ )
+
@trace_method
def handle_llm_error(self, e: Exception) -> Exception:
if isinstance(e, anthropic.APITimeoutError):
diff --git a/letta/llm_api/google_vertex_client.py b/letta/llm_api/google_vertex_client.py
index 32ca5d69a..fa6cf17df 100644
--- a/letta/llm_api/google_vertex_client.py
+++ b/letta/llm_api/google_vertex_client.py
@@ -504,6 +504,9 @@ class GoogleVertexClient(LLMClientBase):
return 1
return 0
+ def is_reasoning_model(self, llm_config: LLMConfig) -> bool:
+ return llm_config.model.startswith("gemini-2.5-flash") or llm_config.model.startswith("gemini-2.5-pro")
+
@trace_method
def handle_llm_error(self, e: Exception) -> Exception:
# Fallback to base implementation
diff --git a/letta/llm_api/llm_client_base.py b/letta/llm_api/llm_client_base.py
index 81ab852bc..038caaa15 100644
--- a/letta/llm_api/llm_client_base.py
+++ b/letta/llm_api/llm_client_base.py
@@ -174,6 +174,10 @@ class LLMClientBase:
"""
raise NotImplementedError(f"Streaming is not supported for {llm_config.model_endpoint_type}")
+ @abstractmethod
+ def is_reasoning_model(self, llm_config: LLMConfig) -> bool:
+ raise NotImplementedError
+
@abstractmethod
def handle_llm_error(self, e: Exception) -> Exception:
"""
diff --git a/letta/llm_api/openai_client.py b/letta/llm_api/openai_client.py
index 2b54a6e35..159372d25 100644
--- a/letta/llm_api/openai_client.py
+++ b/letta/llm_api/openai_client.py
@@ -276,6 +276,9 @@ class OpenAIClient(LLMClientBase):
response: ChatCompletion = await client.chat.completions.create(**request_data)
return response.model_dump()
+ def is_reasoning_model(self, llm_config: LLMConfig) -> bool:
+ return is_openai_reasoning_model(llm_config.model)
+
@trace_method
def convert_response_to_chat_completion(
self,
@@ -298,7 +301,7 @@ class OpenAIClient(LLMClientBase):
)
# If we used a reasoning model, create a content part for the ommitted reasoning
- if is_openai_reasoning_model(llm_config.model):
+ if self.is_reasoning_model(llm_config):
chat_completion_response.choices[0].message.omitted_reasoning_content = True
return chat_completion_response
diff --git a/letta/schemas/llm_config.py b/letta/schemas/llm_config.py
index ecb2c6635..24b7902d5 100644
--- a/letta/schemas/llm_config.py
+++ b/letta/schemas/llm_config.py
@@ -94,6 +94,9 @@ class LLMConfig(BaseModel):
"""
model = values.get("model")
+ if model is None:
+ return values
+
# Define models where we want put_inner_thoughts_in_kwargs to be False
avoid_put_inner_thoughts_in_kwargs = ["gpt-4"]
@@ -107,25 +110,13 @@ class LLMConfig(BaseModel):
if is_openai_reasoning_model(model):
values["put_inner_thoughts_in_kwargs"] = False
- if values.get("enable_reasoner") and values.get("model_endpoint_type") == "anthropic":
+ if values.get("model_endpoint_type") == "anthropic" and (
+ model.startswith("claude-3-7-sonnet") or model.startswith("claude-sonnet-4") or model.startswith("claude-opus-4")
+ ):
values["put_inner_thoughts_in_kwargs"] = False
return values
- @model_validator(mode="after")
- def issue_warning_for_reasoning_constraints(self) -> "LLMConfig":
- if self.enable_reasoner:
- if self.max_reasoning_tokens is None:
- logger.warning("max_reasoning_tokens must be set when enable_reasoner is True")
- if self.max_tokens is not None and self.max_reasoning_tokens >= self.max_tokens:
- logger.warning("max_tokens must be greater than max_reasoning_tokens (thinking budget)")
- if self.put_inner_thoughts_in_kwargs:
- logger.debug("Extended thinking is not compatible with put_inner_thoughts_in_kwargs")
- elif self.max_reasoning_tokens and not self.enable_reasoner:
- logger.warning("model will not use reasoning unless enable_reasoner is set to True")
-
- return self
-
@classmethod
def default_config(cls, model_name: str):
"""
diff --git a/tests/configs/llm_model_configs/claude-3-7-sonnet.json b/tests/configs/llm_model_configs/claude-3-7-sonnet.json
deleted file mode 100644
index beecaa759..000000000
--- a/tests/configs/llm_model_configs/claude-3-7-sonnet.json
+++ /dev/null
@@ -1,8 +0,0 @@
-{
- "model": "claude-3-7-sonnet-20250219",
- "model_endpoint_type": "anthropic",
- "model_endpoint": "https://api.anthropic.com/v1",
- "model_wrapper": null,
- "context_window": 200000,
- "put_inner_thoughts_in_kwargs": true
-}
diff --git a/tests/configs/llm_model_configs/claude-4-sonnet.json b/tests/configs/llm_model_configs/claude-4-sonnet-extended.json
similarity index 65%
rename from tests/configs/llm_model_configs/claude-4-sonnet.json
rename to tests/configs/llm_model_configs/claude-4-sonnet-extended.json
index 12a582e01..e622c7888 100644
--- a/tests/configs/llm_model_configs/claude-4-sonnet.json
+++ b/tests/configs/llm_model_configs/claude-4-sonnet-extended.json
@@ -4,5 +4,7 @@
"model_endpoint": "https://api.anthropic.com/v1",
"model_wrapper": null,
"context_window": 200000,
- "put_inner_thoughts_in_kwargs": true
+ "put_inner_thoughts_in_kwargs": false,
+ "enable_reasoner": true,
+ "max_reasoning_tokens": 1024
}
diff --git a/tests/integration_test_send_message.py b/tests/integration_test_send_message.py
index ed44ed671..e07988411 100644
--- a/tests/integration_test_send_message.py
+++ b/tests/integration_test_send_message.py
@@ -118,9 +118,8 @@ all_configs = [
"openai-o3.json",
"openai-o4-mini.json",
"azure-gpt-4o-mini.json",
- "claude-4-sonnet.json",
+ "claude-4-sonnet-extended.json",
"claude-3-5-sonnet.json",
- "claude-3-7-sonnet.json",
"claude-3-7-sonnet-extended.json",
"bedrock-claude-4-sonnet.json",
"gemini-1.5-pro.json",