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: remove gating from auto mode (#9821)

cthomas1a1846d769df
diff --git a/letta/agents/base_agent.py b/letta/agents/base_agent.py
index c81f760d4..3e6019a4b 100644
--- a/letta/agents/base_agent.py
+++ b/letta/agents/base_agent.py
@@ -56,7 +56,6 @@ class BaseAgent(ABC):
input_messages: List[MessageCreate],
max_steps: int = DEFAULT_MAX_STEPS,
run_id: Optional[str] = None,
- auto_mode_enabled: bool = False,
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
"""
diff --git a/letta/agents/base_agent_v2.py b/letta/agents/base_agent_v2.py
index 0edbee0ce..515edb76b 100644
--- a/letta/agents/base_agent_v2.py
+++ b/letta/agents/base_agent_v2.py
@@ -53,7 +53,6 @@ class BaseAgentV2(ABC):
request_start_timestamp_ns: int | None = None,
client_tools: list["ClientToolSchema"] | None = None,
include_compaction_messages: bool = False, # Not used in V2, but accepted for API compatibility
- auto_mode_enabled: bool = False, # Not used in V2, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
"""
@@ -79,7 +78,6 @@ class BaseAgentV2(ABC):
conversation_id: str | None = None,
client_tools: list["ClientToolSchema"] | None = None,
include_compaction_messages: bool = False, # Not used in V2, but accepted for API compatibility
- auto_mode_enabled: bool = False, # Not used in V2, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> AsyncGenerator[LettaMessage | LegacyLettaMessage | MessageStreamStatus, None]:
"""
diff --git a/letta/agents/letta_agent_v2.py b/letta/agents/letta_agent_v2.py
index b958c9fde..13fb9b07f 100644
--- a/letta/agents/letta_agent_v2.py
+++ b/letta/agents/letta_agent_v2.py
@@ -186,7 +186,6 @@ class LettaAgentV2(BaseAgentV2):
request_start_timestamp_ns: int | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False, # Not used in V2, but accepted for API compatibility
- auto_mode_enabled: bool = False, # Not used in V2, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
"""
@@ -293,7 +292,6 @@ class LettaAgentV2(BaseAgentV2):
conversation_id: str | None = None, # Not used in V2, but accepted for API compatibility
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False, # Not used in V2, but accepted for API compatibility
- auto_mode_enabled: bool = False, # Not used in V2, but accepted for API compatibility
billing_context: BillingContext | None = None,
) -> AsyncGenerator[str, None]:
"""
diff --git a/letta/agents/letta_agent_v3.py b/letta/agents/letta_agent_v3.py
index c92f452ae..429e4e441 100644
--- a/letta/agents/letta_agent_v3.py
+++ b/letta/agents/letta_agent_v3.py
@@ -161,7 +161,6 @@ class LettaAgentV3(LettaAgentV2):
conversation_id: str | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
"""
@@ -273,7 +272,6 @@ class LettaAgentV3(LettaAgentV2):
include_return_message_types=include_return_message_types,
request_start_timestamp_ns=request_start_timestamp_ns,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
)
input_messages_to_persist = [] # clear after first step
@@ -378,7 +376,6 @@ class LettaAgentV3(LettaAgentV2):
conversation_id: str | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
billing_context: BillingContext | None = None,
) -> AsyncGenerator[str, None]:
"""
@@ -507,7 +504,6 @@ class LettaAgentV3(LettaAgentV2):
include_return_message_types=include_return_message_types,
request_start_timestamp_ns=request_start_timestamp_ns,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
)
input_messages_to_persist = [] # clear after first step
async for chunk in response:
@@ -807,7 +803,6 @@ class LettaAgentV3(LettaAgentV2):
dry_run: bool = False,
enforce_run_id_set: bool = True,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
) -> AsyncGenerator[LettaMessage | dict, None]:
"""
Execute a single agent step (one LLM call and tool execution).
@@ -948,7 +943,6 @@ class LettaAgentV3(LettaAgentV2):
if is_auto_mode:
routing_client = await get_llm_routing_client()
active_llm_config, is_primary, primary_handle = await routing_client.resolve_auto_mode_config(
- auto_mode_enabled=auto_mode_enabled,
stored_llm_config=self.agent_state.llm_config,
actor=self.actor,
)
diff --git a/letta/groups/sleeptime_multi_agent_v2.py b/letta/groups/sleeptime_multi_agent_v2.py
index fb833b8d8..e842e0b7f 100644
--- a/letta/groups/sleeptime_multi_agent_v2.py
+++ b/letta/groups/sleeptime_multi_agent_v2.py
@@ -70,7 +70,6 @@ class SleeptimeMultiAgentV2(BaseAgent):
use_assistant_message: bool = True,
request_start_timestamp_ns: int | None = None,
include_return_message_types: list[MessageType] | None = None,
- auto_mode_enabled: bool = False, # Not used in multi-agent, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
run_ids = []
diff --git a/letta/groups/sleeptime_multi_agent_v3.py b/letta/groups/sleeptime_multi_agent_v3.py
index 4c1b95789..7a148eada 100644
--- a/letta/groups/sleeptime_multi_agent_v3.py
+++ b/letta/groups/sleeptime_multi_agent_v3.py
@@ -48,7 +48,6 @@ class SleeptimeMultiAgentV3(LettaAgentV2):
request_start_timestamp_ns: int | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False, # Not used in multi-agent, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
self.run_ids = []
@@ -65,7 +64,6 @@ class SleeptimeMultiAgentV3(LettaAgentV2):
request_start_timestamp_ns=request_start_timestamp_ns,
client_tools=client_tools,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
)
@@ -86,7 +84,6 @@ class SleeptimeMultiAgentV3(LettaAgentV2):
include_return_message_types: list[MessageType] | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
billing_context: "BillingContext | None" = None,
) -> AsyncGenerator[str, None]:
self.run_ids = []
@@ -106,7 +103,6 @@ class SleeptimeMultiAgentV3(LettaAgentV2):
request_start_timestamp_ns=request_start_timestamp_ns,
client_tools=client_tools,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
):
yield chunk
diff --git a/letta/groups/sleeptime_multi_agent_v4.py b/letta/groups/sleeptime_multi_agent_v4.py
index dc1ceaa61..150cd3f7f 100644
--- a/letta/groups/sleeptime_multi_agent_v4.py
+++ b/letta/groups/sleeptime_multi_agent_v4.py
@@ -48,7 +48,6 @@ class SleeptimeMultiAgentV4(LettaAgentV3):
conversation_id: str | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False, # Not used in multi-agent, but accepted for API compatibility
billing_context: "BillingContext | None" = None,
) -> LettaResponse:
self.run_ids = []
@@ -66,7 +65,6 @@ class SleeptimeMultiAgentV4(LettaAgentV3):
conversation_id=conversation_id,
client_tools=client_tools,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
)
@@ -87,7 +85,6 @@ class SleeptimeMultiAgentV4(LettaAgentV3):
conversation_id: str | None = None,
client_tools: list[ClientToolSchema] | None = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
billing_context: "BillingContext | None" = None,
) -> AsyncGenerator[str, None]:
self.run_ids = []
@@ -108,7 +105,6 @@ class SleeptimeMultiAgentV4(LettaAgentV3):
conversation_id=conversation_id,
client_tools=client_tools,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
):
yield chunk
diff --git a/letta/server/rest_api/dependencies.py b/letta/server/rest_api/dependencies.py
index 15772d754..66ea43bb4 100644
--- a/letta/server/rest_api/dependencies.py
+++ b/letta/server/rest_api/dependencies.py
@@ -20,7 +20,6 @@ class ExperimentalParams(BaseModel):
letta_v1_agent: Optional[bool] = None
letta_v1_agent_message_async: Optional[bool] = None
modal_sandbox: Optional[bool] = None
- auto_mode: Optional[bool] = None
class HeaderParams(BaseModel):
@@ -47,7 +46,6 @@ def get_headers(
None, alias="X-Experimental-Letta-V1-Agent-Message-Async", include_in_schema=False
),
modal_sandbox: Optional[str] = Header(None, alias="X-Experimental-Modal-Sandbox", include_in_schema=False),
- auto_mode: Optional[str] = Header(None, alias="X-Experimental-Auto-Mode", include_in_schema=False),
billing_plan_type: Optional[str] = Header(None, alias="X-Billing-Plan-Type", include_in_schema=False),
billing_cost_source: Optional[str] = Header(None, alias="X-Billing-Cost-Source", include_in_schema=False),
billing_customer_id: Optional[str] = Header(None, alias="X-Billing-Customer-Id", include_in_schema=False),
@@ -71,7 +69,6 @@ def get_headers(
letta_v1_agent=(letta_v1_agent == "true") if letta_v1_agent else None,
letta_v1_agent_message_async=(letta_v1_agent_message_async == "true") if letta_v1_agent_message_async else None,
modal_sandbox=(modal_sandbox == "true") if modal_sandbox else None,
- auto_mode=(auto_mode == "true") if auto_mode else None,
),
billing_context=BillingContext(
plan_type=billing_plan_type,
diff --git a/letta/server/rest_api/routers/v1/agents.py b/letta/server/rest_api/routers/v1/agents.py
index 962728d1b..dce472409 100644
--- a/letta/server/rest_api/routers/v1/agents.py
+++ b/letta/server/rest_api/routers/v1/agents.py
@@ -1692,14 +1692,12 @@ async def send_message(
actor = await server.user_manager.get_actor_or_default_async(actor_id=headers.actor_id)
if request.streaming and is_1_0_sdk:
- auto_mode_enabled = bool(headers.experimental_params and headers.experimental_params.auto_mode)
streaming_service = StreamingService(server)
run, result = await streaming_service.create_agent_stream(
agent_id=agent_id,
actor=actor,
request=request,
run_type="send_message",
- auto_mode_enabled=auto_mode_enabled,
billing_context=headers.billing_context,
)
return result
@@ -1761,7 +1759,6 @@ async def send_message(
}
)
- auto_mode_enabled = bool(headers.experimental_params and headers.experimental_params.auto_mode)
agent_loop = AgentLoop.load(agent_state=agent, actor=actor)
result = await agent_loop.step(
request.messages,
@@ -1772,7 +1769,6 @@ async def send_message(
include_return_message_types=request.include_return_message_types,
client_tools=request.client_tools,
include_compaction_messages=request.include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=headers.billing_context,
)
run_status = result.stop_reason.stop_reason.run_status
diff --git a/letta/services/llm_router/client.py b/letta/services/llm_router/client.py
index 87fa858c3..d47cc25f7 100644
--- a/letta/services/llm_router/client.py
+++ b/letta/services/llm_router/client.py
@@ -71,7 +71,6 @@ class LLMRoutingClient(LLMRoutingClientBase):
async def resolve_auto_mode_config(
self,
- auto_mode_enabled: bool,
stored_llm_config: "LLMConfig",
actor: "User",
) -> tuple["LLMConfig", bool, str]:
@@ -81,7 +80,6 @@ class LLMRoutingClient(LLMRoutingClientBase):
Preserves user-configured context_window and max_tokens from the stored config.
Args:
- auto_mode_enabled: Whether the experimental auto_mode header was set.
stored_llm_config: The agent's stored LLM config (with auto mode handle).
actor: The user actor for provider lookups.
@@ -89,22 +87,24 @@ class LLMRoutingClient(LLMRoutingClientBase):
Tuple of (resolved_config, is_primary, primary_handle).
"""
from letta.services.provider_manager import ProviderManager
+ from letta.settings import model_settings
provider_manager = ProviderManager()
primary_handle = AUTO_MODE_PRIMARY
- if auto_mode_enabled:
- if await self._is_healthy(primary_handle):
- config = _build_fireworks_config()
- config = config.model_copy(
- update={
- "context_window": min(stored_llm_config.context_window, config.context_window),
- "max_tokens": min(stored_llm_config.max_tokens, config.max_tokens),
- }
- )
- return config, True, primary_handle
-
- logger.warning(f"Auto mode: primary {primary_handle} unhealthy, using {AUTO_MODE_FALLBACK}")
+ if not model_settings.auto_mode_enabled:
+ logger.info(f"[AUTO MODE]: primary {primary_handle} disabled via kill switch, using {AUTO_MODE_FALLBACK}")
+ elif await self._is_healthy(primary_handle):
+ config = _build_fireworks_config()
+ config = config.model_copy(
+ update={
+ "context_window": min(stored_llm_config.context_window, config.context_window),
+ "max_tokens": min(stored_llm_config.max_tokens, config.max_tokens),
+ }
+ )
+ return config, True, primary_handle
+ else:
+ logger.warning(f"[AUTO MODE]: primary {primary_handle} unhealthy, using {AUTO_MODE_FALLBACK}")
config = await provider_manager.get_llm_config_from_handle(AUTO_MODE_FALLBACK, actor)
config = config.model_copy(
diff --git a/letta/services/llm_router/client_base.py b/letta/services/llm_router/client_base.py
index 3c69c8d01..3bb9f8159 100644
--- a/letta/services/llm_router/client_base.py
+++ b/letta/services/llm_router/client_base.py
@@ -20,14 +20,12 @@ class LLMRoutingClient:
async def resolve_auto_mode_config(
self,
- auto_mode_enabled: bool,
stored_llm_config: "LLMConfig",
actor: "User",
) -> tuple["LLMConfig", bool, str]:
"""Resolve an auto mode handle to an actual model config.
Args:
- auto_mode_enabled: Whether the experimental auto_mode header was set.
stored_llm_config: The agent's stored LLM config (with auto mode handle).
actor: The user actor for provider lookups.
diff --git a/letta/services/streaming_service.py b/letta/services/streaming_service.py
index be925a2d5..18ecbdc8b 100644
--- a/letta/services/streaming_service.py
+++ b/letta/services/streaming_service.py
@@ -81,7 +81,6 @@ class StreamingService:
run_type: str = "streaming",
conversation_id: Optional[str] = None,
should_lock: bool = False,
- auto_mode_enabled: bool = False,
billing_context: "BillingContext | None" = None,
) -> tuple[Optional[PydanticRun], Union[StreamingResponse, LettaResponse]]:
"""
@@ -191,7 +190,6 @@ class StreamingService:
lock_key=lock_key, # For lock release (may differ from conversation_id)
client_tools=request.client_tools,
include_compaction_messages=request.include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
route_class=route_class,
is_background=request.background,
@@ -363,7 +361,6 @@ class StreamingService:
lock_key: Optional[str] = None,
client_tools: Optional[list[ClientToolSchema]] = None,
include_compaction_messages: bool = False,
- auto_mode_enabled: bool = False,
billing_context: BillingContext | None = None,
route_class: str = "foreground",
is_background: bool = False,
@@ -401,7 +398,6 @@ class StreamingService:
conversation_id=conversation_id,
client_tools=client_tools,
include_compaction_messages=include_compaction_messages,
- auto_mode_enabled=auto_mode_enabled,
billing_context=billing_context,
)