stirrup 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

stirrup/__init__.py

@@ -35,6 +35,7 @@ from stirrup.core.models import (
     AssistantMessage,
     AudioContentBlock,
     ChatMessage,
+    EmptyParams,
     ImageContentBlock,
     LLMClient,
     SubAgentMetadata,
@@ -58,6 +59,7 @@ __all__ = [
     "AudioContentBlock",
     "ChatMessage",
     "ContextOverflowError",
+    "EmptyParams",
     "ImageContentBlock",
     "LLMClient",
     "SubAgentMetadata",

stirrup/clients/chat_completions_client.py

@@ -67,7 +67,6 @@ class ChatCompletionsClient(LLMClient):
         *,
         base_url: str | None = None,
         api_key: str | None = None,
-        supports_audio_input: bool = False,
         reasoning_effort: str | None = None,
         timeout: float | None = None,
         max_retries: int = 2,
@@ -82,7 +81,6 @@ class ChatCompletionsClient(LLMClient):
                 Use for OpenAI-compatible providers (e.g., 'http://localhost:8000/v1').
             api_key: API key for authentication. If None, reads from OPENROUTER_API_KEY
                 environment variable.
-            supports_audio_input: Whether the model supports audio inputs. Defaults to False.
             reasoning_effort: Reasoning effort level for extended thinking models
                 (e.g., 'low', 'medium', 'high'). Only used with o1/o3 style models.
             timeout: Request timeout in seconds. If None, uses OpenAI SDK default.
@@ -92,7 +90,6 @@ class ChatCompletionsClient(LLMClient):
         """
         self._model = model
         self._max_tokens = max_tokens
-        self._supports_audio_input = supports_audio_input
         self._reasoning_effort = reasoning_effort
         self._kwargs = kwargs or {}
 

stirrup/clients/litellm_client.py

@@ -7,7 +7,7 @@ Requires the litellm extra: `pip install stirrup[litellm]`
 """
 
 import logging
-from typing import Any
+from typing import Any, Literal
 
 try:
     from litellm import acompletion
@@ -38,6 +38,8 @@ __all__ = [
 
 LOGGER = logging.getLogger(__name__)
 
+type ReasoningEffort = Literal["none", "minimal", "low", "medium", "high", "xhigh", "default"]
+
 
 class LiteLLMClient(LLMClient):
     """LiteLLM-based client supporting multiple LLM providers with unified interface.
@@ -49,8 +51,8 @@ class LiteLLMClient(LLMClient):
         self,
         model_slug: str,
         max_tokens: int,
-        supports_audio_input: bool = False,
-        reasoning_effort: str | None = None,
+        api_key: str | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         kwargs: dict[str, Any] | None = None,
     ) -> None:
         """Initialize LiteLLM client with model configuration and capabilities.
@@ -58,15 +60,13 @@ class LiteLLMClient(LLMClient):
         Args:
             model_slug: Model identifier for LiteLLM (e.g., 'anthropic/claude-3-5-sonnet-20241022')
             max_tokens: Maximum context window size in tokens
-            supports_audio_input: Whether the model supports audio inputs
             reasoning_effort: Reasoning effort level for extended thinking models (e.g., 'medium', 'high')
             kwargs: Additional arguments to pass to LiteLLM completion calls
         """
         self._model_slug = model_slug
-        self._supports_video_input = False
-        self._supports_audio_input = supports_audio_input
         self._max_tokens = max_tokens
-        self._reasoning_effort = reasoning_effort
+        self._reasoning_effort: ReasoningEffort | None = reasoning_effort
+        self._api_key = api_key
         self._kwargs = kwargs or {}
 
     @property
@@ -92,6 +92,8 @@ class LiteLLMClient(LLMClient):
             tools=to_openai_tools(tools) if tools else None,
             tool_choice="auto" if tools else None,
             max_tokens=self._max_tokens,
+            reasoning_effort=self._reasoning_effort,
+            api_key=self._api_key,
             **self._kwargs,
         )
 
@@ -103,14 +105,20 @@ class LiteLLMClient(LLMClient):
             )
 
         msg = choice["message"]
-
         reasoning: Reasoning | None = None
         if getattr(msg, "reasoning_content", None) is not None:
             reasoning = Reasoning(content=msg.reasoning_content)
         if getattr(msg, "thinking_blocks", None) is not None and len(msg.thinking_blocks) > 0:
-            reasoning = Reasoning(
-                signature=msg.thinking_blocks[0]["signature"], content=msg.thinking_blocks[0]["content"]
-            )
+            if len(msg.thinking_blocks) > 1:
+                raise ValueError("Found multiple thinking blocks in the response")
+
+            signature = msg.thinking_blocks[0].get("thinking_signature", None)
+            content = msg.thinking_blocks[0].get("thinking", None)
+
+            if signature is None and content is None:
+                raise ValueError("Signature and content not found in the thinking block response")
+
+            reasoning = Reasoning(signature=signature, content=content)
 
         usage = r["usage"]
 
@@ -119,6 +127,7 @@ class LiteLLMClient(LLMClient):
                 tool_call_id=tc.get("id"),
                 name=tc["function"]["name"],
                 arguments=tc["function"].get("arguments", "") or "",
+                signature=tc.get("provider_specific_fields", {}).get("thought_signature", None),
             )
             for tc in (msg.get("tool_calls") or [])
         ]

stirrup/clients/utils.py

@@ -12,6 +12,7 @@ from stirrup.core.models import (
     AudioContentBlock,
     ChatMessage,
     Content,
+    EmptyParams,
     ImageContentBlock,
     SystemMessage,
     Tool,
@@ -47,7 +48,7 @@ def to_openai_tools(tools: dict[str, Tool]) -> list[dict[str, Any]]:
             "name": t.name,
             "description": t.description,
         }
-        if t.parameters is not None:
+        if t.parameters is not EmptyParams:
             function["parameters"] = t.parameters.model_json_schema()
         tool_payload: dict[str, Any] = {
             "type": "function",
@@ -139,6 +140,10 @@ def to_openai_messages(msgs: list[ChatMessage]) -> list[dict[str, Any]]:
                     tool_dict = tool.model_dump()
                     tool_dict["id"] = tool.tool_call_id
                     tool_dict["type"] = "function"
+                    if tool.signature is not None:
+                        tool_dict["provider_specific_fields"] = {
+                            "thought_signature": tool.signature,
+                        }
                     tool_dict["function"] = {
                         "name": tool.name,
                         "arguments": tool.arguments,

stirrup/constants.py

@@ -1,14 +1,18 @@
+from typing import Literal
+
 # Tool naming
-FINISH_TOOL_NAME = "finish"
+FINISH_TOOL_NAME: Literal["finish"] = "finish"
 
 # Agent execution limits
 AGENT_MAX_TURNS = 30  # Maximum agent turns before forced termination
 CONTEXT_SUMMARIZATION_CUTOFF = 0.7  # Context window usage threshold (0.0-1.0) that triggers message summarization
+TURNS_REMAINING_WARNING_THRESHOLD = 20
 
 # Media resolution limits
 RESOLUTION_1MP = 1_000_000  # 1 megapixel - default max resolution for images
 RESOLUTION_480P = 640 * 480  # 480p video resolution
 
 # Code execution
-SUBMISSION_SANDBOX_TIMEOUT = 60 * 10  # 10 minutes
+SANDBOX_TIMEOUT = 60 * 10  # 10 minutes
+SANDBOX_REQUEST_TIMEOUT = 60 * 3  # 3 minutes
 E2B_SANDBOX_TEMPLATE_ALIAS = "e2b-sandbox"