patch anthropic parsing and improve error logging

letta/llm_api/anthropic.py

@@ -244,9 +244,9 @@ def convert_anthropic_response_to_chatcompletion(
     if isinstance(response_json["content"], list):
         if len(response_json["content"]) > 1:
             # inner mono + function call
-            assert len(response_json["content"]) == 2, response_json
-            assert response_json["content"][0]["type"] == "text", response_json
-            assert response_json["content"][1]["type"] == "tool_use", response_json
+            assert len(response_json["content"]) == 2, f"Unexpected content length: {response_json}"
+            assert response_json["content"][0]["type"] == "text", f"Expected text, got: {response_json['content'][0]}"
+            assert response_json["content"][1]["type"] == "tool_use", f"Expected tool_use, got: {response_json['content'][1]}"
             content = strip_xml_tags(string=response_json["content"][0]["text"], tag=inner_thoughts_xml_tag)
             tool_calls = [
                 ToolCall(
@@ -258,10 +258,31 @@ def convert_anthropic_response_to_chatcompletion(
                     ),
                 )
             ]
+        elif len(response_json["content"]) == 1:
+            # Only tool call or just inner mono
+            first_item = response_json["content"][0]
+            if first_item["type"] == "text":
+                # Just inner mono
+                content = strip_xml_tags(string=first_item["text"], tag=inner_thoughts_xml_tag)
+                tool_calls = None
+            elif first_item["type"] == "tool_use":
+                # Only tool call, no inner mono
+                content = None  # No inner mono to extract
+                tool_calls = [
+                    ToolCall(
+                        id=first_item["id"],
+                        type="function",
+                        function=FunctionCall(
+                            name=first_item["name"],
+                            arguments=json.dumps(first_item["input"], indent=2),
+                        ),
+                    )
+                ]
+            else:
+                raise ValueError(f"Unexpected type in content: {first_item}")
         else:
-            # Just inner mono
-            content = strip_xml_tags(string=response_json["content"][0]["text"], tag=inner_thoughts_xml_tag)
-            tool_calls = None
+            # Empty content list
+            raise ValueError(f"Empty content in response: {response_json}")
     else:
         raise RuntimeError("Unexpected type for content in response_json.")
 

letta/offline_memory_agent.py

@@ -18,8 +18,18 @@ def trigger_rethink_memory(agent_state: "AgentState", message: Optional[str]) ->
 
     """
     from letta import create_client
+    from letta.schemas.embedding_config import EmbeddingConfig
+    from letta.schemas.llm_config import LLMConfig
 
     client = create_client()
+    ANTHROPIC_CONFIG = LLMConfig(
+        model_endpoint_type="anthropic",
+        model_endpoint="https://api.anthropic.com/v1",
+        model="claude-3-5-haiku-20241022",
+        context_window=32000,
+    )
+    client.set_default_llm_config(ANTHROPIC_CONFIG)
+    client.set_default_embedding_config(EmbeddingConfig.default_config(model_name="letta"))
     agents = client.list_agents()
     for agent in agents:
         if agent.agent_type == "offline_memory_agent":
@@ -130,7 +140,7 @@ def __init__(
         # extras
         first_message_verify_mono: bool = False,
         max_memory_rethinks: int = 10,
-        initial_message_sequence: Optional[List[Message]] = None,  
+        initial_message_sequence: Optional[List[Message]] = None,
     ):
         super().__init__(interface, agent_state, user, initial_message_sequence=initial_message_sequence)
         self.first_message_verify_mono = first_message_verify_mono

letta/server/server.py

@@ -521,9 +521,8 @@ def _step(
             # save agent after step
             save_agent(letta_agent, self.ms)
 
-        except Exception as e:
-            logger.error(f"Error in server._step: {e}")
-            print(traceback.print_exc())
+        except Exception:
+            logger.exception("Error in server._step")
             raise
         finally:
             logger.debug("Calling step_yield()")

letta/services/tool_execution_sandbox.py

@@ -203,7 +203,7 @@ def run_local_dir_sandbox_venv(self, sbx_config: SandboxConfig, env: Dict[str, s
             func_result, stdout = self.parse_out_function_results_markers(result.stdout)
             func_return, agent_state = self.parse_best_effort(func_result)
             return SandboxRunResult(
-                func_return=func_return, 
+                func_return=func_return,
                 agent_state=agent_state,
                 stdout=[stdout],
                 stderr=[result.stderr],
@@ -229,7 +229,7 @@ def run_local_dir_sandbox_runpy(
 
         # Execute the temp file
         with self.temporary_env_vars(env_vars):
-            result = runpy.run_path(temp_file_path, init_globals=env_vars)
+            result = runpy.run_path(temp_file_path, init_globals=env_vars, run_name="__main__")
 
         # Fetch the result
         func_result = result.get(self.LOCAL_SANDBOX_RESULT_VAR_NAME)

tests/test_offline_memory_agent.py

@@ -21,8 +21,16 @@
 @pytest.fixture(scope="module")
 def client():
     client = create_client()
-    client.set_default_llm_config(LLMConfig.default_config("gpt-4o-mini"))
-    client.set_default_embedding_config(EmbeddingConfig.default_config(provider="openai"))
+    ANTHROPIC_CONFIG = LLMConfig(
+        model_endpoint_type="anthropic",
+        model_endpoint="https://api.anthropic.com/v1",
+        model="claude-3-5-haiku-20241022",
+        context_window=32000,
+    )
+    # client.set_default_llm_config(LLMConfig.default_config("gpt-4o-mini"))
+    # client.set_default_embedding_config(EmbeddingConfig.default_config(provider="openai"))
+    client.set_default_llm_config(ANTHROPIC_CONFIG)
+    client.set_default_embedding_config(EmbeddingConfig.default_config(model_name="letta"))
 
     yield client
 
@@ -32,6 +40,7 @@ def clear_agents(client):
     for agent in client.list_agents():
         client.delete_agent(agent.id)
 
+
 def test_ripple_edit_anthropic(client, mock_e2b_api_key_none):
     trigger_rethink_memory_tool = client.create_or_update_tool(trigger_rethink_memory)
 
@@ -57,19 +66,19 @@ def test_ripple_edit_anthropic(client, mock_e2b_api_key_none):
     new_memory = Block(name="rethink_memory_block", label="rethink_memory_block", value="[empty]", limit=2000)
     conversation_memory = BasicBlockMemory(blocks=[conversation_persona_block, conversation_human_block, fact_block, new_memory])
     offline_memory = BasicBlockMemory(blocks=[offline_persona_block, offline_human_block, fact_block, new_memory])
-
-    ANTHROPIC_CONFIG = LLMConfig(
-            model_endpoint_type="anthropic",
-            model_endpoint="https://api.anthropic.com/v1",
-            model="claude-3-5-haiku-20241022",
-            context_window=32000,
-        )
+    #
+    # ANTHROPIC_CONFIG = LLMConfig(
+    #         model_endpoint_type="anthropic",
+    #         model_endpoint="https://api.anthropic.com/v1",
+    #         model="claude-3-5-haiku-20241022",
+    #         context_window=32000,
+    #     )
     conversation_agent = client.create_agent(
         name="conversation_agent",
         agent_type=AgentType.memgpt_agent,
         system=gpt_system.get_system_text("memgpt_convo_only"),
-        llm_config=ANTHROPIC_CONFIG,
-        embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
+        # llm_config=ANTHROPIC_CONFIG,
+        # embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
         tools=["send_message", trigger_rethink_memory_tool.name],
         memory=conversation_memory,
         include_base_tools=False,
@@ -85,8 +94,8 @@ def test_ripple_edit_anthropic(client, mock_e2b_api_key_none):
         agent_type=AgentType.offline_memory_agent,
         system=gpt_system.get_system_text("memgpt_offline_memory"),
         memory=offline_memory,
-        llm_config=ANTHROPIC_CONFIG,
-        embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
+        # llm_config=ANTHROPIC_CONFIG,
+        # embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
         tools=[rethink_memory_tool.name, finish_rethinking_memory_tool.name],
         tool_rules=[TerminalToolRule(tool_name=finish_rethinking_memory_tool.name)],
         include_base_tools=False,
@@ -97,18 +106,15 @@ def test_ripple_edit_anthropic(client, mock_e2b_api_key_none):
         agent_id=conversation_agent.id, message="[trigger_rethink_memory]: Messi has now moved to playing for Inter Miami"
     )
     offline_memory_agent = client.get_agent(agent_id=offline_memory_agent.id)
-
-    import pdb; pdb.set_trace()
     assert offline_memory_agent.memory.get_block("rethink_memory_block").value != "[empty]"
     conversation_agent = client.get_agent(agent_id=conversation_agent.id)
     assert conversation_agent.memory.get_block("rethink_memory_block").value != "[empty]"
 
     # Clean up agent
-    '''
+    """
     client.delete_agent(conversation_agent.id)
     client.delete_agent(offline_memory_agent.id)
-    '''
-
+    """
 
 
 def test_ripple_edit(client, mock_e2b_api_key_none):
@@ -214,6 +220,7 @@ def test_chat_only_agent(client, mock_e2b_api_key_none):
     # Clean up agent
     client.delete_agent(chat_only_agent.id)
 
+
 def test_initial_message_sequence(client, mock_e2b_api_key_none):
     """
     Test that when we set the initial sequence to an empty list,
@@ -229,8 +236,6 @@ def test_initial_message_sequence(client, mock_e2b_api_key_none):
         initial_message_sequence=[],
     )
     assert offline_memory_agent is not None
-    assert len(offline_memory_agent.message_ids) == 1 # There should just the system message 
+    assert len(offline_memory_agent.message_ids) == 1  # There should just the system message
 
     client.delete_agent(offline_memory_agent.id)
-
-