langtrace-python-sdk 3.3.11__py3-none-any.whl → 3.3.13__py3-none-any.whl

examples/langgraph_example/main.py

@@ -0,0 +1,173 @@
+from typing import TypedDict, Union, Annotated
+from langchain_core.agents import AgentAction, AgentFinish
+from langchain_core.tools import tool
+import operator
+from dotenv import load_dotenv
+from langchain_openai import ChatOpenAI
+
+from langchain import hub
+from langchain.agents import create_openai_tools_agent
+import json
+from langgraph.graph import StateGraph, END
+from langtrace_python_sdk import langtrace, with_langtrace_root_span
+
+load_dotenv()
+
+langtrace.init(write_spans_to_console=False)
+
+
+class AgentState(TypedDict):
+    input: str
+    agent_out: Union[AgentAction, AgentFinish, None]
+    intermediate_steps: Annotated[list[tuple[AgentAction, str]], operator.add]
+
+
+ehi_information = """Title: EHI: End-to-end Learning of Hierarchical Index for
+Efficient Dense Retrieval
+Summary: Dense embedding-based retrieval is now the industry
+standard for semantic search and ranking problems, like obtaining relevant web
+documents for a given query. Such techniques use a two-stage process: (a)
+contrastive learning to train a dual encoder to embed both the query and
+documents and (b) approximate nearest neighbor search (ANNS) for finding similar
+documents for a given query. These two stages are disjoint; the learned
+embeddings might be ill-suited for the ANNS method and vice-versa, leading to
+suboptimal performance. In this work, we propose End-to-end Hierarchical
+Indexing -- EHI -- that jointly learns both the embeddings and the ANNS
+structure to optimize retrieval performance. EHI uses a standard dual encoder
+model for embedding queries and documents while learning an inverted file index
+(IVF) style tree structure for efficient ANNS. To ensure stable and efficient
+learning of discrete tree-based ANNS structure, EHI introduces the notion of
+dense path embedding that captures the position of a query/document in the tree.
+We demonstrate the effectiveness of EHI on several benchmarks, including
+de-facto industry standard MS MARCO (Dev set and TREC DL19) datasets. For
+example, with the same compute budget, EHI outperforms state-of-the-art (SOTA)
+in by 0.6% (MRR@10) on MS MARCO dev set and by 4.2% (nDCG@10) on TREC DL19
+benchmarks.
+Author(s): Ramnath Kumar, Anshul Mittal, Nilesh Gupta, Aditya Kusupati,
+Inderjit Dhillon, Prateek Jain
+Source: https://arxiv.org/pdf/2310.08891.pdf"""
+
+
+@tool("search")
+def search_tool(query: str):
+    """Searches for information on the topic of artificial intelligence (AI).
+    Cannot be used to research any other topics. Search query must be provided
+    in natural language and be verbose."""
+    # this is a "RAG" emulator
+    return ehi_information
+
+
+@tool("final_answer")
+def final_answer_tool(answer: str, source: str):
+    """Returns a natural language response to the user in `answer`, and a
+    `source` which provides citations for where this information came from.
+    """
+    return ""
+
+
+llm = ChatOpenAI()
+prompt = hub.pull("hwchase17/openai-functions-agent")
+
+
+query_agent_runnable = create_openai_tools_agent(
+    llm=llm, tools=[final_answer_tool, search_tool], prompt=prompt
+)
+
+
+inputs = {"input": "what are EHI embeddings?", "intermediate_steps": []}
+
+agent_out = query_agent_runnable.invoke(inputs)
+
+
+def run_query_agent(state: list):
+    print("> run_query_agent")
+    agent_out = query_agent_runnable.invoke(state)
+    return {"agent_out": agent_out}
+
+
+def execute_search(state: list):
+    print("> execute_search")
+    action = state["agent_out"]
+    tool_call = action[-1].message_log[-1].additional_kwargs["tool_calls"][-1]
+    out = search_tool.invoke(json.loads(tool_call["function"]["arguments"]))
+    return {"intermediate_steps": [{"search": str(out)}]}
+
+
+def router(state: list):
+    print("> router")
+    if isinstance(state["agent_out"], list):
+        return state["agent_out"][-1].tool
+    else:
+        return "error"
+
+
+# finally, we will have a single LLM call that MUST use the final_answer structure
+final_answer_llm = llm.bind_tools([final_answer_tool], tool_choice="final_answer")
+
+
+# this forced final_answer LLM call will be used to structure output from our
+# RAG endpoint
+def rag_final_answer(state: list):
+    print("> final_answer")
+    query = state["input"]
+    context = state["intermediate_steps"][-1]
+
+    prompt = f"""You are a helpful assistant, answer the user's question using the
+    context provided.
+
+    CONTEXT: {context}
+
+    QUESTION: {query}
+    """
+    out = final_answer_llm.invoke(prompt)
+    function_call = out.additional_kwargs["tool_calls"][-1]["function"]["arguments"]
+    return {"agent_out": function_call}
+
+
+# we use the same forced final_answer LLM call to handle incorrectly formatted
+# output from our query_agent
+def handle_error(state: list):
+    print("> handle_error")
+    query = state["input"]
+    prompt = f"""You are a helpful assistant, answer the user's question.
+
+    QUESTION: {query}
+    """
+    out = final_answer_llm.invoke(prompt)
+    function_call = out.additional_kwargs["tool_calls"][-1]["function"]["arguments"]
+    return {"agent_out": function_call}
+
+
+@with_langtrace_root_span("run_graph")
+def run_graph():
+    graph = StateGraph(AgentState)
+
+    # we have four nodes that will consume our agent state and modify
+    # our agent state based on some internal process
+    graph.add_node("query_agent", run_query_agent)
+    graph.add_node("search", execute_search)
+    graph.add_node("error", handle_error)
+    graph.add_node("rag_final_answer", rag_final_answer)
+    # our graph will always begin with the query agent
+    graph.set_entry_point("query_agent")
+    # conditional edges are controlled by our router
+    graph.add_conditional_edges(
+        "query_agent",
+        router,
+        {
+            "search": "search",
+            "error": "error",
+            "final_answer": END,
+        },
+    )
+    graph.add_edge("search", "rag_final_answer")
+    graph.add_edge("error", END)
+    graph.add_edge("rag_final_answer", END)
+
+    runnable = graph.compile()
+
+    return runnable.invoke({"input": "what are EHI embeddings?"})
+
+
+if __name__ == "__main__":
+    run_graph()

langtrace_python_sdk/instrumentation/crewai/patch.py

@@ -223,7 +223,7 @@ class CrewAISpanAttributes:
         for task in tasks:
             self.crew["tasks"].append(
                 {
-                    "agent": task.agent.role,
+                    "agent": task.agent.role if task.agent else None,
                     "description": task.description,
                     "async_execution": task.async_execution,
                     "expected_output": task.expected_output,

langtrace_python_sdk/instrumentation/dspy/instrumentation.py

@@ -27,12 +27,12 @@ class DspyInstrumentation(BaseInstrumentor):
     The DspyInstrumentor class represents the DSPy instrumentation"""
 
     def instrumentation_dependencies(self) -> Collection[str]:
-        return ["dspy-ai >= 2.0.0"]
+        return ["dspy >= 2.0.0"]
 
     def _instrument(self, **kwargs):
         tracer_provider = kwargs.get("tracer_provider")
         tracer = get_tracer(__name__, "", tracer_provider)
-        version = v("dspy-ai")
+        version = v("dspy")
         _W(
             "dspy.teleprompt.bootstrap",
             "BootstrapFewShot.compile",

langtrace_python_sdk/instrumentation/langgraph/instrumentation.py

@@ -41,7 +41,8 @@ class LanggraphInstrumentation(BaseInstrumentor):
         # List of modules to patch, with their corresponding patch names
         modules_to_patch = [
             (
-                "langgraph.graph.graph",
+                "langgraph.graph.state",  # Updated module path
+                "StateGraph",  # Updated class name
                 [
                     "add_node",
                     "add_edge",
@@ -49,26 +50,20 @@ class LanggraphInstrumentation(BaseInstrumentor):
                     "set_finish_point",
                     "add_conditional_edges",
                 ],
-            ),
+            )
         ]
 
-        for module_name, methods in modules_to_patch:
-            module = importlib.import_module(module_name)
-            for name, obj in inspect.getmembers(
-                module,
-                lambda member: inspect.isclass(member)
-                and member.__module__ == module.__name__,
-            ):
-                for method_name, _ in inspect.getmembers(
-                    obj, predicate=inspect.isfunction
-                ):
-                    if method_name in methods:
-                        module = f"{name}.{method_name}"
-                        wrap_function_wrapper(
-                            module_name,
-                            module,
-                            patch_graph_methods(module, tracer, version),
-                        )
+        for module_name, class_name, methods in modules_to_patch:
+            for method_name in methods:
+                # Construct the correct path for the method
+                method_path = f"{class_name}.{method_name}"
+                wrap_function_wrapper(
+                    module_name,
+                    method_path,
+                    patch_graph_methods(
+                        f"{module_name}.{method_path}", tracer, version
+                    ),
+                )
 
     def _uninstrument(self, **kwargs):
         pass

langtrace_python_sdk/instrumentation/langgraph/patch.py

@@ -30,6 +30,7 @@ from langtrace_python_sdk.constants.instrumentation.common import (
 from importlib_metadata import version as v
 
 from langtrace_python_sdk.constants import LANGTRACE_SDK_NAME
+from langtrace_python_sdk.utils.llm import set_span_attributes
 
 
 def patch_graph_methods(method_name, tracer, version):
@@ -57,9 +58,7 @@ def patch_graph_methods(method_name, tracer, version):
             kind=SpanKind.CLIENT,
             context=set_span_in_context(trace.get_current_span()),
         ) as span:
-            for field, value in attributes.model_dump(by_alias=True).items():
-                if value is not None:
-                    span.set_attribute(field, value)
+            set_span_attributes(span, attributes)
             try:
                 # Attempt to call the original method
                 result = wrapped(*args, **kwargs)

langtrace_python_sdk/langtrace.py

@@ -275,7 +275,7 @@ def init(
         "weaviate-client": WeaviateInstrumentation(),
         "sqlalchemy": SQLAlchemyInstrumentor(),
         "ollama": OllamaInstrumentor(),
-        "dspy-ai": DspyInstrumentation(),
+        "dspy": DspyInstrumentation(),
         "crewai": CrewAIInstrumentation(),
         "vertexai": VertexAIInstrumentation(),
         "google-cloud-aiplatform": VertexAIInstrumentation(),

langtrace_python_sdk/version.py

@@ -1 +1 @@
-__version__ = "3.3.11"
+__version__ = "3.3.13"

{langtrace_python_sdk-3.3.11.dist-info → langtrace_python_sdk-3.3.13.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: langtrace-python-sdk
-Version: 3.3.11
+Version: 3.3.13
 Summary: Python SDK for LangTrace
 Project-URL: Homepage, https://github.com/Scale3-Labs/langtrace-python-sdk
 Author-email: Scale3 Labs <engineering@scale3labs.com>

{langtrace_python_sdk-3.3.11.dist-info → langtrace_python_sdk-3.3.13.dist-info}/RECORD

@@ -61,6 +61,7 @@ examples/langchain_example/langgraph_example.py,sha256=7C2a4Sg0PKbbab03CVkStO3Mz
 examples/langchain_example/langgraph_example_tools.py,sha256=rFwgQYRngeyCz9PuBxnllp5t5PIHk8d-UDKwCQTgkxw,4208
 examples/langchain_example/sagemaker.py,sha256=V-rTZRyaErHCuo3kfrrZD8AELHJVi3wF7n1YrixfF1s,2330
 examples/langchain_example/tool.py,sha256=8T8_IDbgA58XbsfyH5_xhA8ZKQfyfyFxF8wor-PsRjA,2556
+examples/langgraph_example/main.py,sha256=G1bvJ83KmBk6ibs6UFHNaLFTeg9-72dEy2UyhGd7ssI,6031
 examples/litellm_example/basic.py,sha256=UDbv6-SV7H5_Ogk_IOL22ZX4hv5I_CKCyEHl3r8pf7c,2338
 examples/litellm_example/config.yaml,sha256=kSAAspBhibtc4D7Abd2vYqm3uIqHR1kjE8nZrSTJqYQ,303
 examples/litellm_example/proxy_basic.py,sha256=glQvcQ3rYD1QTyQfTwmzlPdzGMiIceRhAzE3_O94_1U,366
@@ -104,8 +105,8 @@ examples/vertexai_example/main.py,sha256=gndId5X5ksD-ycxnAWMdEqIDbLc3kz5Vt8vm4YP
 examples/weaviate_example/__init__.py,sha256=8JMDBsRSEV10HfTd-YC7xb4txBjD3la56snk-Bbg2Kw,618
 examples/weaviate_example/query_text.py,sha256=wPHQTc_58kPoKTZMygVjTj-2ZcdrIuaausJfMxNQnQc,127162
 langtrace_python_sdk/__init__.py,sha256=VZM6i71NR7pBQK6XvJWRelknuTYUhqwqE7PlicKa5Wg,1166
-langtrace_python_sdk/langtrace.py,sha256=TtRWuUiWUB0S7JiQpUsF9lZsiyqPG3m9mMDX-QlDgAw,12601
-langtrace_python_sdk/version.py,sha256=8z892wHTeofZFdOEAH717nrRAZ1euxjj23yT75bY2Kg,23
+langtrace_python_sdk/langtrace.py,sha256=AN6ecuL47c5eIkgYLW-0nDyEZPaqKfOYRbw7ceZzJss,12598
+langtrace_python_sdk/version.py,sha256=NrcIAr0BNQR9mCN28fqqP9XcjFrymE0bA3_rqg6Oem8,23
 langtrace_python_sdk/constants/__init__.py,sha256=3CNYkWMdd1DrkGqzLUgNZXjdAlM6UFMlf_F-odAToyc,146
 langtrace_python_sdk/constants/exporter/langtrace_exporter.py,sha256=d-3Qn5C_NTy1NkmdavZvy-6vePwTC5curN6QMy2haHc,50
 langtrace_python_sdk/constants/instrumentation/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -152,9 +153,9 @@ langtrace_python_sdk/instrumentation/cohere/instrumentation.py,sha256=YQFHZIBd7S
 langtrace_python_sdk/instrumentation/cohere/patch.py,sha256=AnRWIy00XaLdQg670s8FDXoVWai3sF1JKeR70pDuZ7I,20986
 langtrace_python_sdk/instrumentation/crewai/__init__.py,sha256=_UBKfvQv7l0g2_wnmA5F6CdSAFH0atNOVPd49zsN3aM,88
 langtrace_python_sdk/instrumentation/crewai/instrumentation.py,sha256=5Umzq8zjEnMEtjZZiMB4DQOPkxZ-1vts7RKC6JWpn24,2969
-langtrace_python_sdk/instrumentation/crewai/patch.py,sha256=C2OKKPC-pzfzZWxPc74kHdYsKTX9yRhOgVY47WY9KN8,9109
+langtrace_python_sdk/instrumentation/crewai/patch.py,sha256=VoyOtGKYzaOIu7UnVNTemZeB3LrCIodrrYwmXLdxRw8,9133
 langtrace_python_sdk/instrumentation/dspy/__init__.py,sha256=tM1srfi_QgyCzrde4izojMrRq2Wm7Dj5QUvVQXIJzkk,84
-langtrace_python_sdk/instrumentation/dspy/instrumentation.py,sha256=o8URiDvCbZ8LL0I-4xKHkn_Ms2sETBRpn-gOliv3xzQ,2929
+langtrace_python_sdk/instrumentation/dspy/instrumentation.py,sha256=qx2vBeuODI7rubf-0bkuNzDWu4bLI-E5uabrWTEuH6k,2923
 langtrace_python_sdk/instrumentation/dspy/patch.py,sha256=H7zF4PVdtepOSpzJuEcckKUjnZQYKlY7yhn3dk6xbpY,10458
 langtrace_python_sdk/instrumentation/embedchain/__init__.py,sha256=5L6n8-brMnRWZ0CMmHEuN1mrhIxrYLNtxRy0Ujc-hOY,103
 langtrace_python_sdk/instrumentation/embedchain/instrumentation.py,sha256=dShwm0duy25IvL7g9I_v-2oYuyh2fadeiJqXtXBay-8,1987
@@ -175,8 +176,8 @@ langtrace_python_sdk/instrumentation/langchain_core/__init__.py,sha256=kumE_reeq
 langtrace_python_sdk/instrumentation/langchain_core/instrumentation.py,sha256=szTCveG4IP64rlaY4iZATWv2f38k1_DtfbBU60YlfYk,6730
 langtrace_python_sdk/instrumentation/langchain_core/patch.py,sha256=CXEfbq6E88X_y3JF7CaEEbNCYzSfig5ztNHW-aiiTic,10918
 langtrace_python_sdk/instrumentation/langgraph/__init__.py,sha256=eitlHloY-aZ4ZuIEJx61AadEA3G7siyecP-V-lziAr8,101
-langtrace_python_sdk/instrumentation/langgraph/instrumentation.py,sha256=SUZZhWSIbcfsF1S5NtEqW8QzkRM_pKAuXB7pwk5tsOU,2526
-langtrace_python_sdk/instrumentation/langgraph/patch.py,sha256=PGe1ZywXctB_yYqnp8AtD8Xqj7EZ087-S5_2vLRYhEQ,4987
+langtrace_python_sdk/instrumentation/langgraph/instrumentation.py,sha256=lEm_rcOU4JqXGmhG1C2yrIiPbt9vntvxmU7pZg8NYtE,2313
+langtrace_python_sdk/instrumentation/langgraph/patch.py,sha256=e1cFCDUB8Dwl2ekxgnZ36S2XkWROagRGtxF3Rz5F8RM,4931
 langtrace_python_sdk/instrumentation/litellm/__init__.py,sha256=8uziCc56rFSRiPkYcrcBRbtppOANkZ7uZssCKAl2MKk,97
 langtrace_python_sdk/instrumentation/litellm/instrumentation.py,sha256=Km2q_yfZU6nSqPEXG2xbtTSjqv7xSS92Kxqzw-GtQno,2655
 langtrace_python_sdk/instrumentation/litellm/patch.py,sha256=wGPOlrLo4RHj1lXNv6wOz5H_p4G0XtzhVjgc-2m7Gik,24469
@@ -264,8 +265,8 @@ tests/pinecone/cassettes/test_query.yaml,sha256=b5v9G3ssUy00oG63PlFUR3JErF2Js-5A
 tests/pinecone/cassettes/test_upsert.yaml,sha256=neWmQ1v3d03V8WoLl8FoFeeCYImb8pxlJBWnFd_lITU,38607
 tests/qdrant/conftest.py,sha256=9n0uHxxIjWk9fbYc4bx-uP8lSAgLBVx-cV9UjnsyCHM,381
 tests/qdrant/test_qdrant.py,sha256=pzjAjVY2kmsmGfrI2Gs2xrolfuaNHz7l1fqGQCjp5_o,3353
-langtrace_python_sdk-3.3.11.dist-info/METADATA,sha256=ReGla3YZNI-O-8R3BP4gt1Pu8IyfJOer19fvSuz7XJg,15643
-langtrace_python_sdk-3.3.11.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
-langtrace_python_sdk-3.3.11.dist-info/entry_points.txt,sha256=1_b9-qvf2fE7uQNZcbUei9vLpFZBbbh9LrtGw95ssAo,70
-langtrace_python_sdk-3.3.11.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-langtrace_python_sdk-3.3.11.dist-info/RECORD,,
+langtrace_python_sdk-3.3.13.dist-info/METADATA,sha256=PTvFNsHskew8hssq1CAxynLul99E0Z3jZpJRPw9qXWk,15643
+langtrace_python_sdk-3.3.13.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
+langtrace_python_sdk-3.3.13.dist-info/entry_points.txt,sha256=1_b9-qvf2fE7uQNZcbUei9vLpFZBbbh9LrtGw95ssAo,70
+langtrace_python_sdk-3.3.13.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+langtrace_python_sdk-3.3.13.dist-info/RECORD,,