skip more tests

.github/workflows/badge-spider.yml

@@ -24,6 +24,6 @@ jobs:
           script: |
             const badgeAggregation = require('./scripts/badge_aggregation.js');
             const dependencies = [
-              { workflow: 'examples-spider.yml', label: 'spider', variants: ['stable', 'legacy'] },
+              { workflow: 'examples-spider.yml', label: 'spider', variants: ['stable'] },
             ];
             await badgeAggregation({ github, context, core, dependencies });

.github/workflows/examples-spider.yml

@@ -33,8 +33,7 @@ jobs:
     strategy:
       matrix:
         include:
-        - python-version: '3.10'
-          setup-script: 'legacy'
+        # legacy is omitted because langchain doesn't work with legacy vllm versions
         - python-version: '3.12'
           setup-script: 'stable'
         - python-version: '3.13'
@@ -58,13 +57,13 @@ jobs:
       - name: Sync dependencies (latest)
         run: |
           uv sync --frozen --no-default-groups --extra verl \
-            --group dev --group experiment --group agents --group torch-gpu-stable
+            --group dev --group experiment --group agents --group langchain --group torch-gpu-stable
         if: matrix.setup-script == 'latest'
-      - name: Sync dependencies (stable & legacy)
+      - name: Sync dependencies (stable)
         run: |
           uv sync --frozen --no-default-groups --extra verl \
-            --group dev --group experiment --group agents --group torch-gpu-${{ matrix.setup-script }}
-        if: matrix.setup-script != 'latest'
+            --group dev --group experiment --group agents --group langchain --group torch-gpu-${{ matrix.setup-script }}
+        if: matrix.setup-script == 'stable'
       - name: Freeze dependencies
         run: |
           set -ex

.github/workflows/tests-full.yml

@@ -56,11 +56,15 @@ jobs:
         run: uv lock --upgrade
         if: matrix.setup-script == 'latest'
       - name: Sync dependencies (latest)
-        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group torch-gpu-stable
+        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group langchain --group torch-gpu-stable
         if: matrix.setup-script == 'latest'
-      - name: Sync dependencies (stable & legacy)
-        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group torch-gpu-${{ matrix.setup-script }}
-        if: matrix.setup-script != 'latest'
+      - name: Sync dependencies (stable)
+        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group langchain --group torch-gpu-${{ matrix.setup-script }}
+        if: matrix.setup-script == 'stable'
+      # Don't install langchain for legacy dependency because it has conflicts with torch.
+      - name: Sync dependencies (legacy)
+        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group torch-gpu-legacy
+        if: matrix.setup-script == 'legacy'
       - name: Freeze dependencies
         run: |
           set -ex
@@ -178,11 +182,15 @@ jobs:
         run: uv lock --upgrade
         if: matrix.setup-script == 'latest'
       - name: Sync dependencies (latest)
-        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group torch-gpu-stable
+        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group langchain --group torch-gpu-stable
         if: matrix.setup-script == 'latest'
-      - name: Sync dependencies (stable & legacy)
-        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group torch-gpu-${{ matrix.setup-script }}
-        if: matrix.setup-script != 'latest'
+      - name: Sync dependencies (stable)
+        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group langchain --group torch-gpu-${{ matrix.setup-script }}
+        if: matrix.setup-script == 'stable'
+      # Don't install langchain for legacy dependency because it has conflicts with torch.
+      - name: Sync dependencies (legacy)
+        run: uv sync --frozen --no-default-groups --extra apo --extra mongo --group dev --group agents --group torch-gpu-legacy
+        if: matrix.setup-script == 'legacy'
       - name: Freeze dependencies
         run: |
           set -ex

.github/workflows/tests.yml

@@ -44,6 +44,7 @@ jobs:
             --group trl \
             --group tinker \
             --group agents \
+            --group langchain \
             --no-default-groups
         if: matrix.setup == 'slow'
       # This pre-commit skips JavaScript on purpose.
@@ -139,10 +140,10 @@ jobs:
         run: uv lock --upgrade
         if: matrix.setup-script == 'latest'
       - name: Sync dependencies (latest)
-        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group core-stable
+        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group langchain --group core-stable
         if: matrix.setup-script == 'latest'
       - name: Sync dependencies (stable & legacy)
-        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group core-${{ matrix.setup-script }}
+        run: uv sync --frozen --no-default-groups --extra apo --group dev --group agents --group langchain --group core-${{ matrix.setup-script }}
         if: matrix.setup-script != 'latest'
       - name: Freeze dependencies
         run: |

pyproject.toml

@@ -183,10 +183,10 @@ anthropic = [
   "anthropic",
 ]
 langchain = [
-  "langgraph<1.0",
-  "langchain[openai]<1.0",
-  "langchain-community",
-  "langchain-text-splitters<1.0",
+  "langgraph>=1.0.0",
+  "langchain[openai]>=1.0.0",
+  "langchain-community>=0.4.0",
+  "langchain-text-splitters>=1.0.0",
 ]
 sql = [
   "sqlparse",
@@ -206,10 +206,11 @@ swebench = [
 ]
 
 # Summarize into large installable groups.
+# Note: langchain is excluded from agents to avoid conflicts with torch-legacy.
+# Use --group langchain explicitly when needed.
 agents = [
   {include-group = "autogen"},
   {include-group = "openai-agents"},
-  {include-group = "langchain"},
   {include-group = "sql"},
   {include-group = "anthropic"},
   {include-group = "crewai"},
@@ -231,6 +232,12 @@ conflicts = [
     { group = "torch-stable" },
     { group = "torch-legacy" },
   ],
+  # langchain >= 1.0 requires openai >= 1.109.1 (via langchain-openai),
+  # but torch-legacy uses vllm==0.9.2 which requires openai<=1.90.0
+  [
+    { group = "langchain" },
+    { group = "torch-legacy" },
+  ],
 ]
 environments = [
   "sys_platform == 'linux'",
@@ -250,6 +257,8 @@ override-dependencies = [
   # Conflicts between litellm and fastmcp
   "websockets>=15.0.1",
   "rich>=13.9.4",
+  # verl's numpy<2.0.0 constraint is related to Docker images, not code incompatibility
+  "numpy>=2.0.0,<2.3.0",
 ]
 
 [tool.uv.sources]

tests/assets/prompt_caches.jsonl

@@ -28,3 +28,5 @@
 {"request": {"messages": [{"content": "Return 1.0 if the answer is 8, else 0.0.", "role": "system"}, {"role": "user", "content": "8"}], "model": "gpt-4.1-mini", "response_format": {"type": "json_schema", "json_schema": {"name": "final_output", "strict": true, "schema": {"properties": {"response": {"title": "Response", "type": "number"}}, "required": ["response"], "title": "OutputType", "type": "object", "additionalProperties": false}}}}, "response": {"choices": [{"content_filter_results": {"hate": {"filtered": false, "severity": "safe"}, "protected_material_code": {"filtered": false, "detected": false}, "protected_material_text": {"filtered": false, "detected": false}, "self_harm": {"filtered": false, "severity": "safe"}, "sexual": {"filtered": false, "severity": "safe"}, "violence": {"filtered": false, "severity": "safe"}}, "finish_reason": "stop", "index": 0, "logprobs": null, "message": {"annotations": [], "content": "{\"response\":1.0}", "refusal": null, "role": "assistant"}}], "created": 1763565701, "id": "chatcmpl-CdeI1ifu2oOt06oTodWPpSpNFAqIN", "model": "gpt-4.1-mini-2025-04-14", "object": "chat.completion", "prompt_filter_results": [{"prompt_index": 0, "content_filter_results": {"hate": {"filtered": false, "severity": "safe"}, "jailbreak": {"filtered": false, "detected": false}, "self_harm": {"filtered": false, "severity": "safe"}, "sexual": {"filtered": false, "severity": "safe"}, "violence": {"filtered": false, "severity": "safe"}}}], "system_fingerprint": "fp_3dcd5944f5", "usage": {"completion_tokens": 8, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0}, "prompt_tokens": 65, "prompt_tokens_details": {"audio_tokens": 0, "cached_tokens": 0}, "total_tokens": 73}}}
 {"request": {"messages": [{"content": "If the question is about math, handoff to MathAgent. Otherwise, handoff to HistoryAgent.", "role": "system"}, {"role": "user", "content": "Who was the first president of the US?"}], "model": "gpt-4.1-mini", "tools": [{"type": "function", "function": {"name": "transfer_to_mathagent", "description": "Handoff to the MathAgent agent to handle the request. ", "parameters": {"additionalProperties": false, "type": "object", "properties": {}, "required": []}}}, {"type": "function", "function": {"name": "transfer_to_historyagent", "description": "Handoff to the HistoryAgent agent to handle the request. ", "parameters": {"additionalProperties": false, "type": "object", "properties": {}, "required": []}}}]}, "response": {"choices": [{"content_filter_results": {}, "finish_reason": "tool_calls", "index": 0, "logprobs": null, "message": {"annotations": [], "content": null, "refusal": null, "role": "assistant", "tool_calls": [{"function": {"arguments": "{}", "name": "transfer_to_historyagent"}, "id": "call_tXXmrsNYcFlKrG9hMo2Yk0E3", "type": "function"}]}}], "created": 1763565706, "id": "chatcmpl-CdeI6hxlaAwzbn7Sx4ZTXZs0Hx3E5", "model": "gpt-4.1-mini-2025-04-14", "object": "chat.completion", "prompt_filter_results": [{"prompt_index": 0, "content_filter_results": {"hate": {"filtered": false, "severity": "safe"}, "jailbreak": {"filtered": false, "detected": false}, "self_harm": {"filtered": false, "severity": "safe"}, "sexual": {"filtered": false, "severity": "safe"}, "violence": {"filtered": false, "severity": "safe"}}}], "system_fingerprint": "fp_3dcd5944f5", "usage": {"completion_tokens": 13, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0}, "prompt_tokens": 97, "prompt_tokens_details": {"audio_tokens": 0, "cached_tokens": 0}, "total_tokens": 110}}}
 {"request": {"messages": [{"content": "Answer history questions.", "role": "system"}, {"role": "assistant", "content": "For context, here is the conversation so far between the user and the previous agent:\n<CONVERSATION HISTORY>\n1. user: Who was the first president of the US?\n2. function_call: {\"arguments\": \"{}\", \"call_id\": \"call_tXXmrsNYcFlKrG9hMo2Yk0E3\", \"name\": \"transfer_to_historyagent\", \"id\": \"__fake_id__\"}\n3. function_call_output: {\"call_id\": \"call_tXXmrsNYcFlKrG9hMo2Yk0E3\", \"output\": \"{\\\"assistant\\\": \\\"HistoryAgent\\\"}\"}\n</CONVERSATION HISTORY>"}, {"role": "assistant", "tool_calls": [{"id": "call_tXXmrsNYcFlKrG9hMo2Yk0E3", "type": "function", "function": {"name": "transfer_to_historyagent", "arguments": "{}"}}]}, {"role": "tool", "tool_call_id": "call_tXXmrsNYcFlKrG9hMo2Yk0E3", "content": "{\"assistant\": \"HistoryAgent\"}"}], "model": "gpt-4.1-mini"}, "response": {"choices": [{"content_filter_results": {"hate": {"filtered": false, "severity": "safe"}, "protected_material_text": {"filtered": false, "detected": false}, "self_harm": {"filtered": false, "severity": "safe"}, "sexual": {"filtered": false, "severity": "safe"}, "violence": {"filtered": false, "severity": "safe"}}, "finish_reason": "stop", "index": 0, "logprobs": null, "message": {"annotations": [], "content": "The first president of the United States was George Washington. He served as president from 1789 to 1797. If you have more questions about U.S. history or any other historical events, feel free to ask!", "refusal": null, "role": "assistant"}}], "created": 1763565707, "id": "chatcmpl-CdeI78NuMQmjmeuwLhx3I0Qqcem99", "model": "gpt-4.1-mini-2025-04-14", "object": "chat.completion", "prompt_filter_results": [{"prompt_index": 0, "content_filter_results": {}}], "system_fingerprint": "fp_3dcd5944f5", "usage": {"completion_tokens": 46, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0}, "prompt_tokens": 179, "prompt_tokens_details": {"audio_tokens": 0, "cached_tokens": 0}, "total_tokens": 225}}}
+{"request": {"messages": [{"role": "system", "content": "You are a helpful assistant. Use the multiply tool to answer math questions."}, {"role": "user", "content": "what is 42 * 12"}], "model": "gpt-4.1-mini", "tools": [{"type": "function", "function": {"name": "multiply", "description": "A simple calculator tool that multiplies two integers.", "parameters": {"type": "object", "properties": {"a_and_b": {"type": "string"}}, "required": ["a_and_b"]}}}]}, "response": {"id": "chatcmpl-test1", "model": "gpt-4.1-mini", "object": "chat.completion", "created": 1750758616, "choices": [{"index": 0, "message": {"role": "assistant", "content": null, "tool_calls": [{"id": "call_abc123", "type": "function", "function": {"name": "multiply", "arguments": "{\"a_and_b\": \"42 and 12\"}"}}]}, "finish_reason": "tool_calls"}], "usage": {"prompt_tokens": 100, "completion_tokens": 20, "total_tokens": 120}}}
+{"request": {"messages": [{"role": "system", "content": "You are a helpful assistant. Use the multiply tool to answer math questions."}, {"role": "user", "content": "what is 42 * 12"}, {"role": "assistant", "content": null, "tool_calls": [{"id": "call_abc123", "type": "function", "function": {"name": "multiply", "arguments": "{\"a_and_b\": \"42 and 12\"}"}}]}, {"role": "tool", "content": "504", "tool_call_id": "call_abc123"}], "model": "gpt-4.1-mini", "tools": [{"type": "function", "function": {"name": "multiply", "description": "A simple calculator tool that multiplies two integers.", "parameters": {"type": "object", "properties": {"a_and_b": {"type": "string"}}, "required": ["a_and_b"]}}}]}, "response": {"id": "chatcmpl-test2", "model": "gpt-4.1-mini", "object": "chat.completion", "created": 1750758617, "choices": [{"index": 0, "message": {"role": "assistant", "content": "The result of 42 * 12 is 504."}, "finish_reason": "stop"}], "usage": {"prompt_tokens": 150, "completion_tokens": 15, "total_tokens": 165}}}

tests/tracer/test_integration.py

@@ -30,7 +30,7 @@ import socket
 import threading
 import time
 from contextlib import asynccontextmanager, contextmanager
-from typing import Any, Callable, Dict, Iterator, List, Literal, Optional, Tuple
+from typing import TYPE_CHECKING, Any, Callable, Dict, Iterator, List, Literal, Optional, Tuple
 from unittest.mock import AsyncMock, MagicMock, Mock, patch
 
 import agentops
@@ -50,17 +50,6 @@ from autogen_agentchat.teams import RoundRobinGroupChat
 from autogen_ext.models.openai import OpenAIChatCompletionClient
 from autogen_ext.tools.mcp import McpWorkbench, StdioServerParams
 from fastapi import FastAPI
-from langchain import hub
-from langchain.agents import AgentExecutor, create_react_agent, tool
-from langchain.chat_models import init_chat_model
-from langchain_community.agent_toolkits import SQLDatabaseToolkit
-from langchain_community.utilities import SQLDatabase
-from langchain_core.messages import AIMessage, BaseMessage, HumanMessage, ToolMessage
-from langchain_core.output_parsers import StrOutputParser
-from langchain_core.prompts import ChatPromptTemplate
-from langchain_openai import ChatOpenAI
-from langgraph.graph import END, START, MessagesState, StateGraph
-from langgraph.prebuilt import ToolNode
 from openai import AsyncOpenAI, OpenAI
 from opentelemetry.sdk.trace import ReadableSpan
 from pydantic import BaseModel, Field
@@ -74,6 +63,25 @@ from agentlightning.types import Span, Triplet
 
 from ..common.tracer import clear_agentops_init, clear_tracer_provider
 
+try:
+    import langchain
+
+    LANGCHAIN_INSTALLED = True
+except ImportError:
+    LANGCHAIN_INSTALLED = False
+
+if TYPE_CHECKING or LANGCHAIN_INSTALLED:
+    from langchain.agents import create_agent
+    from langchain.chat_models import init_chat_model
+    from langchain_community.agent_toolkits import SQLDatabaseToolkit
+    from langchain_community.utilities import SQLDatabase
+    from langchain_core.messages import AIMessage, BaseMessage, HumanMessage, ToolMessage
+    from langchain_core.output_parsers import StrOutputParser
+    from langchain_core.prompts import ChatPromptTemplate
+    from langchain_core.tools import tool
+    from langchain_openai import ChatOpenAI
+    from langgraph.graph import END, START, MessagesState, StateGraph
+
 USE_OPENAI = os.environ.get("USE_OPENAI", "false").lower() == "true"
 OPENAI_BASE_URL = "http://127.0.0.1:58000/v1"
 OPENAI_MODEL = "gpt-4.1-mini"
@@ -301,13 +309,16 @@ def agent_langchain_tooluse() -> None:
         disable_streaming=True,
     )
     tools = [multiply]
-    agent = create_react_agent(llm, tools, hub.pull("hwchase17/react"))
-    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=False)
-    result = agent_executor.invoke(
-        {"input": "what is 42 * 12"},
+    agent = create_agent(
+        model=llm,
+        tools=tools,
+        system_prompt="You are a helpful assistant. Use the multiply tool to answer math questions.",
+    )
+    result = agent.invoke(
+        {"messages": [{"role": "user", "content": "what is 42 * 12"}]},
         {"callbacks": [_langchain_callback_handler]} if _langchain_callback_handler else None,
     )
-    assert "504" in result["output"]
+    assert "504" in result["messages"][-1].content
 
 
 def agent_langgraph() -> None:
@@ -321,10 +332,25 @@ def agent_langgraph() -> None:
         return next(t for t in tools if t.name == name)
 
     get_schema_tool = next(tool for tool in tools if tool.name == "sql_db_schema")
-    get_schema_node = ToolNode([get_schema_tool], name="get_schema")
-
     run_query_tool = next(tool for tool in tools if tool.name == "sql_db_query")
-    run_query_node = ToolNode([run_query_tool], name="run_query")
+
+    def get_schema(state: MessagesState):
+        """Execute the get_schema tool based on the last message's tool calls."""
+        last_message = state["messages"][-1]
+        tool_messages = []
+        for tool_call in getattr(last_message, "tool_calls", []):
+            result = get_schema_tool.invoke(tool_call)
+            tool_messages.append(result)
+        return {"messages": tool_messages}
+
+    def run_query(state: MessagesState):
+        """Execute the run_query tool based on the last message's tool calls."""
+        last_message = state["messages"][-1]
+        tool_messages = []
+        for tool_call in getattr(last_message, "tool_calls", []):
+            result = run_query_tool.invoke(tool_call)
+            tool_messages.append(result)
+        return {"messages": tool_messages}
 
     def list_tables(state: MessagesState):
         tool_call = {
@@ -381,10 +407,10 @@ def agent_langgraph() -> None:
     builder = StateGraph(MessagesState)
     builder.add_node(list_tables)
     builder.add_node(call_get_schema)
-    builder.add_node(get_schema_node, "get_schema")
+    builder.add_node(get_schema)
     builder.add_node(generate_query)
     builder.add_node(check_query)
-    builder.add_node(run_query_node, "run_query")
+    builder.add_node(run_query)
     builder.add_edge(START, "list_tables")
     builder.add_edge("list_tables", "call_get_schema")
     builder.add_edge("call_get_schema", "get_schema")
@@ -797,6 +823,8 @@ def test_run_with_agentops_tracer(agent_func_name: str):
     """AgentOps tracer tests are notoriously problematic and does not work well with other tests."""
     if agent_func_name in ["openai_agents_sdk_mcp_tool_use", "agent_autogen_mcp"]:
         pytest.skip("Async MCP server is problematic with AgentOps tracer in multiprocessing mode.")
+    if ("langchain" in agent_func_name or "langgraph" in agent_func_name) and not LANGCHAIN_INSTALLED:
+        pytest.skip("LangChain is not installed. Skip langchain related tests.")
 
     ctx = multiprocessing.get_context("spawn")
     proc = ctx.Process(target=_test_run_with_agentops_tracer_impl, args=(agent_func_name,))
@@ -824,7 +852,9 @@ def _test_run_with_agentops_tracer_impl(agent_func_name: str):
     tracer.init_worker(0)
 
     global _langchain_callback_handler
-    _langchain_callback_handler = tracer.get_langchain_callback_handler()
+
+    if LANGCHAIN_INSTALLED:
+        _langchain_callback_handler = tracer.get_langchain_callback_handler()
 
     try:
         tracer.trace_run(