Issue in langtrace integration with langgraph supervisor

[ashahid98-24]

Bug Report: IndexError in LangGraph instrumentation when handling variable argument patterns
Environment:
LangTrace Python SDK Version: 3.8.18
LangGraph Version: Latest (used with LangGraph Supervisor)
Python Version: 3.13
Error Location: [langtrace_python_sdk/instrumentation/langgraph/patch.py] in get_atrribute_key_value() function
Problem Description:
The LangTrace SDK's LangGraph instrumentation fails with IndexError: tuple index out of range when [add_node]) is called with different argument patterns than expected. The current implementation assumes [add_node] always receives exactly 2 arguments (name, node_function), but LangGraph's [add_node] method can be called with various patterns.

Root Cause:
In get_atrribute_key_value() function (lines 80-101), the code directly accesses args[0] and args[1] without checking if these indices exist:

def get_atrribute_key_value(method_name, args):
if args is None or len(args) == 0:
return None

if "add_node" in method_name:
    return {
        "langgraph.node": json.dumps({
            "name": args[0],  # ❌ Can fail if len(args) < 1
            "action": (
                args[1].json()  # ❌ Can fail if len(args) < 2
                if hasattr(args[1], "json")
                else (
                    args[1].__name__
                    if hasattr(args[1], "__name__")
                    else str(args[1])
                )
            ),
        }),
        "langgraph.task.name": "add_node",
    }

When the Error Occurs:
Single-argument [add_node] calls: When LangGraph Supervisor calls [add_node(compiled_graph_object) with only one argument
Object-based node addition: When passing CompiledStateGraph objects directly instead of (name, function) pairs
Keyword arguments: When additional parameters like destinations are passed as kwargs (not captured in args)

These cause IndexError:

builder.add_node(supervisor_agent, destinations=tuple(agent_names) + (END,))
builder.add_node(compiled_graph_object)

Debug output showing the issue:

[DEBUG] method_name: langgraph.graph.state.StateGraph.add_node
[DEBUG] args content: (<langgraph.graph.state.CompiledStateGraph object at 0x7f94950b8490>,)
[DEBUG] args length: 1

❌ Crashes when trying to access args[1]

Suggested Fix:
Add bounds checking and handle different argument patterns:

def get_atrribute_key_value(method_name, args):
try:
if args is None or len(args) == 0:
return None

    if "add_node" in method_name:
        name = "unknown"
        action = "unknown"
        
        if len(args) >= 2:
            # Normal case: add_node(name, node_function)
            name = str(args[0])
            if hasattr(args[1], "json"):
                action = args[1].json()
            elif hasattr(args[1], "__name__"):
                action = args[1].__name__
            else:
                action = str(args[1])
        elif len(args) == 1:
            # Single argument case: add_node(node_object)
            node_obj = args[0]
            if hasattr(node_obj, 'name') and node_obj.name:
                name = str(node_obj.name)
            else:
                name = str(type(node_obj).__name__)
            action = str(type(node_obj).__name__)
        
        return {
            "langgraph.node": json.dumps({
                "name": name,
                "action": action,
            }),
            "langgraph.task.name": "add_node",
        }
except Exception as e:
    # Fallback to prevent crashes
    return {
        "langgraph.task.name": "add_node",
        "langgraph.node": json.dumps({"name": "unknown", "action": "unknown"})
    }

Similar Issues:
The same pattern exists in other methods [add_edge], add_conditional_edges, etc.) that also need bounds checking.

[karthikscale3]

Hi, thanks for reporting these issues. We will take a look and get back to you soon.

[olascale3]

🔍 Triage Assessment

Priority: High
Category: Bug - LangGraph Integration
Impact: IndexError in LangGraph instrumentation causing failures

Root Cause Analysis:
The error occurs in get_atrribute_key_value() function in langtrace_python_sdk/instrumentation/langgraph/patch.py when handling variable argument patterns. This suggests the function expects a specific tuple structure but encounters different argument patterns from LangGraph supervisor nodes.

Recommended Fix:

1. Add bounds checking in get_atrribute_key_value() function
2. Handle variable argument patterns more gracefully
3. Add comprehensive testing for different LangGraph node types and argument patterns

Next Steps:

• Review the get_atrribute_key_value() function implementation
• Add defensive programming for tuple index access
• Test with various LangGraph supervisor configurations

This affects users integrating LangTrace with LangGraph workflows, which is a growing use case.

[olascale3]

Triage Notes:

• Type: Bug - Critical
• Priority: High
• Area: LangGraph instrumentation, supervisor pattern
• Labels needed: bug, langgraph, high-priority

Critical IndexError in LangGraph instrumentation affecting the supervisor pattern. Error in get_atrribute_key_value() function when handling variable argument patterns in add_node() calls.

Root Cause: Tuple index out of range suggests the instrumentation code assumes a specific argument structure that doesn't match actual usage.

Action Required: Fix argument handling in LangGraph patch to support variable argument patterns.

[olascale3]

Triage Assessment

Priority: High
Category: Bug - LangGraph Integration
Labels Needed: bug, langgraph, priority:high, indexerror

Analysis: Critical IndexError in LangGraph instrumentation causing failures with supervisor patterns. This is a runtime crash that blocks LangGraph users. The issue is in the get_atrribute_key_value() function when handling variable argument patterns.

Next Steps:

1. URGENT: Fix IndexError in patch.py line handling tuple access
2. Add bounds checking for tuple operations
3. Test with LangGraph Supervisor examples
4. Add regression tests for variable argument patterns