error-troubleshooting

name	error-troubleshooting
description	Kailash errors — Nexus hangs, connection, runtime, cycles, missing .build(), validation.

Kailash Error Troubleshooting

Common error patterns and solutions for Kailash SDK.

When to Use

Use when encountering errors, debugging issues, or asking about error, troubleshooting, debugging, not working, hangs, timeout, validation error, connection error, runtime error, cycle not converging, missing build, or template syntax. Covers Nexus blocking issues, connection parameter errors, runtime execution errors, cycle convergence problems, missing .build() calls, parameter validation errors, and DataFlow template syntax errors.

Sub-File Index

Critical Errors

• error-nexus-blocking - Nexus hangs or blocks
  • Symptom: API hangs forever | Cause: LocalRuntime in Docker/FastAPI | Fix: Use AsyncLocalRuntime
• error-missing-build - Forgot .build()
  • Symptom: TypeError: execute() expects Workflow, got WorkflowBuilder | Fix: runtime.execute(workflow.build())

Connection & Parameter Errors

• error-connection-exhaustion - Database connection exhaustion
  • Symptom: "too many connections" | Fix: Use external_pool parameter, set max_pool_size = DB max / worker count
• error-connection-params - Invalid connections
  • Symptom: Node gets wrong data | Fix: Use 4-param format (source_id, source_param, target_id, target_param)
• error-parameter-validation - Invalid node parameters
  • Symptom: ValidationError: Missing required parameter | Fix: Check node docs for required params

Runtime & Cycle Errors

• error-runtime-execution - Runtime failures
  • Check logs, validate inputs, test nodes individually, add LoggerNode for visibility
• error-cycle-convergence - Cycles don't converge
  • Symptom: Infinite loop / max iterations exceeded | Fix: Add cycle_complete convergence check

DataFlow Errors

• error-dataflow-template-syntax - Template string errors
  • Symptom: SyntaxError in template strings | Fix: Use {{variable}} syntax

Kaizen Provider Errors

• error-kaizen-provider-config - Provider configuration issues
  • Azure 400 "messages must contain 'json'" | Use json_prompt_suffix() or set response_format
  • "Missing required parameter: response_format.type" | Don't put api_version in response_format
  • DeprecationWarning about provider_config | Migrate to response_format field
  • Azure env var deprecation warnings | Switch to canonical names (AZURE_ENDPOINT, AZURE_API_KEY)

Quick Error Reference

Symptom	Error Type	Quick Fix
API hangs forever	Nexus blocking	Use AsyncLocalRuntime
TypeError: expects Workflow	Missing .build()	Add .build() call
Node gets wrong data	Connection params	Check 4-parameter format
ValidationError	Parameter validation	Check required params
Infinite loop	Cycle convergence	Add convergence condition
Template SyntaxError	DataFlow template	Use {{variable}} syntax
Runtime fails	Runtime execution	Check logs, validate inputs
"too many connections"	Connection exhaustion	Use external_pool injection
Azure 400 "messages must contain 'json'"	Kaizen provider	Use json_prompt_suffix() or response_format
"Missing required parameter: response_format.type"	Kaizen provider	Don't put api_version in response_format
DeprecationWarning about provider_config	Kaizen provider	Migrate to response_format field

Error Prevention Checklist

• Called .build() on WorkflowBuilder?
• Using AsyncLocalRuntime for Docker/FastAPI?
• All connections use 4 parameters?
• All required node parameters provided?
• Cyclic workflows have convergence checks?
• Template strings use {{variable}} syntax?
• Using external_pool in multi-worker deployments?
• Structured output in response_format (not provider_config)?
• Azure using canonical env vars (AZURE_ENDPOINT, AZURE_API_KEY)?
• structured_output_mode="explicit" for new agents?

Static Analysis False Positives

CodeQL: __getattr__ Lazy Loading

Symptom: CodeQL reports "Explicit export is not defined" (py/undefined-export) for names in __all__ resolved by a module-level __getattr__ function.

Fix: Add a TYPE_CHECKING-guarded import for the reported name:

from typing import TYPE_CHECKING

if TYPE_CHECKING:
    from mypackage.submodule import LazyName as LazyName  # explicit re-export

def __getattr__(name: str):
    if name == "LazyName":
        from mypackage.submodule import LazyName
        return LazyName
    raise AttributeError(...)

__all__ = ["LazyName"]  # CodeQL now sees the TYPE_CHECKING import

The as LazyName re-export syntax prevents "unused import" warnings from other linters.

CodeQL: Empty Except Clauses

Fix: Add a comment explaining intent. CodeQL accepts except: pass with an inline comment:

except ValueError:
    pass  # Callback already unregistered; silently ignore duplicate removal

except asyncio.CancelledError:
    pass  # Expected: we just cancelled this task above

CodeQL: Overly Complex __del__

Fix: Use the _warnings=warnings default parameter pattern and minimize branching:

def __del__(self, _warnings=warnings) -> None:
    if not getattr(self, "_closed", True):
        _warnings.warn("Resource was not closed.", ResourceWarning, stacklevel=1)

Avoids conditional import warnings inside __del__ that CodeQL flags as complex. The default parameter captures the module at definition time, surviving interpreter shutdown.

Git Hook Traps

Pre-Commit Auto-Stash Phantom Failure

Symptom: git commit fails with "stash pop" errors or silently drops staged changes, even though running pre-commit run --all-files directly passes every hook.

Root cause: Pre-commit's auto-stash feature stashes unstaged changes before running hooks, then pops after. When a hook modifies the working tree (e.g., a formatter), the pop conflicts with the hook's changes, causing the commit to abort or lose staged hunks.

Workaround: Bypass the hooks directory for this commit only:

git -c core.hooksPath=/dev/null commit -m "fix(scope): description"

Mandatory follow-up: Document the bypass in the commit body AND file a todo against the pre-commit configuration. See rules/git.md "Pre-Commit Hook Workarounds" for the full rule. Silent --no-verify retries are BLOCKED.

Frequency: Recurring across sessions; the stash interaction is non-deterministic and depends on which files have unstaged changes at commit time.

Debugging Tips

1. Always check .build() was called
2. Never ignore connection validation errors
3. Always verify absolute imports when seeing import errors
4. Never assume mock tests found real issues -- use real infrastructure

Related Skills

• 16-validation-patterns - Validation patterns
• 17-gold-standards - Best practices to avoid errors
• 01-core-sdk - Core patterns
• 02-dataflow - DataFlow specifics
• 03-nexus - Nexus specifics

Support

• pattern-expert - Pattern validation
• gold-standards-validator - Check compliance
• testing-specialist - Test debugging