Skip to main content
Pytest failures in PraisonAI fall into predictable categories that help distinguish real regressions from outdated test mocks.

​
Quick Start

1

Identify the Category

Run the one-liner check to identify which category your test failure belongs to.
2

Determine Real vs Stale

Use the verification commands to distinguish between real regressions and outdated mocks.
3

Apply the Fix

Update mock targets or escalate to SDK maintainers based on the analysis.

​
Failure Categories

​
Registry get_instance

Symptom: AttributeError where registry mock returns None instead of registry instance One-liner check:
Likely root cause: Stale mock — registry API changed in SDK, mock needs updating Common failure signature:
Fix approach:
  • Update mock patch target to match current registry location
  • Verify mock returns actual registry instance, not None
  • Check for singleton pattern changes in LLMProviderRegistry.get_instance()

​
execute_code Export

Symptom: ImportError on from praisonaiagents.tools import execute_code One-liner check:
Likely root cause: Export removed/renamed in SDK Common failure signature:
Fix approach:
  • Check TOOL_MAPPINGS in praisonaiagents/tools/__init__.py
  • Verify execute_code is mapped to .python_tools module
  • Update import paths if module structure changed

​
Bot Lazy Loaders

Symptom: First call returns None instead of bot instance due to lazy loading bypass One-liner check:
Likely root cause: Mock targets the eager path; lazy loader bypasses it Common failure signature:
Fix approach:
  • Mock the lazy loading mechanism itself
  • Ensure mocks cover both eager and lazy initialization paths
  • Check for __getattr__ method changes in tools modules

​
aiui Patch Path

Symptom: Mock patch silently no-ops, test passes when it should fail One-liner check:
Likely root cause: Patch target moved; update mock.patch path Common failure signature:
Fix approach:
  • Verify praisonaiui module location has not changed
  • Update patch paths from praisonai.aiui to praisonaiui if needed
  • Check for import alias changes (import praisonaiui as aiui)

​
conftest Key Behavior

Symptom: Tests pass locally but fail in CI, or vice versa One-liner check:
Likely root cause: Real key exported in one environment but not the other Context: Since May 2026 (PR #1663), setup_test_environment only injects placeholder API keys when variables are unset or empty. Real keys from shell environment are preserved. Fix approach:
  • unset API keys before running unit tests to force placeholder injection
  • Use clean shell environment for consistent test behavior
  • Mark tests as real/network/e2e if they need real API keys

​
Troubleshooting Workflow

  • Re-pull latest from main branch
  • Re-install editable package: pip install -e .
  • Check installed version: pip show praisonai-agents
  • Run with verbose output: pytest -vv --tb=long
  • Verify environment isolation (no stale imports)
Real Regression Signs:
  • One-liner checks fail on fresh install
  • Multiple unrelated tests failing simultaneously
  • Recent SDK changes in related modules
  • Failures persist after mock updates
Stale Mock Signs:
  • One-liner checks pass, only mocked tests fail
  • Single test or module affected
  • Mock targets non-existent paths/attributes
  • Failures resolve with mock path updates
  1. Identify current API: Run one-liner to see actual behavior
  2. Update patch target: Match current module location
  3. Verify mock return: Ensure mock returns expected type
  4. Test both paths: Cover eager and lazy loading scenarios
Escalate to SDK maintainers when:
  • One-liner checks confirm real API breakage
  • Breaking changes affect multiple test categories
  • Public API contracts have changed unexpectedly
  • Fixes require SDK-level changes, not just test updates

Testing Guide

Complete testing documentation and environment setup

Development Setup

Local development environment configuration