Examples

Real-world dependency conflict scenarios demonstrating depwhy's analysis and fix suggestions.

Scenario 1: requests + urllib3 (Direct Pin Conflict)

A common conflict when pinning both requests and urllib3 to incompatible versions.

Input requirements.txt:

requests==2.28.0
urllib3==2.0.0

Command and output:

$ depwhy requirements.txt
1 conflict(s) found
urllib3
Problem: requests==2.28.0 requires urllib3>=1.21.1,<1.27, but you have urllib3==2.0.0 pinned
Solution: Remove your pin — use urllib3==1.26.18 (latest compatible)
  → pip install urllib3==1.26.18
Alternative: Upgrade requests to >=2.29.0 (loosens urllib3 upper bound)
  → pip install requests==2.29.0

JSON output with --json:

json

{
"has_conflicts": true,
"analyzed_packages": 6,
"conflicts": [
  {
    "package": "urllib3",
    "problem": "requests==2.28.0 requires urllib3>=1.21.1,<1.27, but you have urllib3==2.0.0 pinned",
    "solution": {
      "description": "Remove your pin — use urllib3==1.26.18 (latest compatible)",
      "pip_command": "pip install urllib3==1.26.18",
      "is_alternative": false
    },
    "alternative": {
      "description": "Upgrade requests to >=2.29.0 (loosens urllib3 upper bound)",
      "pip_command": "pip install requests==2.29.0",
      "is_alternative": true
    }
  }
],
"warnings": []
}

Scenario 2: ML Stack (numpy/scipy/scikit-learn)

A typical machine learning stack where an older numpy pin conflicts with scikit-learn.

Input requirements.txt:

numpy==1.21.0
scipy==1.10.0
scikit-learn==1.2.0

$ depwhy requirements.txt
1 conflict(s) found
numpy
Problem: scikit-learn==1.2.0 requires numpy>=1.22.4, but you have numpy==1.21.0 pinned
Solution: Upgrade numpy to satisfy scikit-learn's constraint
  → pip install numpy==1.22.4
Alternative: Downgrade scikit-learn to a version compatible with numpy==1.21.0
  → pip install scikit-learn==1.1.3

Scenario 3: Django + Celery + Kombu (Three-way Conflict)

A transitive conflict that surfaces through the Celery dependency chain.

Input requirements.txt:

django==4.1
celery==5.2.0
amqp==5.1.0

$ depwhy requirements.txt
1 conflict(s) found
amqp
Problem: kombu==5.2.4 requires amqp>=5.0.9,<5.1.0, but you have amqp==5.1.0 pinned
Solution: Downgrade amqp to satisfy kombu's constraint
  → pip install amqp==5.0.9
Alternative: Upgrade celery to >=5.3.0 (uses kombu>=5.3.0 which accepts amqp>=5.1)
  → pip install celery==5.3.0

This is a transitive conflict — celery depends on kombu, which constrains amqp.

Scenario 4: Circular Dependencies

Two packages that depend on each other, with a conflict on a shared transitive dependency.

Input requirements.txt:

alpha==1.0.0
beta==1.0.0

$ depwhy requirements.txt
1 conflict(s) found
shared-lib
Problem: alpha==1.0.0 requires shared-lib>=2.0.0,<3.0.0, but beta==1.0.0 requires shared-lib>=1.0.0,<2.0.0 — no version satisfies both
Solution: Upgrade beta to a version that accepts shared-lib>=2.0
  → pip install beta==1.1.0
Alternative: Downgrade alpha to a version that accepts shared-lib<2.0
  → pip install alpha==0.9.0

depwhy handles circular dependencies gracefully — alpha depends on beta and vice versa, but the conflict is on their shared transitive dependency.

Scenario 5: No Conflicts (Clean Baseline)

When all dependencies are compatible, depwhy confirms a clean state.

Input requirements.txt:

flask==3.0.0
jinja2>=3.1.2

$ depwhy requirements.txt
✓ No conflicts found

Scenario 6: Environment Marker Conflict (Windows-only)

A conflict that only manifests on a specific platform due to environment markers.

Input requirements.txt:

winapp==1.0.0
desktop-notify==1.0.0

$ depwhy requirements.txt --python-version 3.11
1 conflict(s) found
pywin32
Problem: winapp==1.0.0 requires pywin32>=300,<304, but desktop-notify==1.0.0 requires pywin32>=304 — no version satisfies both
Solution: Upgrade winapp to a version with a looser pywin32 constraint
  → pip install winapp==1.1.0
Alternative: Downgrade desktop-notify to accept pywin32<304
  → pip install desktop-notify==0.8.0

This conflict only appears on Windows. On macOS/Linux, the sys_platform == 'win32' markers exclude these dependencies entirely.