Resolving "Solving environment: failed with initial frozen solve" in Conda

Problem Statement

When installing packages with Conda, you may encounter the following error:

Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.

This occurs because Conda's default dependency resolver struggles with complex package relationships. Common triggers include:

• Conflicting package versions or dependencies
• Package incompatibilities in your environment
• Challenging dependencies from conda-forge or other channels
• Large environments with many interdependent packages

The error indicates Conda is attempting multiple resolution strategies but failing to find a consistent set of packages that satisfy all requirements.

Recommended Solutions

1. Use Conda's libmamba Solver (Official Recommendation)

Conda's developers recommend the faster, more efficient libmamba solver:

# Install the solver plugin
conda install -n base conda-libmamba-solver

# Use for a single installation
conda install tensorflow --solver=libmamba

# OR set as default solver
conda config --set solver libmamba

Benefits

• Solves complex environments in seconds instead of minutes
• Maintains full compatibility with existing Conda packages
• Officially supported (see Conda documentation)

2. Replace Conda with Mamba

Mamba is a drop-in alternative to Conda with a faster dependency resolver:

# Install mamba in base environment
conda install -n base mamba -c conda-forge

# Use mamba for installations
mamba install python-pdfkit -c conda-forge

Compatibility Note

• Installs packages from standard Conda channels
• Requires installing mamba in your current environment first
• May resolve conflicts Conda's older solvers can't handle

3. Adjust Channel Priority Settings

If you're using strict channel priority, relax the setting:

conda config --set channel_priority flexible

After changing, retry your installation:

conda install python-pdfkit -c conda-forge

4. Environment Isolation

For persistent issues, create a clean environment:

# Create environment with specific channels
conda create -n pdf_env python=3.10 -c conda-forge

# Activate environment
conda activate pdf_env

# Install package
conda install python-pdfkit

Channel Consistency

• Avoid mixing packages from defaults and conda-forge channels
• Explicitly specify all channels during environment creation

Why These Solutions Work

1. Modern Solvers (libmamba/mamba):

   • Use advanced SAT algorithms for dependency resolution
   • Handle version conflicts more efficiently
   • Reduce solve times from minutes to seconds

2. Channel Priority Adjustment:

   • Allows using packages from lower-priority channels if needed
   • Helps when strict channel priorities cause artificial conflicts

3. Environment Isolation:

   • Eliminates conflicts from existing packages
   • Provides a clean dependency tree for new installations

Best Practices

1. Update Conda regularly:

   conda update -n base conda

2. Prefer minimal environments over installing everything in base

3. Always specify channels explicitly during installation

Avoid This Anti-Pattern

Using pip exclusively in your base environment often compounds dependency issues. Only use pip when packages are unavailable through Conda.

4. Prioritize solutions in this order:
   • Use libmamba solver (official solution)
   • Try Mamba (community favorite)
   • Adjust channel settings
   • Create clean environments

These solutions address the core dependency resolution challenges causing "failed with initial frozen solve" errors while maintaining environment stability.