Skip to main content

Troubleshooting

Common issues and solutions.

Token count not decreasing​

Problem: Tokens increase instead of decrease Cause: Using v1.5.0 (had accumulation bug) Solution: Upgrade to v1.5.1+

pip install --upgrade vidurai
python -c "import vidurai; print(vidurai.__version__)"  # Should show 1.5.1+

High-threshold recall returns few results​

Problem: recall(min_importance=0.7) returns 1/5 items Cause: Importance decay drops memories below threshold Solution: Disable decay or use lower threshold

# Option 1: Disable decay
memory = Vidurai(enable_decay=False)

# Option 2: Use lower threshold
results = memory.recall(min_importance=0.5)

# Option 3: Use count-based recall
results = memory.recall(limit=10)  # Get top 10 regardless of threshold

RL agent not learning​

Problem: Q-table not growing, epsilon not decaying Cause: Not enough episodes Solution: Give it 50-100 episodes

stats = memory.get_rl_agent_stats()
if stats['episodes'] < 50:
    print("Agent still learning, needs more episodes")
    print(f"Current episode: {stats['episodes']}")
    print(f"Exploration rate: {stats['epsilon']:.2%}")

Memory not being compressed​

Problem: Token count keeps growing, no compression triggered Cause: Threshold not reached or compression disabled Solution: Check configuration

# Check current stats
stats = memory.get_memory_stats()
print(f"Tokens: {stats['tokens']}")
print(f"Threshold: {memory.compression_threshold}")

# Manual trigger
memory.trigger_compression()

# Or lower threshold
memory.compression_threshold = 0.7

Import errors​

Problem: ModuleNotFoundError: No module named 'vidurai' Cause: Not installed or wrong environment Solution: Install in correct environment

# Check current environment
which python

# Install
pip install vidurai

# Verify
python -c "import vidurai; print(vidurai.__version__)"

Performance issues​

Problem: Slow recall or compression Cause: Large memory store, inefficient queries Solution: Optimize queries and storage

# Use semantic search instead of scanning all memories
results = memory.recall(query="specific topic", limit=5)

# Periodic cleanup
memory.clear_below_threshold(0.2)

# Enable indexing (if available)
memory = Vidurai(enable_indexing=True)

Compression quality issues​

Problem: Compressed memories lose critical information Cause: Too aggressive compression settings Solution: Use quality-focused profile

from vidurai.core.data_structures_v2 import RewardProfile

memory = Vidurai(
    reward_profile=RewardProfile.QUALITY_FOCUSED,
    compression_threshold=0.85
)

API key errors​

Problem: OpenAI API key not found Cause: Environment variable not set Solution: Set API key

# Option 1: Environment variable
export OPENAI_API_KEY="sk-..."

# Option 2: In code
import os
os.environ["OPENAI_API_KEY"] = "sk-..."

# Option 3: Pass to constructor
memory = Vidurai(api_key="sk-...")

Getting Help​

Still stuck? Get help from:

When reporting issues, include:

  1. Vidurai version (vidurai.__version__)
  2. Python version
  3. Minimal reproducible code
  4. Error messages and stack traces