Summary

Fixes #865 (Bug 1)

Problem

When using Ollama as embedding provider with models like nomic-embed-text, get_effective_dimension() would fallback to 2048 (default). This caused:

1. Collection created with Dim: 2048
2. Ollama returns 768-dim vectors
3. Qdrant similarity scores become inf
4. JSON serialization fails: Out of range float values are not JSON compliant: inf

Solution

Added Ollama-specific dimension resolution in get_effective_dimension():

• Known Ollama models return their correct dimension
• Unknown models raise clear error requiring explicit dimension config

ollama_model_dimensions = {
    "nomic-embed-text": 768,
    "mxbai-embed-large": 1024,
    "all-minilm": 384,
    "snowflake-arctic-embed": 1024,
    # ...
}

Testing

# ov.conf
embedding:
  dense:
    provider: ollama
    model: nomic-embed-text
    api_base: http://localhost:11434/v1

• Before: inf similarity scores, JSON error
• After: Collection created with 768-dim, search works correctly