embedding

mdvdb embedding

Inspect a provider's live model catalog or make one minimal inference to verify the configured provider, model, credentials, and vector dimensions.

Usage

mdvdb embedding <COMMAND> [OPTIONS]

Command	Description
models [--provider <PROVIDER>]	List models reported by a provider's live catalog
probe	Embed one short probe input and report the resolved dimensions and latency

The commands also accept all global options, including --json and --root.

List models

# Use the configured provider
mdvdb embedding models

# Temporarily select a provider for discovery
mdvdb embedding models --provider gemini

# Consume the catalog as JSON
mdvdb embedding models --provider huggingface --json

Canonical provider names are openai, openrouter, gemini, azure, bedrock, huggingface, ollama, and custom. Accepted aliases include google, azure-openai, aws-bedrock, and hf.

Catalog support varies by provider. OpenAI, Azure OpenAI, Ollama, and Custom currently return discovery_available: false; enter their model IDs directly. Other providers may still return no catalog if the remote service does not expose one.

{
  "provider": "gemini",
  "discovery_available": true,
  "models": [
    {
      "id": "models/gemini-embedding-001",
      "name": "Gemini Embedding 001",
      "input_token_limit": 2048
    }
  ]
}

Each model has an opaque id; name and input_token_limit may be null because provider catalogs expose different metadata.

Probe the configured model

mdvdb embedding probe
mdvdb embedding probe --json

Human-readable output has the form:

openrouter · openai/text-embedding-3-small · 1536 dimensions · 184 ms

JSON output:

{
  "provider": "openrouter",
  "model": "openai/text-embedding-3-small",
  "dimensions": 1536,
  "latency_ms": 184
}

probe uses the configured provider; it has no provider override. It performs a real one-input embedding request, so it requires valid credentials/connectivity and may count toward provider usage. The returned vector itself is never printed.