Skip to content

Embeddings

Generate vector embeddings from text using embed() and aembed().

Basic usage

from llmgate import embed

# Single string
resp = embed("text-embedding-3-small", "Hello world")
vector: list[float] = resp.embeddings[0]
print(len(vector))   # 1536

# Batch — list of strings
resp = embed("text-embedding-3-small", ["Hello", "world", "foo"])
vectors: list[list[float]] = resp.embeddings   # one per input

Async

from llmgate import aembed

resp = await aembed("text-embedding-3-small", "Hello world")

Provider examples

resp = embed("text-embedding-3-small", text)
resp = embed("text-embedding-3-large", text, dimensions=256)  # truncate

resp = embed("gemini/text-embedding-004", text)

resp = embed("cohere/embed-english-v3.0", text)
resp = embed("cohere/embed-multilingual-v3.0", text)

resp = embed("mistral/mistral-embed", text)

resp = embed("bedrock/amazon.titan-embed-text-v2:0", text)

resp = embed("ollama/nomic-embed-text", text)

resp = embed("azure/my-embedding-deployment", text)

The EmbeddingResponse

resp.embeddings    # list[list[float]] — one vector per input string
resp.model         # str — model name
resp.provider      # str — "openai" | "gemini" | ...
resp.usage         # TokenUsage
resp.usage.prompt_tokens
resp.usage.total_tokens

Provider support

Provider Supported Notes
OpenAI ✅ dimensions param to truncate
Google Gemini ✅ Uses gemini/ prefix
Azure OpenAI ✅ Uses azure/ prefix
Cohere ✅ Batch supported
Mistral ✅
Bedrock ✅ Titan models
Ollama ✅ Local models
Anthropic ❌ Raises EmbeddingsNotSupported
Groq ❌ Raises EmbeddingsNotSupported