Model Plugins¶
A model plugin is a Python package for a specific embedding model or model vendor. All model plugins contain the same set of commands, and multiple model plugins can be simultaneously installed in the same environment.
embcli-openai for OpenAI Models¶
The embcli-openai plugin provides access to OpenAI's embedding models.
You need an OpenAI API key to use this plugin. Set the OPENAI_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
OPENAI_API_KEY=<YOUR_OPENAI_KEY>
Installation: pip install embcli-openai
emb models command shows the available models with there short aliases and supported model options.
emb models
OpenAIEmbeddingModel
Vendor: openai
Models:
* text-embedding-3-small (aliases: 3-small)
* text-embedding-3-large (aliases: 3-large)
* text-embedding-ada-002 (aliases: ada-002)
Model Options:
* dimensions (int) - The number of dimensions the resulting output embeddings should have. Only supported in text-embedding-3 and later models.
Example usage: get an embedding for an input text by text-embedding-3-small model with an option dimensions=512.
emb embed -m 3-small -o dimensions 512 \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-cohere for Cohere Models¶
The embcli-cohere plugin provides access to Cohere's embedding models.
You need a Cohere API key to use this plugin. Set the COHERE_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
COHERE_API_KEY=<YOUR_COHERE_KEY>
Installation: pip install embcli-cohere
emb models command shows the available models with there short aliases and supported model options.
emb models
CohereEmbeddingModel
Vendor: cohere
Models:
* embed-v4.0 (aliases: embed-v4)
* embed-english-v3.0 (aliases: embed-en-v3)
* embed-english-light-v3.0 (aliases: embed-en-light-v3)
* embed-multilingual-v3.0 (aliases: embed-multiling-v3)
* embed-multilingual-light-v3.0 (aliases: embed-multiling-light-v3)
Model Options:
* input_type (str) - The type of input, affecting how the model processes it. Options include 'search_document', 'search_query', 'classification', 'clustering', 'image'.
* embedding_type (str) - The type of embeddings to return. Options include 'float', 'int8', 'uint8', 'binary', 'ubinary'
* truncate (str) - How to handle text inputs that exceed the model's token limit. Options include 'none', 'start', 'end', 'middle'.
Example usage: get an embedding for an input text by embed-v4.0 model with an option input_type=search_query and embedding_type=binary.
emb embed -m embed-v4 -o input_type search_query -o embedding_type binary \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-gemini for Gemini Models¶
The embcli-gemini plugin provides access to Gemini's embedding models.
You need a Gemini API key to use this plugin. Set the GEMINI_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
GEMINI_API_KEY=<YOUR_GEMINI_KEY>
Installation: pip install embcli-gemini
emb models command shows the available models with there short aliases and supported model options.
GeminiEmbeddingModel
Vendor: gemini
Models:
* gemini-embedding-exp-03-07 (aliases: exp-03-07)
* text-embedding-004 (aliases: text-004)
* embedding-001 (aliases: )
Model Options:
* task_type (str) - The type of task for the embedding. Supported task types: 'semantic_similarity', 'classification', 'clustering', 'retrieval_document', 'retrieval_query', 'question_answering', 'fact_verification', 'code_retrieval_query'
Example usage: get an embedding for an input text by text-embedding-004 model with an option task_type=classification.
emb embed -m text-004 -o task_type classification \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-jina for Jina Models¶
The embcli-jina plugin provides access to Jina's embedding models.
You need a Jina API key to use this plugin. Set the JINA_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
JINA_API_KEY=<YOUR_JINA_KEY>
Installation: pip install embcli-jina
emb models command shows the available models with their short aliases and supported model options.
emb models
JinaEmbeddingModel
Vendor: jina
Models:
* jina-embeddings-v3 (aliases: jina-v3)
* jina-colbert-v2 (aliases: colbert-v2)
* jina-embeddings-v2-base-code (aliases: jina-v2-code)
Model Options:
* task (str) - Downstream task for which the embeddings are used. Supported tasks: 'text-matching', 'retrieval.query', 'retrieval.passage', 'separation', 'classification'. Only supported in jina-embeddings-v3.
* late_chunking (bool) - Whether if the late chunking is applied. Only supported in jina-embeddings-v3.
* truncate (bool) - When enabled, the model will automatically drop the tail that extends beyond the maximum context length allowed by the model instead of throwing an error. Only supported in jina-embeddings-v3.
* dimensions (int) - The number of dimensions the resulting output embeddings should have. Only supported in jina-embeddings-v3 and jina-colbert-v2.
* input_type (str) - The type of input to the model. Supported types: 'query', 'document' Only supported in jina-corebert-v2.
* embedding_type (str) - The type of embeddings to return. Options include 'float', 'binary', 'ubinary'. Default is 'float'.
JinaClipModel
... (snip)
Example usage: get an embedding for an input text by jina-embeddings-v3 model model with an option dimensions=512 and embedding_type=binary.
emb embed -m jina-v3 -o dimensions 512 -o embedding_type binary \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-mistral for Mistral Models¶
The embcli-mistral plugin provides access to Mistral's embedding models.
You need a Mistral API key to use this plugin. Set the MISTRAL_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
MISTRAL_API_KEY=<YOUR_MISTRAL_KEY>
Installation: pip install embcli-mistral
emb models command shows the available models with their short aliases and supported model options.
emb models
MistralEmbeddingModel
Vendor: mistral
Models:
* mistral-embed (aliases: )
* codestral-embed (aliases: )
Model Options:
* output_dimension (int) - The dimesions of the output embedding, defaults to 1536 and has a maximum value of 3072. Only supprted in codestral-embed model.
* output_dtype (str) - The precision and format of the output embedding. Supported values are 'float' (default), 'int8', 'uint8', 'binary', and 'ubinary'. Only supported in codestral-embed model.
Example usage: get an embedding for an input text by mistral-embed model.
emb embed -m mistral-embed \
"Owls can rotate their necks 270 degrees without injury🦉"
Example usage: get an embedding for an input text by codestral-embed model with output_dimension=3072 output_dtype=binary.
emb embed -m codestral-embed -o output_dimension 3072 -o output_dtype binary \
"print('hello, code embeddings')"
embcli-voyage VoyageAI Models¶
The embcli-voyage plugin provides access to VoyageAI's embedding models.
You need a VoyageAI API key to use this plugin. Set the VOYAGE_API_KEY environment variable in a .env file in the current directory, or specify the path to the env file using the -e option.
cat .env
VOYAGE_API_KEY=<YOUR_VOYAGE_KEY>
Installation: pip install embcli-voyage
emb models command shows the available models with their short aliases and supported model options.
emb models
oyageEmbeddingModel
Vendor: voyage
Models:
* voyage-3-large (aliases: )
* voyage-3.5 (aliases: )
* voyage-3.5-lite (aliases: )
* voyage-3 (aliases: )
* voyage-3-lite (aliases: )
* voyage-code-3 (aliases: )
* voyage-finance-2 (aliases: )
* voyage-law-2 (aliases: )
* voyage-code-2 (aliases: )
Model Options:
* input_type (str) - Type of the input text. Options: 'None', 'query', 'document' Defaults to 'None'.
* truncation (bool) - Whether to truncate the input texts to fit within the context length. Defaults to True.
* output_dimension (int) - The number of dimensions for resulting output embeddings.
* output_dtype (str) - The data type for the embeddings to be returned. Options: float, int8, uint8, binary, ubinary. float is supported for all models. int8, uint8, binary, and ubinary are supported by voyage-3-large, voyage-3.5, voyage-3.5-lite, and voyage-code-3.
Example usage: get an embedding for an input text by voyage-3.5 model with options input_type=query and output_dimensions=512 and output_dtype=binary.
emb embed -m voyage-3.5 -o input_type query -o output_dimension 512 -o output_dtype binary \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-sbert for Sentence Transformers Models¶
The embcli-sbert plugin provides access to Sentence Transformers models.
Installation: pip install embcli-sbert
emb models command shows the available models with their short aliases and supported model options.
emb models
SentenceTransformerModel
Vendor: sbert
Models:
* sentence-transformers (aliases: sbert)
Default Local Model: all-MiniLM-L6-v2
See https://sbert.net/docs/sentence_transformer/pretrained_models.html for available models.
Model Options:
Example usage: get an embedding for an input text by an original sentence-transformers model (e.g. all-MiniLM-L6-v2) or a community model.
Unlike proprietary models, you need to add sbert/ to the model prefix.
# Use the default sentence-transformers model, all-MiniLM-L6-v2.
emb embed -m sbert \
"Owls can rotate their necks 270 degrees without injury🦉"
# Use another sentence-transformers model
emb embed -m sbert/all-mpnet-base-v2 \
"Owls can rotate their necks 270 degrees without injury🦉"
# Use a community model
emb embed -m sbert/intfloat/multilingual-e5-small \
"Owls can rotate their necks 270 degrees without injury🦉"
embcli-llamacpp for llama.cpp Models¶
The embcli-llamacpp plugin provides access to locally installed models via llama.cpp and llama-cpp-python.
You need to have gguf model files in your local machine. See this tutorial for instructions on how to convert original Hugging Face transformer models to gguf format.
Installation: pip install embcli-llamacpp
emb models command shows the available models with their short aliases and supported model options.
LlamaCppModel
Vendor: llama-cpp
Models:
* llama-cpp (aliases: llamacpp)
Model Options:
Example usage: get an embedding for an input text by running the GGUF converted model.
# Assume you have a GGUF converted all-MiniLM-L6-v2 model in the current directory.
emb embed -m llamacpp -p ./all-MiniLM-L6-v2.F16.gguf \
"Owls can rotate their necks 270 degrees without injury🦉"