Model Providers

Agentkit supports a wide range of LLM providers through elsai-model v2.1.0 — either the LLM factory or direct *Model classes from elsai Core.

For standalone invoke / stream usage (without agents), see LLM Models in elsai Core. Prefer the LLM factory when switching providers.

Install

pip install --extra-index-url https://elsai-agents.elsai.ai/root/ elsai-agents==0.3.1
pip install --extra-index-url https://core-packages.elsai.ai/root/ "elsai-model[openai]==2.1.0"

Add more provider extras as needed (for example "elsai-model[openai,bedrock]==2.1.0") — see LLM Models — Optional package extras.

Supported providers

Provider	Model class	Docs
Amazon Bedrock	BedrockModel	Bedrock
Anthropic (direct)	AnthropicModel	Anthropic
Anthropic via Bedrock SDK	AnthropicBedrockModel	Anthropic
OpenAI	OpenAIModel	OpenAI
Google Gemini	GeminiModel	Gemini
LiteLLM	LiteLLMModel	LiteLLM
Ollama (local)	OllamaModel	Ollama
Mistral	MistralModel	Mistral
Writer	WriterModel	Writer
Meta Llama API	LlamaAPIModel	Llama API
llama.cpp (local)	LlamaCppModel	llama.cpp
Amazon SageMaker	SageMakerAIModel	SageMaker

Quick switch

Using the factory (recommended). Set API keys in the environment (do not hard-code secrets):

import os
from elsai import Agent
from elsai_model import LLM, Provider

# OpenAI — requires OPENAI_API_KEY
agent = Agent(
    model=LLM(
        provider=Provider.OPENAI,
        model="gpt-4o",
        api_key=os.environ["OPENAI_API_KEY"],
    )
)

# Gemini — requires GEMINI_API_KEY
agent = Agent(
    model=LLM(
        provider=Provider.GEMINI,
        model="gemini-2.5-flash",
        api_key=os.environ["GEMINI_API_KEY"],
    )
)

# Amazon Bedrock — requires AWS credentials; region defaults to us-west-2
agent = Agent(
    model=LLM(
        provider=Provider.BEDROCK,
        model="global.anthropic.claude-sonnet-4-6",
        region_name=os.getenv("AWS_DEFAULT_REGION", "us-west-2"),
    )
)

# Ollama (local) — no API key
agent = Agent(
    model=LLM(
        provider=Provider.OLLAMA,
        model="llama3.2",
        host="http://localhost:11434",
    )
)

Direct *Model construction still works:

import os
from elsai import Agent
from elsai_model.openai import OpenAIModel
from elsai_model.gemini import GeminiModel

agent = Agent(
    model=OpenAIModel(
        model_id="gpt-4o",
        client_args={"api_key": os.environ["OPENAI_API_KEY"]},
    )
)
agent = Agent(
    model=GeminiModel(
        model_id="gemini-2.5-flash",
        client_args={"api_key": os.environ["GEMINI_API_KEY"]},
    )
)

Common parameters

Parameter	Description
provider	Provider enum or string when using LLM(provider=…, model=…)
model_id	Provider model or deployment ID (on *Model classes)
model	Same role when using LLM(provider=…, model=…)
api_key	Factory shorthand merged into client_args
client_args	SDK client options (API keys, endpoints)
params	Generation settings (temperature, max_tokens, …)

Related

• LLM Models — LLM factory
• Amazon Bedrock
• Anthropic
• OpenAI
• Gemini
• Ollama
• LiteLLM
• Mistral
• Writer
• Llama API
• llama.cpp
• SageMaker
• LLM Models (invoke/stream)