Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.oasis.camel-ai.org/llms.txt

Use this file to discover all available pages before exploring further.

​
Models

OASIS supports all models listed in camel here: https://docs.camel-ai.org/key_modules/models. Note that only models with tool-calling support can successfully perform actions in OASIS. You can pass a ModelBackend, List[ModelBackend] or ModelManager as needed. For example, the OPENAI model: 
from camel.models import ModelFactory
from camel.types import ModelPlatformType, ModelType
from camel.configs import ChatGPTConfig

# Define the model, here in this case we use gpt-4o-mini
model = ModelFactory.create(
    model_platform=ModelPlatformType.OPENAI,
    model_type=ModelType.GPT_4O_MINI,
    model_config_dict=ChatGPTConfig().as_dict(),
)
For the vLLM model, note that to deploy the model with tool-calling capabilities, you should refer to the documentation here: https://docs.vllm.ai/en/latest/features/tool_calling.html. 
from camel.models import ModelFactory
from camel.types import ModelPlatformType

vllm_model = ModelFactory.create(
    model_platform=ModelPlatformType.VLLM,
    model_type="microsoft/Phi-3-mini-4k-instruct",
    url="http://localhost:8000/v1", # Optional
    model_config_dict={"temperature": 0.0}, # Optional
)
For the ModelManager, you can define the scheduling strategy as needed. round_robin is the recommended strategy for load balancing on multiple models among agents. 
from camel.models import ModelFactory, ModelManager
from camel.types import ModelPlatformType

vllm_model_1 = ModelFactory.create(
    model_platform=ModelPlatformType.VLLM,
    model_type="qwen-2",
    url="http://localhost:8000/v1", # Optional
)

vllm_model_2 = ModelFactory.create(
    model_platform=ModelPlatformType.VLLM,
    model_type="qwen-2",
    url="http://localhost:8001/v1", # Optional
)

model_manager = ModelManager(
    models=[vllm_model_1, vllm_model_2],
    scheduling_strategy='round_robin',
)