steamship.agents.llms package

Submodules

steamship.agents.llms.openai module

class steamship.agents.llms.openai.ChatOpenAI(client, model_name: str = 'gpt-4-0613', *args, generator: PluginInstance)

Bases: ChatLLM, OpenAI

ChatLLM that uses Steamship’s OpenAI plugin to generate chat completions.

chat(messages: List[Block], tools: List[Tool] | None, **kwargs) → List[Block]

Sends chat messages to the LLM with functions from the supplied tools in a side-channel.

Supported kwargs include: - max_tokens (controls the size of LLM responses)

class steamship.agents.llms.openai.OpenAI(client, model_name: str = 'gpt-3.5-turbo', temperature: float = 0.4, *args, generator: PluginInstance)

Bases: LLM

LLM that uses Steamship’s OpenAI plugin to generate completions.

NOTE: By default, this LLM uses the gpt-3.5-turbo model. Valid model choices are gpt-3.5-turbo and gpt-4.

client: Steamship

complete(prompt: str, stop: str | None = None, **kwargs) → List[Block]

Completes the prompt, respecting the supplied stop sequence.

Supported kwargs include: - max_tokens (controls the size of LLM responses)

generator: PluginInstance

steamship.agents.llms.steamship_llm module

class steamship.agents.llms.steamship_llm.SteamshipLLM(plugin_instance: PluginInstance)

Bases: object

A class wrapping LLM plugins.

class Impls

Bases: object

static gpt(client: Steamship, plugin_version: str | None = None, model: str = 'gpt-4', temperature: float = 0.4, max_tokens: int = 256, **kwargs)

generate(messages: List[Block], capabilities: List[Capability] = None, assert_capabilities: bool = True, **kwargs) → List[Block]

Call the LLM plugin’s generate method. Generate requests for plugin capabilities based on input parameters.

Parameters:

• messages – Messages to be passed to the LLM to construct the prompt.

• capabilities –

  Capabilities that will be asked of the LLM. See the docstring for steamship.plugins.capabilities.

  If default_capabilities was provided at construction, those capabilities will be requested unless overridden by this parameter.

• assert_capabilities – If True (default), raise a SteamshipError if the LLM plugin did not respond with a block that asserts what level capabilities were fulfilled at.

• kwargs – Options that can be passed to the LLM model as other parameters.

Returns:

a List of Blocks that are returned from the plugin.

Module contents

class steamship.agents.llms.OpenAI(client, model_name: str = 'gpt-3.5-turbo', temperature: float = 0.4, *args, generator: PluginInstance)

Bases: LLM

LLM that uses Steamship’s OpenAI plugin to generate completions.

NOTE: By default, this LLM uses the gpt-3.5-turbo model. Valid model choices are gpt-3.5-turbo and gpt-4.

client: Steamship

complete(prompt: str, stop: str | None = None, **kwargs) → List[Block]

Completes the prompt, respecting the supplied stop sequence.

Supported kwargs include: - max_tokens (controls the size of LLM responses)

generator: PluginInstance