LLM API Reference¶

LLMConfig¶

LLM provider configuration.

from flowgentra_ai import LLMConfig

Constructor¶

LLMConfig(
    provider: str,          # "openai", "anthropic", "mistral", "groq", "ollama", "huggingface", "azure"
    model: str,             # e.g. "gpt-4", "claude-3-opus-20240229"
    api_key: str = "",
    temperature: float | None = None,   # 0.0-2.0
    max_tokens: int | None = None,
    top_p: float | None = None,         # 0.0-1.0
)

Properties (read/write)¶

Property	Type	Description
`provider`	`str`	Provider name (read-only)
`model`	`str`	Model identifier (read-only)
`api_key`	`str`	API key (read-only)
`temperature`	`float \\| None`	Response randomness
`max_tokens`	`int \\| None`	Max response tokens
`top_p`	`float \\| None`	Nucleus sampling

Methods¶

Method	Description
`set_response_format(format)`	Set structured output format

LLMClient¶

Client for sending messages to an LLM.

from flowgentra_ai import LLMClient

Class Methods¶

Method	Returns	Description
`LLMClient.from_config(config)`	`LLMClient`	Create from an `LLMConfig`

Methods¶

Method	Returns	Description
`chat(messages)`	`Message`	Send messages, get response
`chat_with_usage(messages)`	`(Message, TokenUsage \\| None)`	Chat with token usage stats
`chat_with_tools(messages, tools)`	`Message`	Chat with function calling
`cached(max_entries=100)`	`LLMClient`	Wrap with response cache
`with_fallback(client)`	`LLMClient`	Add fallback provider
`with_retry(max_retries=3)`	`LLMClient`	Add retry with backoff

Message¶

A message in a conversation.

from flowgentra_ai import Message

Constructor¶

Message(role: str, content: str, tool_call_id: str | None = None)

Factory Methods¶

Method	Returns	Description
`Message.system(content)`	`Message`	System message
`Message.user(content)`	`Message`	User message
`Message.assistant(content)`	`Message`	Assistant message
`Message.tool(content, tool_call_id=None)`	`Message`	Tool result message

Properties¶

Property	Type	Description
`role`	`str`	`"system"`, `"user"`, `"assistant"`, or `"tool"`
`content`	`str`	Message text (read/write)
`tool_call_id`	`str \\| None`	Associated tool call ID

Methods¶

Method	Returns	Description
`is_system()`	`bool`	Check role
`is_user()`	`bool`	Check role
`is_assistant()`	`bool`	Check role
`is_tool()`	`bool`	Check role
`has_tool_calls()`	`bool`	Whether message has tool calls
`tool_calls()`	`list[ToolCall]`	Get tool calls

ToolCall¶

A tool call from an LLM response.

from flowgentra_ai import ToolCall

Constructor¶

ToolCall(id: str, name: str, arguments: Any)

Properties¶

Property	Type	Description
`id`	`str`	Unique call ID
`name`	`str`	Tool name
`arguments`	`Any`	Call arguments (dict)

ToolDefinition¶

Tool definition for LLM function calling.

from flowgentra_ai import ToolDefinition

Constructor¶

ToolDefinition(name: str, description: str, parameters: Any)

Properties¶

Property	Type	Description
`name`	`str`	Tool name
`description`	`str`	Tool description
`parameters`	`Any`	JSON Schema for parameters

TokenUsage¶

Token usage statistics.

from flowgentra_ai import TokenUsage

Constructor¶

TokenUsage(prompt_tokens: int, completion_tokens: int)

Properties¶

Property	Type	Description
`prompt_tokens`	`int`	Input token count
`completion_tokens`	`int`	Output token count
`total_tokens`	`int`	Total tokens

Methods¶

Method	Returns	Description
`estimated_cost(model)`	`float \\| None`	Estimated cost in USD

ResponseFormat¶

Structured output format for LLM responses.

from flowgentra_ai import ResponseFormat

Factory Methods¶

Method	Description
`ResponseFormat.text()`	Plain text (default)
`ResponseFormat.json()`	Force JSON output
`ResponseFormat.json_schema(name, schema)`	JSON matching a schema