Common Parameters
All model classes accept these parameters:| Parameter | Type | Default | Description |
|---|---|---|---|
id | str | — | Model identifier (e.g., "gpt-4o"). |
api_key | str | env var | API key. Defaults to provider’s env var. |
base_url | str | None | Override API base URL. |
temperature | float | None | Sampling temperature (0.0–2.0). |
max_tokens | int | None | Max output tokens. |
top_p | float | None | Nucleus sampling. |
timeout | float | None | Request timeout in seconds. |
retries | int | 0 | Retry attempts. |
delay_between_retries | int | 1 | Base retry delay in seconds. |
exponential_backoff | bool | False | Exponential backoff for retries. |
cache_response | bool | False | Cache responses locally. |
cache_ttl | int | 3600 | Cache TTL in seconds. |
cache_dir | str | ".cache/models" | Cache directory. |
Methods
| Method | Returns | Description |
|---|---|---|
invoke(messages, assistant_message, **kwargs) | ModelResponse | Sync call. |
ainvoke(messages, assistant_message, **kwargs) | ModelResponse | Async call. |
invoke_stream(messages, assistant_message, **kwargs) | Iterator[ModelResponse] | Sync streaming. |
ainvoke_stream(messages, assistant_message, **kwargs) | AsyncIterator[ModelResponse] | Async streaming. |
ModelResponse
| Field | Type | Description |
|---|---|---|
content | str | Response text. |
tool_calls | List[ToolCall] | Tool calls requested by model. |
usage | Usage | Token usage metrics. |
stop_reason | str | Why generation stopped. |
Provider Classes
| Import | Provider |
|---|---|
from definable.model.openai import OpenAIChat | OpenAI |
from definable.model.anthropic import Claude | Anthropic |
from definable.model.google import Gemini | |
from definable.model.deepseek import DeepSeekChat | DeepSeek |
from definable.model.mistral import MistralChat | Mistral |
from definable.model.moonshot import MoonshotChat | Moonshot |
from definable.model.xai import xAI | xAI |
from definable.model.perplexity import Perplexity | Perplexity |
from definable.model.ollama import Ollama | Ollama |
from definable.model.openrouter import OpenRouter | OpenRouter |
from definable.model.openai_like import OpenAILike | Custom |
String Shorthand
from definable.model.utils import resolve_model_string
model = resolve_model_string("openai/gpt-4o")
model = resolve_model_string("gpt-4o") # defaults to OpenAI
"provider/model-id" or bare "model-id" (defaults to OpenAI).