LLM Integration

LLM Module

LLM Implementation

openchatbi.llm.llm.list_llm_providers() → list[str]

List configured LLM provider names (if any).

openchatbi.llm.llm.get_embedding_model(provider: str | None = None)

Get embedding model from config (optionally scoped to a provider).

openchatbi.llm.llm.get_default_llm(provider: str | None = None)

Get default LLM from config (optionally scoped to a provider).

openchatbi.llm.llm.get_llm(provider: str | None = None)

Get the chat model to use (alias for get_default_llm).

openchatbi.llm.llm.get_text2sql_llm(provider: str | None = None)

Get text2sql LLM from config (optionally scoped to a provider).

openchatbi.llm.llm.get_analysis_llm(provider: str | None = None)

Get the data analysis LLM from config (optionally scoped to a provider).

Falls back to the default LLM when no dedicated analysis_llm is configured.

openchatbi.llm.llm.call_llm_chat_model_with_retry(chat_model: BaseChatModel, messages, streaming_tokens=False, bound_tools=None, parallel_tool_call=False, metadata: dict | None = None)

Calls a language model chat endpoint with retry logic.

Retries up to 3 times if there are errors or invalid tool calls.

Parameters:

• chat_model – The chat model to invoke.

• messages (list) – List of messages to send to the model.

• streaming_tokens (bool, optional) – flag to indicate whether or not to show streaming tokens in UI.

• bound_tools (list, optional) – List of valid tool names that can be called.

• parallel_tool_call (bool, optional) – whether or not to call multiple tools in parallel.

Returns:

The model response or None if all retries failed.

Return type:

AIMessage or None