handle_tool_call
async handle_tool_call(func_name: str, func_args: Dict[str, Any], session_id: str | None = None, *, sub_user_id: str | None = None) → Tuple[Any, bool, bool]
Handle a built-in Qveris tool call from an LLM response.
- 参数:
- func_name -- The name of the function/tool to call
- func_args -- The arguments parsed from the LLM response
- session_id -- Optional session ID for tracking
- sub_user_id -- Host-controlled OAuth identity; never read from model arguments.
- 返回:
- result: the tool output (None if not handled)
- is_error: True if an error occurred
- handled: True if this was a Qveris tool and was processed
- 返回类型: Tuple of (result, is_error, handled) where
备注
- params_to_tool may be either a dict (canonical) or a JSON string (legacy).
- If func_name is not a Qveris built-in, (None, False, False) is returned so that callers can route to their own tool handlers.
class qveris.PublicDiscoveryClient(config: QverisConfig | None = None, *, transport: AsyncBaseTransport | None = None)
Explicit credential-free Discover/Inspect. Authenticate with a new client.
Re-inspect selected IDs and review current inputs and a fresh quote before explicitly calling through that authenticated client. Never reuse pub_ IDs. Use transport for injection; borrowed HTTP clients may retain identity.
