1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
| ### Arguments of Doc_QA class ###
"""
Args:
**embeddings (str, optional)**: the embeddings used in query and vector storage. Defaults to "text-embedding-ada-002".\n
**chunk_size (int, optional)**: chunk size of texts from documents. Defaults to 1000.\n
**model (str, optional)**: llm model to use. Defaults to "gpt-3.5-turbo".\n
**verbose (bool, optional)**: show log texts or not. Defaults to False.\n
**language (str, optional)**: the language of documents and prompt, use to make sure docs won't exceed
max token size of llm input.\n
**search_type (str, optional)**: search type to find similar documents from db, default 'merge'.
includes 'merge', 'mmr', 'svm', 'tfidf', also, you can custom your own search_type function, as long as your
function input is (query_embeds:np.array, docs_embeds:list[np.array], k:int, relevancy_threshold:float, log:dict)
and output is a list [index of selected documents].\n
**record_exp (str, optional)**: use aiido to save running params and metrics to the remote mlflow or not if record_exp not empty, and set record_exp as experiment name. default "".\n
**system_prompt (str, optional)**: the system prompt that you assign special instruction to llm model, so will not be used
in searching relevant documents. Defaults to "".\n
**max_input_tokens (int, optional)**: max token length of llm input. Defaults to 3000.\n
**temperature (float, optional)**: temperature of llm model from 0.0 to 1.0 . Defaults to 0.0.\n
"""
|