Open hongjin-su opened 9 months ago
We have added the tool matching function, i.e., api embedding during tool registration (./multiagents/tools/api_retrieval.py) and api matching at the start of tree search (./multiagents/tools/retriever.py).
We will (1) release the local embedding model (sentence_transformer) after adding sufficient optimization tools and (2) further test for the most suitable time to apply the "tool matching" function (e.g., at the beginning of diagnosis or for every llm inference).
Pls. the training data for localized text embedding model will be released soon
As titled, thanks!