Skip to main content
Back
2023–2024Open-SourceLLMs

Reverse Chain

Implemented the Reverse Chain algorithm to enhance LLM capabilities in multi-API action plan, integrating langchain, FAISS (vector database), BERT embeddings, and OpenAI API.

Flow diagram showing a goal, API retrieval, and an ordered action plan.

Project Overview

Implemented the Reverse Chain algorithm to enhance LLM capabilities in multi-API action plan, integrating langchain, FAISS (vector database), BERT embeddings, and OpenAI API.

BERT embeddings and FAISS retrieve candidate APIs. The planner uses their inputs and outputs to work backward through dependencies.

The resulting plan connects multiple API actions needed for a request.

Engineering decision

BERT embeddings and FAISS retrieve candidate APIs. The planner uses their inputs and outputs to work backward through dependencies.

Outcome

The resulting plan connects multiple API actions needed for a request.

Key Features

  • Plans backward from the requested result
  • BERT embeddings with FAISS retrieve candidate APIs
  • LangChain and the OpenAI API drive the planner
  • Chains several dependent API calls into one plan

Technology Stack

PythonLangChainFAISSBERTOpenAI API
Next projectProcureJs Website