Provide a MQL Database tool in LangChain.
LangChain provides a range of agents and toolkits that enable interaction with different tools. For instance, LangChain offers SQL Database Tool, which can be used to interact with SQL databases. These tools are then utilized by LLMs (Large Language Models) such as GPT-4 to perform a series of tasks as an agent.
Retrieving data from MongoDB using a Natural Language query can be difficult when the data is stored in collections without vectors. To perform this we prompt the LLM with database schema, and user query has is described in the doc https://docs.google.com/document/d/1A5qFxFUtEkwk7dQfYn7nxFrziiJMpXkDAJcViveMPzY/edit. MongoDB Compass has a Natural Language query feature which is in the beta version, which uses similar methodology.
The scope of this work is
(a) Create a brief design doc and get review from the product stakeholders
(b) Commit code based on lgtm'ed design doc for MQL_agent_tool similar to code in `langchain_community.agent_toolkits.create_sql_agent`
- is related to
-
PYTHON-4602 LangChain: Native Parent Child retriever for MongoDB
-
- Closed
-
- related to
-
PYTHON-4339 LangChain: Enhance support for Self querying retriever
-
- Backlog
-