AI Agent 007 :Tooling for success

The project aims to enhance the capabilities of Large Language Models by enabling them to generate a specific set of tools in response to a given query or task. This challenge was presented as a problem statement during the DevRev at Inter IIT Tech Meet 12.0.

Problem Statement

A Language model L has a set of tools T, and a user query Q is given. To answer query Q, we need to use existing tools. You need to output the subset of tools to be used to answer the query, the arguments that these tools should be called with, and how to compose the tools to answer the query.

Report

The repository Report contains the details and approaches used for the project.

Code/Final Model/2_step_prompting.py

code for prompting technique using two steps

data_generation_and_cleaning.py

final code for dataset generation for fine tuning

Final_traindata.csv

Training dataset

Final_testdata.csv

Testing dataset

RAG_Final.py

code containing RAG model

fine_tune.py

general code for fine tuning models

gpt_3_5_fine_tuning.py

specific code for gpt 3.5 finetuning

Metric_Created.py

code for metric

Final_Pipeline.py

final pipeline compiling all above codes

Inference.py

code for inference on test data and real word usage