Artificial Intelligence Tools for Knowledge-Intensive Tasks (AIKIT)
################################################################################ Author: Darrell O. Ricke, Ph.D. (mailto: [email protected]) Copyright: Copyright (c) 2024 Massachusetts Institute of Technology License: GNU GPL license (http://www.gnu.org/licenses/gpl.html)
RAMS request ID 1028809
DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited. This material is based upon work supported by the Department of the Air Force under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Department of the Air Force.
© 2024 Massachusetts Institute of Technology.
Subject to FAR52.227-11 Patent Rights - Ownership by the contractor (May 2014)
The software/firmware is provided to you on an As-ls basis Delivered to the U.S. Government with Unlimited Rights, as defined in DFARS Part 252.227-7013 or 7014 (Feb 2014). Notwithstanding any copyright notice, U.S. Government rights in this work are defined by DFARS 252.227-7013 or DFARS 252.227-7014 as detailed above. Use of this work other than as specifically authorized by the U.S. Government may violate any copyrights that exist in this work.
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. ################################################################################
Summary
This is the Artificial Intelligence Tools for Knoledge-Intensive Tasks (AIKIT) system. AIKIT is packaged as both Docker and Singularity continers. AIKIT includes Large Language Models (LLM), LangChain, Jupyter notebooks, Retrieval-Augmented Generation (RAG), and Ruby-on-Rails web interface.
Setup
Customize AIKIT_UI/config/database.yml for your choice of relational database
tar -cf AIKIT_UI.tar AIKIT_UI # Note builds AIKIT_UI.tar file for the Ruby on Rails GUI
Conda setup
cd aikit_conda
sh < aikit_conda.cmds
Singularity
To build:
singularity build aikit.sif aikit.def
singularity build --sandbox aikit_box aikit.def Note: builds Singularity sandbox
To run:
singularity run --nv -B io/:/io/ -B <squashfs>:/io/hub/<model>:image-src=/ aikit.sif <Your program details>
Example LLM Python application:
singularity run --nv -B io/:/io/ -B Llama-2-7b-chat-hf.sqsh:/io/hub/models--meta-llama--Llama-2-7b-chat-hf:image-src=/ aikit.sif python /io/llama2_cli.py "How to cook fish?"
Jupyter notebook example:
singularity run --nv -B io/:/io/ -B falcon-7b.sqsh:/io/hub/models--tiiuae--falcon-7b:image-src=/ aikit_box jupyter notebook --allow-root --ip='*' --NotebookApp.token='' --NotebookApp.password=''
Singularity AIKIT Rails interface example:
Place LLM models in io/hub
singularity run --nv -B io/:/io/ aikit.sif "rails s -b 0.0.0.0"
AIKIT web interface on port 3000 on hosting server
Docker
To build:
docker build . -t aikit:latest
To run:
Note that Docker will accidently collide with the bridge network port, so the following file is highly recommended for running:
./docker_up.cmds
Docker LLM Python example:
docker run --gpus all -v /data/da23452/llm/aikit/io:/io -v /data/da23452/llm/llama2/models--meta-llama--Llama-2-7b-chat-hf:/io/hub/models--meta-llama--Llama-2-7b-chat-hf aikit:latest python llama2_cli.py "How to cook pasta?"
Example Jupyter notebook using LangChain:
LangChain_example.ipynb
Docker AIKIT Rails interface example:
Place LLM models in io/hub
docker run --gpus all -v ${PWD}/io:/io aikit:latest rails s -b 0.0.0.0
or
docker volume create --name io -o type=none -o o=bind -o device=${PWD}/io
./docker_up.cmds
AIKIT web interface on port 3000 on hosting server
REST API Documentation:
The AIKIT REST API is a standard Ruby on Rails REST API.
GET /
.json returns all records for a table GET /Examples: POST /books/1.json {"authenticity_token"=>"[FILTERED]", "book"=>{"title"=>"Book title", "pages"=>"12"}, "commit"=>"Create Book"} PATCH /books/1.json {"authenticity_token"=>"[FILTERED]", "book"=>{"title"=>"Book title 2", "pages"=>"14"}, "commit"=>"Update Book", "id"=>"1"}
Creating a vector store collection: GET /documents/add_favorites {"utf8"=>"✓", "example_length"=>"10", "col_ids"=>["1", "2", "3", "4", "5", "6"], "fav_list_name"=>"", "list_name"=>"default", "col_name"=>"demo", "vec_name"=>"FAISS", "parameter_set_id"=>{"parameter_set_id"=>"2"}, "commit"=>"Create collection"}
Running LLM RAG question(s): POST "/llm_questions/1/query {"utf8"=>"✓", "authenticity_token"=>"JYaf...", "llm_id"=>{"llm_id"=>"4"}, "parameter_set_id"=>{"parameter_set_id"=>"1"}, "collection_id"=>{"collection_id"=>"1"}, "controller"=>"llm_questions", "action"=>"query", "id"=>"1"}
AIKIT Table names are: chains groups qualifications user_groups collection_documents images responses user_qualifications collections layouts roles user_questions differences llm_evaluations sources user_reads documents llm_questions templates user_roles favorite_lists llms test_questions user_tests favorites parameter_sets test_sets users folders parameters topics
