jsrinivasa / blaze

This is the repository for composable NLP Inference Pipeline tool Blaze

BLAZE - Building Language Applications with eaZE

BLAZE Drag-and-Drop - README | BLAZE Conversational AI - README

  

The goal of BLAZE is to make the lives of data scientists easier, filling-in a required niche.

Natural Language Processing (NLP) pipelines reuse many of the same components arranged in different orders. The purpose each NLP model serves varies from use-case to use-case. However, these NLP models are not standardized in terms of their inputs, outputs, and hardware requirements. As a result, it is very difficult to interchange and combine NLP models, especially without introducing significant amounts of code. This lack of standardization causes NLP pipelines to be very rigid. Their lack of flexibility makes it difficult to compose, modify, and add functionality. 

To solve this problem, Blaze that allows for the modular creation and composition of NLP pipelines. Each component of the NLP Pipeline can be implemented as "building block" (for example, a microservice). These building blocks will have standardized inputs and outputs, and they can easily be assembled in varying orders. The order and choice of these specific blocks result in varying pipelines, built for unique use-cases.

Features

Here's BLAZE's current functionalities:

Use Case #1 - Developer

• Drag-and-Drop Pipeline Builder

  • Visual builder for configuring and deploying pipelines
  • Allows for building from scratch (adding, connecting components)
  • Allows for uploading existing pipelines (visualize "recipes")
  • Converts custom recipe into downloadable YAML config file
  • Generates and launches custom NLP pipeline solution

• Conversational AI (Webex Bot)

  • Interface with BLAZE to specify pipeline components
  • Generate and launch custom NLP pipeline solution

    

Use Case #2 - Business

• Visual Dashboard Web App

  • Visual representation of generated pipeline
  • Supports semantic search, summarization, file upload, etc.

• Conversational AI (Webex Bot)

  • Interact with generated pipeline through natural language
  • Can choose knowledge base (ex. upload file, view all files)
  • Can index model and retrieve results (ex. summarize this doc)
  • Can retrieve knowledge base, model, and metrics info

    

Use Case #3 - Researcher

• Model/Knowledge Base Benchmarking

  • Benchmark selected model on selected knowledge base
  • Gives latency (avg time/question, as well as generates real-time graph)
  • Gives accuracy (num correct, num total, % correct, % progress)
  • Displays incorrect questions

• Model/Knowledge Base Comparison

  • Benchmark multiple models on selected knowledge base
  • Gives latency (avg time/question, as well as generates real-time graph)
  • Gives side-by-side accuracy (num correct, num total, % correct, % progress)
  • Displays incorrect questions

• Model/Knowledge Base Metrics

  • Compute and display scientific metrics
  • Currently Supported Metrics: BertScore, Bleu, Rouge

    

GIFS/Screenshots Coming Soon!

Over the coming weeks, this platform will be further fleshed out with more exciting features 😄.

  

Install using Pyenv: Local Development

Install Dependencies For Ubuntu/Debian sudo apt-get install -y make build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev python-openssl

For Fedora/CentOS sudo yum install gcc zlib-devel bzip2 bzip2-devel readline-devel sqlite sqlite-devel openssl-devel xz xz-devel libffi-devel

Install Pyenv curl https://pyenv.run | bash

Install Python version 3.9.16 using below command pyenv install -v 3.9.16

check python version installed in pyenv using command pyenv versions

Create virtual environment with specific version using below command pyenv virtualenv 3.9.16 venv

Activate virtual environment using pyenv local venv or pyenv activate <environment_name>

You can verify this by running the following: pyenv which python

Install packages and modules using pip install -r requirements.txt

Run the Program with: python build.py

Now, a link should appear (ex. Dash is running on http://127.0.0.1:5000/). Run this link in your browser to open the dashboard!

python run.py yaml/<file-name>.yaml

  

Install using conda: Local development

NOTE: The use of Conda may require the use of Anaconda Commercial Edition to comply with Anaconda's Terms of Service if your use is considered commercial according to Anaconda. More information about Anaconda's Terms of Service and what qualifies as commercial usage can be found here: https://www.anaconda.com/blog/anaconda-commercial-edition-faq/

Create your conda environment with

conda env create -f aski_env.yml

Then, activate your conda environment with

conda activate aski

Using the Conversational AI

Build the conversational model with

python -m conv build

(This step may take about 15 minutes without a GPU)

And run the AI in the command line with

python -m conv converse

To run the webex bot server, start an ngrok tunnel with

ngrok http 8080 --region=eu

It is important that the region is not 'us'.

Set the environment variables BOT_ACCESS_TOKEN (received when you register your bot) and WEBHOOK_URL (generated by ngrok) and change to the conv directory. Then run

python webex_bot_server.py

Sources for training data

• greet: food_ordering/greet/train.txt
• exit: food_ordering/exit/train.txt
• ask_question: manual data entry + mindmeld data augmentation
• get_summary: manual data entry + mindmeld data augmentation
• upload_data: manual data entry + mindmeld data augmentation

Supported Models, Knowledge Bases

Installing Elasticsearch

Navigate to Elasticsearch Installation and follow the instructions according to your specific setup.

NOTE: ASKI does not currently support Elasticsearch 8 or higher!

In order to launch elasticsearch, open a new terminal, navigate to the elasticsearch directory, and run either of the following:

place the elastic folder as same hirerachy as ASKI project folder to start elastic service when running run.py file insted of running it manually using the commands below

• ./bin/elasticsearch (Linux/Mac)
• .\bin\elasticsearch.bat (Windows)

Now, leave this terminal window open!

  

Installing ColBERT

Clone the following GitHub Repo (new-api branch) into the ColBERT folder.

NOTE: There might be some issues with environments, these will be resolved by next commit!

Once downloading the ColBERT files, make sure to uncomment the three lines near the top of the file ColBERTSearch.py. More instructions are detailed at the top of this file.

After this, navigate to the get_sidebar() function in app_elements.py and make sure to toggle the "Disabled" option next to ColBERT. You should be good to go now!

  

Installing Knowledge Graph

Stay tuned, support for this is coming soon!