Current Features

BLAZE currently supports the following building blocks and sample applications: | Category | Current Components| |:--------------:|:-----------------------------------------------------:| | Data(sets) | SQUAD 2.0, BillSum, XSum, CNN Dailymail, HuggingFace | | User Data | TXT files, WebEx Transcripts (Live, Pre-Recorded) | | Processing | Enterprise LLM Gateway, coming soon! | | Models | GPT-3 Variants, ColBERT, ElasticBERT, HuggingFace | | Metrics | BertScore, Bleu, Rouge, Accuracy, Latency | | Interfaces | React App, Dash App, WebEx Bot, WebEx Plugin, POSTMAN |

HF* - BLAZE's flexible design enables seamless integration of new components, many of which are underway!

Data(sets)* - Summarization datasets are currently being stabilized, but will be ready with a few tweaks!

In addition, we have several sample applications, each of which are detailed below:

1. Semantic Search on User-Uploaded Documents - TechBlog: Building WebApps with BLAZE
2. Benchmarking ElasticBERT Q/A on SQUAD 2.0 - TechBlog: Building WebApps with BLAZE
3. Transcript Analysis with WebEx ChatBot - TechBLog: Building ChatBot with BLAZE
4. Embedded WebEx Meeting Assistant Plugin - TechBlog: Building Plugins with BLAZE

Some of our more-specific offerings include:

Feature	Sample
(Tool) Drag-and-Drop Builder Visual builder for configuring and deploying pipelines Allows for building from scratch or uploading existing pipelines (visualize "recipes") Converts recipe into downloadable YAML config file Generates and launches custom NLP pipeline solution
(Interface) Dashboard WebApp Visual representation of generated pipeline Supports semantic search, summarization, file upload, etc.
(Interface) Conversational AI Interact with generated pipeline through natural language Can be purely functional or conversational Powered by Cisco MindMeld
(Function) Model Benchmarking Benchmark selected model on select knowledge base(s) Gives metrics such as latency, accuracy (real-time graphs) Displays incorrect questions and trends in performance
(Function) Model Comparison Benchmark multiple models on selected knowledge base(s) Gives metrics such as latency, accuracy (real-time graphs) Compares trends in performance (ex. areas of strength)