Skip to content

Updated readme to include architecture diagram #113

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tekrajchhetri merged 2 commits into from
Apr 15, 2026
Merged

Updated readme to include architecture diagram #113

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c4525df
Add architecture section to README
tekrajchhetri Apr 15, 2026
b8c81ee
Add citation information to README
tekrajchhetri Apr 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 28 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,14 +2,28 @@

`structsense` is a multi-agent system for extracting structured information from unstructured text and documents. It orchestrates a configurable pipeline of AI agents — extractor → alignment → judge → human feedback — each driven by a single YAML config file.

**Documentation:** [docs.brainkb.org](http://docs.brainkb.org/structsense_overview.html)
**License:** [Apache 2.0](LICENSE.txt)

If you find this work useful or build upon it, please consider citing:

```bibtex
@misc{chhetri2025structsensetaskagnosticagenticframework,
title = {STRUCTSENSE: A Task-Agnostic Agentic Framework for Structured Information Extraction with Human-In-The-Loop Evaluation and Benchmarking},
author = {Tek Raj Chhetri and Yibei Chen and Puja Trivedi and Dorota Jarecka and Saif Haobsh and Patrick Ray and Lydia Ng and Satrajit S. Ghosh},
year = {2025},
eprint = {2507.03674},
archivePrefix= {arXiv},
primaryClass = {cs.CL},
url = {https://arxiv.org/abs/2507.03674}
}
```

---

## Table of Contents

- [Features](#features)
- [Architecture](#architecture)
- [Installation](#installation)
- [Quick Start](#quick-start)
- [Usage](#usage)
Expand Down Expand Up @@ -46,6 +60,18 @@

---

## Architecture

The figure below illustrates the overall architecture of **StructSense**.


<img width="2571" height="823" alt="with_arch_search-with_search_api drawio (1)" src="https://github.com/user-attachments/assets/8f652c10-4301-4193-b1c8-f750229f972c" />

@gemini-code-assist gemini-code-assist Bot Apr 15, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The alt text for the architecture diagram is currently a raw filename. Using a descriptive title like "StructSense Architecture Diagram" improves accessibility for users with screen readers and provides better context if the image fails to load.

Suggested change
<img width="2571" height="823" alt="with_arch_search-with_search_api drawio (1)" src="https://github.com/user-attachments/assets/8f652c10-4301-4193-b1c8-f750229f972c" />
<img width="2571" height="823" alt="StructSense Architecture Diagram" src="https://github.com/user-attachments/assets/8f652c10-4301-4193-b1c8-f750229f972c" />


StructSense integrates a local concept mapping service, which can also be used independently. The service is available here:
- https://github.com/sensein/search_hybrid

---

## Installation

```bash
Expand Down Expand Up @@ -464,7 +490,7 @@ By default (`skip_alignment_llm=None`), the alignment LLM is **automatically byp
- `CONCEPT_MAPPING_BACKEND=local` (which is the default)
- Task type is `ner`, `keyphrase_extraction`, `resource`, or `structured_extraction`

When bypassed, the concept mapping tool is called directly from Python in one batch (4000 concept/request--see [https://github.com/sensein/search_hybrid/tree/dev](https://github.com/sensein/search_hybrid/tree/dev)) — much
When bypassed, the concept mapping tool is called directly from Python in one batch (4000 concept/request--see [https://github.com/sensein/search_hybrid](https://github.com/sensein/search_hybrid)) — much
faster than running the LLM. The output records `alignment_method: "direct_tool_call"`.

```bash
Expand Down
Loading