DeveloperWeek New York 2026 Hackathon Submission Domain Roulette Track — clocks.works
"Turn Every Hour Into a Connected Intelligence."
graph_structured_hackathon is an autonomous, graph-powered intelligence system that maps relationships, surfaces hidden connections, and onboards complex knowledge structures through AI-driven graph reasoning. Built for the DeveloperWeek New York 2026 Hackathon, it demonstrates how graph-structured AI combined with an intelligent onboarding pipeline can transform raw, disconnected data into a navigable, queryable knowledge network — in real time.
Clone the repo, run the onboarding step once, then launch the Streamlit dashboard. Full setup takes under 5 minutes.
Modern data is inherently relational — but most tools treat it as flat:
graph_structured_hackathon automates the full pipeline — from raw data ingestion to interactive graph exploration:
Every hour of work becomes a richer, more connected intelligence layer.
clocks.works is not a clock application.
It is a system where every hour adds new nodes, new edges, and new intelligence to a living knowledge graph. The domain captures the core idea:
graph_structured_hackathon makes clocks.works literal: as the clock works, so does your data.
The system consists of two tightly integrated components:
onboard_graph.py): Ingests raw data, extracts entities and relationships, and constructs the initial graph structure — run once to bootstrap the knowledge basemain.py): An interactive interface for exploring the graph, querying relationships, visualizing node connections, and surfacing multi-hop insights| Component | Technology | Purpose |
|---|---|---|
| Onboarding Engine | Python | Ingests data, builds graph structure, stores nodes and edges |
| Graph Layer | Graph-structured data model | Relationship mapping, path discovery, reasoning |
| Interactive Dashboard | Streamlit | Query interface, visualization, insight surfacing |
| Dependencies | requirements.txt / requirements_onboarding.txt | Isolated dependency management per component |
flowchart TD
A[ Raw Data / Knowledge Input] --> B[ Onboarding Pipeline\nonboard_graph.py]
B --> C[ Entity Extraction\nNodes + Attributes]
B --> D[ Relationship Mapping\nEdges + Weights]
C --> E[ Graph Store\nStructured Knowledge Base]
D --> E
E --> F[ Graph Query Engine\nPath Discovery · Multi-hop Reasoning]
F --> G[ Streamlit Dashboard\nmain.py]
G --> H[ Visualization Layer\nNode Maps · Connection Explorer · Insights]
H --> A
The onboarding pipeline is the foundation of the system. Run once per knowledge domain:
onboard_graph.pyThis step transforms unstructured or semi-structured data into a queryable knowledge graph — without manual configuration.
Once onboarding is complete, the Streamlit dashboard (main.py) provides a full interactive interface:
The dashboard is designed to be accessible to both technical and non-technical users — graph intelligence without graph expertise.
Scenario: A team needs to understand how concepts in a new knowledge domain connect.
onboard_graph.py runs and extracts 150+ nodes and 400+ relationships in ~30 secondsgraph_structured_hackathon addresses a fundamental gap in knowledge management:
| Innovation | Detail |
|---|---|
| Automated Graph Onboarding | Zero-config pipeline that structures any knowledge domain into a graph without manual entity definition |
| Multi-hop Reasoning | Query engine that traverses indirect relationships invisible in flat data |
| Dual-Pipeline Architecture | Separate onboarding and runtime pipelines — clean separation of concerns, independently scalable |
| Streamlit-Native UX | Full graph exploration in a browser — no graph database expertise required |
| Isolated Dependency Management | Two requirements files ensure onboarding and runtime environments never conflict |
requirements_onboarding.txt — onboarding pipeline dependenciesrequirements.txt — Streamlit dashboard dependenciesProgress The full pipeline is functional: onboarding runs end-to-end, the graph is queryable, and the Streamlit dashboard renders live. Both components are independently runnable with documented setup steps.
Concept Automating the construction of knowledge graphs — historically a complex, expert-driven task — makes graph intelligence accessible to every team. The clocks.works framing adds a compelling dimension: the graph is never static, it grows smarter with every cycle. This positions the project at the intersection of AI, data infrastructure, and autonomous knowledge management.
Feasibility
The stack is entirely Python-based with standard dependencies. No proprietary infrastructure required. Deployment is a single streamlit run command after a one-time onboarding step. The architecture scales from a local laptop to a cloud-hosted service without code changes.
pip install -r requirements_onboarding.txt
pip install -r requirements.txt
python src/onboard_graph.py
Run this once to ingest and structure your knowledge domain into the graph.
streamlit run src/main.py
Open the URL shown in your terminal (default: http://localhost:8501) to explore the graph.
DeveloperWeek New York 2026 Hackathon Submission Domain Roulette Track — clocks.works
"Turn Every Hour Into a Connected Intelligence."
graph_structured_hackathon is an autonomous, graph-powered intelligence system that maps relationships, surfaces hidden connections, and onboards complex knowledge structures through AI-driven graph reasoning. Built for the DeveloperWeek New York 2026 Hackathon, it demonstrates how graph-structured AI combined with an intelligent onboarding pipeline can transform raw, disconnected data into a navigable, queryable knowledge network — in real time.
Clone the repo, run the onboarding step once, then launch the Streamlit dashboard. Full setup takes under 5 minutes.
Modern data is inherently relational — but most tools treat it as flat:
graph_structured_hackathon automates the full pipeline — from raw data ingestion to interactive graph exploration:
Every hour of work becomes a richer, more connected intelligence layer.
clocks.works is not a clock application.
It is a system where every hour adds new nodes, new edges, and new intelligence to a living knowledge graph. The domain captures the core idea:
graph_structured_hackathon makes clocks.works literal: as the clock works, so does your data.
The system consists of two tightly integrated components:
onboard_graph.py): Ingests raw data, extracts entities and relationships, and constructs the initial graph structure — run once to bootstrap the knowledge basemain.py): An interactive interface for exploring the graph, querying relationships, visualizing node connections, and surfacing multi-hop insights| Component | Technology | Purpose |
|---|---|---|
| Onboarding Engine | Python | Ingests data, builds graph structure, stores nodes and edges |
| Graph Layer | Graph-structured data model | Relationship mapping, path discovery, reasoning |
| Interactive Dashboard | Streamlit | Query interface, visualization, insight surfacing |
| Dependencies | requirements.txt / requirements_onboarding.txt | Isolated dependency management per component |
flowchart TD
A[ Raw Data / Knowledge Input] --> B[ Onboarding Pipeline\nonboard_graph.py]
B --> C[ Entity Extraction\nNodes + Attributes]
B --> D[ Relationship Mapping\nEdges + Weights]
C --> E[ Graph Store\nStructured Knowledge Base]
D --> E
E --> F[ Graph Query Engine\nPath Discovery · Multi-hop Reasoning]
F --> G[ Streamlit Dashboard\nmain.py]
G --> H[ Visualization Layer\nNode Maps · Connection Explorer · Insights]
H --> A
The onboarding pipeline is the foundation of the system. Run once per knowledge domain:
onboard_graph.pyThis step transforms unstructured or semi-structured data into a queryable knowledge graph — without manual configuration.
Once onboarding is complete, the Streamlit dashboard (main.py) provides a full interactive interface:
The dashboard is designed to be accessible to both technical and non-technical users — graph intelligence without graph expertise.
Scenario: A team needs to understand how concepts in a new knowledge domain connect.
onboard_graph.py runs and extracts 150+ nodes and 400+ relationships in ~30 secondsgraph_structured_hackathon addresses a fundamental gap in knowledge management:
| Innovation | Detail |
|---|---|
| Automated Graph Onboarding | Zero-config pipeline that structures any knowledge domain into a graph without manual entity definition |
| Multi-hop Reasoning | Query engine that traverses indirect relationships invisible in flat data |
| Dual-Pipeline Architecture | Separate onboarding and runtime pipelines — clean separation of concerns, independently scalable |
| Streamlit-Native UX | Full graph exploration in a browser — no graph database expertise required |
| Isolated Dependency Management | Two requirements files ensure onboarding and runtime environments never conflict |
requirements_onboarding.txt — onboarding pipeline dependenciesrequirements.txt — Streamlit dashboard dependenciesProgress The full pipeline is functional: onboarding runs end-to-end, the graph is queryable, and the Streamlit dashboard renders live. Both components are independently runnable with documented setup steps.
Concept Automating the construction of knowledge graphs — historically a complex, expert-driven task — makes graph intelligence accessible to every team. The clocks.works framing adds a compelling dimension: the graph is never static, it grows smarter with every cycle. This positions the project at the intersection of AI, data infrastructure, and autonomous knowledge management.
Feasibility
The stack is entirely Python-based with standard dependencies. No proprietary infrastructure required. Deployment is a single streamlit run command after a one-time onboarding step. The architecture scales from a local laptop to a cloud-hosted service without code changes.
pip install -r requirements_onboarding.txt
pip install -r requirements.txt
python src/onboard_graph.py
Run this once to ingest and structure your knowledge domain into the graph.
streamlit run src/main.py
Open the URL shown in your terminal (default: http://localhost:8501) to explore the graph.