Continuing the discussion from 
OPEN Round of 32 Challenge: GDS 32! July 2–8, 2026:
GraphAcademy Cup Submission
Topic Title: Round of 32: TirthaPathAI — South Indian Pilgrimage Route Planner – India (
)
GraphAcademy Cup Team Profile Link
Team Profile Link: Neo4j team member
GraphAcademy Public Profile Username
Public Profile Username: guna
Anonymous GraphAcademy profiles are not el@gunagible for weekly LEGO prizes.
Country
Country: India
Learning Path Completed
Learning Path: Graph Data Science
-
Get Started with Graph Data Science — https://graphacademy.neo4j.com/c/c2499149-a6e2-4c31-a2f2-7806cf9c1131/ -
Path Finding with GDS — https://graphacademy.neo4j.com/c/fcff190b-1ea3-4e01-aec4-dd52575ac1c0/
Project Name
Project Name: TirthaPathAI — South Indian Pilgrimage Route Planner using Neo4j GDS
Project Description
Project Description (32–320 words):
My mother asked: "Which pilgrimage places can we visit, and what is the best route?" I had no answer. India has a billion people and thousands of sacred sites — yet most young Indians in their 20s have little idea where these places are, how far apart, or which city is the best base.
After completing KiranaAI for Week 3, the GDS Round of 32 clicked: model South Indian sacred sites as a 32-node knowledge graph and apply Graph Data Science to answer real pilgrim questions with real algorithms.
TirthaPathAI is a LangGraph ReAct agent where devotees ask in Hindi, Telugu, or English. The graph has exactly 32 nodes (20 temples + 12 city hubs) and 32 ROAD_CONNECTS relationships with distance_km and duration_hours. Five GDS algorithms answer five distinct questions:
·····················································
· Devotee (Hindi / Telugu / English) ·
·····················································
│
▼
·····················································
· LangGraph ReAct Agent · DeepSeek LLM ·
·····················································
│ │ │
▼ ▼ ▼
··············· ··············· ···············
· gds_path · · gds_analysis· ·gds_community·
· · · · · ·
· Dijkstra · · PageRank · · Louvain ·
· BFS · · Betweenness · · (circuits) ·
··············· ··············· ···············
│ │ │
·································
│
▼
·····················
· Neo4j GDS ·
· 32 nodes ·
· 32 edges ·
·····················
The GDS courses revealed something surprising: Vijayawada ranks #1 as best hub not because it has the most roads, but because it sits at the crossroads of AP, Telangana, and Tamil Nadu — exactly what Betweenness Centrality measures. The three pilgrimage circuits emerged from Louvain without a single geographic label being given. The algorithms only see connections. The geography emerged on its own.
Additional Notes (Optional)
Additional Notes:
There is no cultural bias in this project. The same graph architecture works for any country's sacred or heritage sites — shrine routes in Japan, pilgrimage paths in Italy, heritage circuits in Egypt. The algorithms don't care what the nodes represent.
I am also personally planning to visit Tirumala soon — perhaps the universe had a hand in this project landing exactly here. 
Special thanks to my mother — her question about where to visit next was the seed of this entire project.
Screenshots
The Pilgrimage Network
Screenshot 1 — Full 32-Node Knowledge Graph
The complete network: 20 temples + 12 city hubs, color-coded by deity (blue = city hub, teal = Vishnu, purple = Shiva, red = Shakti, amber = Saraswati). Edge labels show road distances in km.
Q1 — "Which Jyotirlinga is closest to Vijayawada?"
Screenshot 2 — Agent Thinking: Two Dijkstra Calls
The thinking panel shows two Dijkstra calls fired autonomously — Vijayawada → Srisailam and Vijayawada → Rameswaram — without the user asking for a comparison. The agent decided comparison was needed to answer "which is closest."
Screenshot 3 — Answer: Srisailam Wins
Natural language answer from raw GDS output: Srisailam at 487 km / 2 hops vs Rameswaram at 1045 km / 6 hops. The algorithm ran; the agent explained.
Screenshot 4 — Both Paths Highlighted on the Live Graph
Both Dijkstra paths rendered orange simultaneously — short Srisailam route and long Rameswaram route. Route nodes accumulate across calls so the full comparison shows in one view.
Q2 — "Best city to be based in for pilgrims?"
Screenshot 5 — Agent Thinking: Betweenness + Closeness Called Together
Agent calls Betweenness and Closeness centrality together — recognising that "best base city" needs both: who appears most on shortest paths (Betweenness) and who can reach everything fastest (Closeness).
Screenshot 6 — Hub Rankings: Vijayawada #1
Vijayawada #1 on both metrics: Betweenness 313, Closeness 0.344. Agent also surfaces regional alternatives — Tirupati for Tamil Nadu, Hyderabad for Telangana. No rules hardcoded — graph structure produced this.
Screenshot 7 — Pilgrimage Circuits + Recommended Trip Plan
Louvain reveals three natural circuits — AP, Telangana, Tamil Nadu — without any geographic labels being given. The agent wraps the community output into a recommended multi-day trip plan table.
Repository / Demo Link (Optional)
GitHub / Demo URL (Optional): GitHub - chakka-guna-sekhar-venkata-chennaiah/tirtha-path-ai · GitHub
Next: node coloring by PageRank/Betweenness score on the live graph, and expanding the network northward to cover all four Char Dham circuits.
Submission Checklist
-
I completed the Graph Data Science learning path.
-
I am participating in the GraphAcademy Cup.
-
I included my Team Profile Link.
-
I included my GraphAcademy Public Profile Username.
-
My GraphAcademy profile is public and eligible for prize verification.
-
My project contains 32 nodes or fewer.
-
My project contains 32 relationships or fewer.
-
I included 1–3 screenshots (added 7 screenshots for depth explainability
). -
My project description is 32–320 words.
Contest Reminder
One LEGO prize winner will be selected from the Round of 32 submissions based on creativity, application of Graph Data Science concepts, and overall presentation.
Terms & Conditions: Terms & Conditions | GraphAcademy Cup
FAQ: