LlamaIndex Urban Land Tender Valuation Agent for Singapore URA Sites
Urban land tender participants struggle to rapidly parse complex site requirements and land plot ratio limits set by municipal planners. Build a LlamaIndex RAG agent to digest Singapore URA land sale notices and estimate optimal bid ranges. Achieve gross development value extraction accuracy of 92%.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a LlamaIndex Advanced RAG agent with vector index routing to parse Singapore Urban Redevelopment Authority (URA) tender packages and generate bid metrics.
How work is evaluated
Evaluates accuracy of extracted Gross Floor Area (GFA) and projected land rate valuation.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
gfa_calculation_match
Checks exact match of GFA calculation (site area * plot ratio)
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
extraction_accuracy
Accuracy of extracted tender planning attributes • target: 0.92 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement a LlamaIndex vector store index (`VectorStoreIndex`, `SummaryIndex`) over URA tender release PDFs
Create custom query engine tools for querying land plot ratio (GPR), maximum gross floor area (GFA), and designated building uses
Parse land tender guidelines for special planning conditions (e.g., green mark standards, public transit connections)
Formulate baseline bid evaluation summaries with estimated land rate per square meter per plot ratio
Reference links and supporting material
Summarized land sale tender documentation released by Singapore Urban Redevelopment Authority (URA).
Challenge setup
Evaluates accuracy of extracted Gross Floor Area (GFA) and projected land rate valuation.
Challenge input
JSON containing site_area_sqm, gross_plot_ratio, land_use_zone
LlamaIndex
Data framework for structured indexing and retrieval over tender documents.
Evaluated output
JSON containing total_allowable_gfa_sqm, recommended_bid_psm, target_use
- Checks exact match of GFA calculation (site area * plot ratio)
- Accuracy of extracted tender planning attributes • target: 0.92 • range: 0-1
- Benchmark: Urban Land Tender RAG Benchmark
- Extraction Accuracy target: 0.92
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Llama Index · Required
- ZenML · Optional
- Synthflow · Optional
Evaluation contract
- gfa_calculation_match · Weight 1
- extraction_accuracy · Weight 1
Metrics
- extraction_accuracy · Target 0.92
Benchmarks
- Urban Land Tender RAG Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing site_area_sqm, gross_plot_ratio, land_use_zone
Output: JSON containing total_allowable_gfa_sqm, recommended_bid_psm, target_use
Execution harness: Python
Evaluation tools
- LlamaIndex · agent_framework
Recipe freeze marker
No recipe freeze time is recorded.