tsan.ai has won the bid for a high-quality dataset project with the Cultural Relics Press. Built on the press's six decades of archaeological publishing, it constructs a topic dataset and RAG knowledge base covering the full span of early Chinese civilization.
Background
Data in archaeology is inherently multi-sourced, heterogeneous and inconsistently standardized — and traditional custom development can't keep pace with a knowledge system in constant flux. With six decades of archaeological publishing behind it, the Cultural Relics Press needed a data foundation that could keep expanding as research went deeper, not a one-off delivered system.
Key Requirements
- Archaeological material is multi-sourced and heterogeneous, with no common standard
- The knowledge system keeps evolving while custom development can't keep up
- A data foundation that keeps expanding, rather than a one-off system
Solution
The answer is tsanCenter, the ontology system. Unlike form- and front-end-driven low-code or AI development tools, it takes a “data back-end to front-end” path — solidifying the data-management and business-logic layer first so the model can read the data and understand the business, letting front-end applications grow naturally from there. The platform has accumulated more than 1,500 enterprise-grade application modules, and intelligent solutions built on it have taken on multiple large AI transformation projects, over half of which have already been delivered.
Three Core Capabilities
Intelligent Data Management
An LLM understands the relationships within business data, automatically recommends and optimizes data models, and quickly forms a complete data-management back end.
Hybrid-Model Business Logic Orchestration
A rule engine combined with machine learning turns complex business scenarios into visual, executable, definable workflows.
Controllable Intelligent Development
On that data foundation, development agents build applications, optimize code and run system operations around the clock.
Implementation Path
The project moves through four stages — standards, data, knowledge, services — decoupling standard-setting, data processing, knowledge construction and service output so that industry experts do professional work through configuration alone.
Unified Standards
A thesaurus unifies the data standard across multi-sourced, heterogeneous material.
Data Processing
The pipeline (collection, cleaning, tagging, enrichment, rights clearance, QA) weaves AI processing into the workflow, with human and agent tasks mixed together — agents can even decide the processing direction for a given object on their own.
Knowledge Construction
A topic dataset and RAG knowledge base covering the full span of early Chinese civilization; data accumulates year over year and expert know-how keeps turning into model capability.
Knowledge Services
The built-in agent-management module (tsanClaw) lets business staff configure search, Q&A and writing services themselves, with no engineering hand-off.
Capability Expansion
An open, node-based interface lets new data sources, tagging strategies and QA rules register through one standard API, so the platform grows in step with the data it accumulates.
Value
For the press, the project marks a shift from traditional publishing services to intelligent knowledge services — and the method carries over to other knowledge-dense industries.
What Changes
- Archaeological material once scattered across paper reports and scans becomes a structured asset organized to one standard
- Cross-topic, cross-era knowledge retrieval is now at hand
- Knowledge-base-verified Q&A and report drafts that follow archaeological conventions sharply raise research efficiency