Children's Services Knowledge Base

An open-source, Git-native 'map of the world' or knowledge base for connected people, projects, and organisations in the children’s services sector, aligned to the Smart City Concept Model/framework (SCCM).

---

Notes / Dev story

• How can/should we structure the (meta) data about every element within the map (csv, db, flat file, .yml...)
• Can|should we store the names of people around the various initiatives/projects (is that relevant/useful and would they want that)
• How do we store the data for retrieval, as we scale up with larger volumes what impact will this have (esp on load/search times)

Structure

/knowledge_base
├── app/
│   ├── Home.py
│   └── pages/
│       └── 1_relationships.py
│       └── 2_network_view.py
│       └── 3_map_elements.py
├── data/
│   ├── index_data.parquet  ← cached index lives here
│   └── organizations/...   ← sccm framework folder struct with flat files
│   └── relationships/...   
├── scripts/
│   ├── app.py                  # Streamlit UI
│   ├── build_runtime_index.py  # Build DuckDB index from YAML
│   ├── validate_schema.py      # Validate YAMLs against schema
│   └── nlp_to_sql.py           # Stub for natural-language to SQL
└── README.md

---

Dev Notes

Dependencies

Python 3.9+ and the following packages:

pip install streamlit duckdb pyyaml cerberus

Run Streamlit App

streamlit run app/Home.py

Will launch local UI in browser for exploring|searching people, projects, and orgs.

---

Validate YAML Structure

Use the built-in schema validation script:

python scripts/validate_schema.py

This checks for required fields and types based on a lightweight SCCM-aligned schema.

Rebuild Search Index

DuckDB is used as a fast in-memory index:

python scripts/build_runtime_index.py

You’ll see all records loaded and printed to terminal. If you’re running streamlit run app/Home.py you don't need to do this as build_duckdb_index(records) is automatically called — no need to run it separately.

Text-to-SQL (Prototype)

You can simulate natural language queries:

python scripts/nlp_to_sql.py

Example input:

Who worked on safeguarding in the North West?

---

Future Plans

• SCCM-compliant JSON-LD export
• GitHub Pages + search index
• Relationship mapping (RELATIONSHIP, EVENT)
• Collaborative edit + merge workflow

---

© D2I