# HHAI RAG Skill File
**Hybrid Human Agentic Intelligence — Full Resource Catalog**
Version 1.1 · hhai.city · July 2026

---

## WHAT THIS FILE IS

A self-contained skill file for any AI assistant. Paste the entire contents into a system prompt, project instructions, knowledge base, or `.cursorrules` file. Once loaded, the assistant can:

- Recommend the right resource for any research or project question
- Generate working Python / JavaScript / curl code for all 6 live HK APIs
- Explain how to ingest any resource into a RAG pipeline
- Match resources to project stages (beginner → advanced)
- Suggest stacks of resources for common workflow goals

**No account, password, or installation required to use this file.**

---

## AGENT SYSTEM PROMPT (copy-paste this block)

```
You are an HHAI Research Assistant with access to the full HHAI open-data catalog.

HHAI = Hybrid Human Agentic Intelligence. This catalog covers Hong Kong urban data,
APIs, AI/RAG tools, architecture accreditation, and digital twin precedents.

When a user asks about a resource, dataset, API, or RAG workflow:
1. Identify the most relevant resource(s) by ID (e.g. C01, A02, S01)
2. Explain what it is, why it is relevant, and how to access it
3. For API resources (C01–C06): generate working Python code immediately
4. For RAG resources (A-series, B-series PDFs): explain how to chunk and ingest
5. Suggest related resources using the "Alternates" field
6. For stacks: recommend the named stack that best fits the user's goal

Always cite resource IDs. Always prefer beginner resources unless the user
specifies otherwise. When generating code, use requests (Python) or fetch (JS).
```

---

## CATALOG STRUCTURE

| Series | Category | Description |
|--------|----------|-------------|
| A01–A05 | rag_ready / context_precedent | PDFs and repos for RAG vector ingestion |
| B01–B17 | skill_ready / context_precedent | Code, notebooks, Postman, tools, precedents |
| C01–C06 | api_ready | Live HK government endpoints — call directly |
| S01–S03 | accreditation_policy | HKIA/ARB criteria and professional documents |
| Stack 1–6 | — | Recommended resource combinations for common goals |

---

## SECTION A — RAG-READY DOCUMENTS

---

**ID: A01**
**Name:** Hong Kong Smart City Blueprint 2.0
**Type:** PDF · Source: Hong Kong Government / Smart City Office
**URL:** https://www.smartcity.gov.hk/modules/custom/custom_global_js_css/assets/files/HKSmartCityBlueprint(ENG)v2.pdf
**Category:** rag_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** smart_city, hong_kong, policy, rag

**How to ingest:**
Download the PDF → split by section headings (each policy domain = one chunk) → embed into any vector store (Chroma, Pinecone, Weaviate, OpenAI). Each domain section becomes a retrievable chunk. Good for policy Q&A, domain framing, and smart-city terminology.

```python
# Quick ingest with LangChain
from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

loader = PyPDFLoader("HKSmartCityBlueprint(ENG)v2.pdf")
pages = loader.load()
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
chunks = splitter.split_documents(pages)
# → embed chunks into your vector store
```

**Use cases:** Course framing · policy reference · project domain selection · smart-city terminology
**Alternates:** A02, C05
**Note:** Start here. Foundational smart-city reference for all Hong Kong urban research.

---

**ID: A02**
**Name:** Development Bureau Consolidated Annual Open Data Plan 2026–2028
**Type:** PDF · Source: Development Bureau, Hong Kong
**URL:** https://www.devb.gov.hk/filemanager/en/content_2374/Consolidated%20Annual%20Open%20Data%20Plan%20for%202026-2028%20(EN)_DEVB.pdf
**Category:** rag_ready | **Difficulty:** intermediate | **Agent-ready:** YES
**Tags:** open_data, hong_kong, digital_twin, mapping, rag

**How to ingest:**
Download PDF → ingest as a bureau-level open-data roadmap → chunk by bureau / chapter. Good for dataset discovery queries and identifying which government bureau releases what data.

**Use cases:** Project scoping · dataset discovery · bureau mapping · urban policy interpretation
**Alternates:** A01, C05, C06
**Note:** Use alongside A01 — Blueprint gives policy vision; this gives the data delivery roadmap.

---

**ID: A03**
**Name:** Awesome Urban LLM Agents
**Type:** GitHub Repo (curated reading list)
**Source:** HKUST / usail-hkust
**URL:** https://github.com/usail-hkust/Awesome-Urban-LLM-Agents
**Category:** rag_ready | **Difficulty:** intermediate | **Agent-ready:** YES
**Tags:** urban_ai, rag, agent, smart_city

**How to ingest:**
```bash
git clone --depth 1 https://github.com/usail-hkust/Awesome-Urban-LLM-Agents
# Ingest README.md and linked paper abstracts as a curated reading-list RAG source
```

**Use cases:** AI project ideas · agent workflow reference · literature scanning · benchmark awareness
**Alternates:** B10, B11, B12
**Note:** Best as a curated index. Ingest README first; follow linked papers selectively.

---

**ID: A04**
**Name:** Urban Model Platform
**Type:** GitHub Repo · Source: City Science Lab / HCU Hamburg
**URL:** https://github.com/citysciencelab/urban-model-platform
**Category:** context_precedent | **Difficulty:** advanced | **Agent-ready:** PARTIAL
**Tags:** digital_twin, urban_ai, mapping, smart_city, architecture

**How to ingest:**
Ingest README and docs as a platform precedent for urban digital twin systems and model sharing. Good for architectural / systems framing.

**Use cases:** Platform precedent · urban model thinking · system architecture inspiration
**Alternates:** A05, B13

---

**ID: A05**
**Name:** City-Scale Digital Twin Framework
**Type:** GitHub Repo · Source: University of Iowa / uihilab
**URL:** https://github.com/uihilab/CityDigitalTwin
**Category:** context_precedent | **Difficulty:** advanced | **Agent-ready:** PARTIAL
**Tags:** digital_twin, simulation, smart_city, architecture

**How to ingest:**
Ingest repo docs and linked demo text as a precedent for city-scale digital twin architecture. Use for framing presentations and site-level reasoning.

**Use cases:** Precedent study · digital twin concept · platform comparison
**Alternates:** A04, B13

---

## SECTION B — SKILL-READY REPOS, NOTEBOOKS & TOOLS

---

**ID: B01**
**Name:** Autodesk Platform Services — GitHub Organisation
**Type:** GitHub Org (parent of B02–B05)
**URL:** https://github.com/autodesk-platform-services
**Category:** skill_ready | **Difficulty:** intermediate | **Agent-ready:** PARTIAL
**Tags:** bim, autodesk, aps, smart_city

**Use cases:** BIM API ecosystem overview · workflow research · tool discovery
**Alternates:** B02, B03, B04, B05

---

**ID: B02**
**Name:** APS BIM360 Issue Walkthrough
**Type:** GitHub Repo · Source: Autodesk Platform Services
**URL:** https://github.com/autodesk-platform-services/aps-bim360-issue-walkthrough
**Category:** skill_ready | **Difficulty:** advanced | **Agent-ready:** PARTIAL
**Tags:** bim, autodesk, aps, skill_ready

**How to ingest:** Ingest README and code structure as a hands-on BIM issue workflow example.
**Use cases:** Sample project study · BIM cloud workflow concept · issue tracking logic
**Alternates:** B03, B04, B05

---

**ID: B03**
**Name:** APS Model Properties — Postman Collection
**Type:** Postman Collection (JSON) · Source: Autodesk Platform Services
**URL:** https://github.com/autodesk-platform-services/aps-model.properties-postman.collection
**Category:** skill_ready | **Difficulty:** intermediate | **Agent-ready:** YES
**Tags:** aps, autodesk, api, postman, skill_ready

**How to use:**
Download the JSON collection → import into Postman OR parse as a structured API skill file. Each request = one agent capability. Feed the endpoint list and parameters to an agent to teach it the APS Model Properties API.

**Use cases:** API learning · request structure study · workflow prototyping
**Alternates:** B02, B04
**Note:** A Postman collection IS a skill file. This is the clearest example in the catalog.

---

**ID: B04**
**Name:** APS BIM360 Cost Management — Postman Collection
**Type:** Postman Collection (JSON) · Source: Autodesk Platform Services
**URL:** https://github.com/autodesk-platform-services/aps-bim360.costmanagement.api-postman.collection
**Category:** skill_ready | **Difficulty:** intermediate | **Agent-ready:** YES
**Tags:** bim, autodesk, aps, api, postman

**Use cases:** API study · BIM cost workflow concept · construction-management framing
**Alternates:** B03, B02

---

**ID: B05**
**Name:** Download Updated Content from Autodesk Construction Cloud — Notebook
**Type:** Jupyter Notebook · Source: Esri / GitHub
**URL:** https://github.com/Esri/cad-bim-scripts/blob/main/Samples/Download%20Updated%20Content%20from%20Autodesk%20Construction%20Cloud/DownloadUpdatedContentFromAutodeskConstructionCloud.ipynb
**Category:** skill_ready | **Difficulty:** advanced | **Agent-ready:** YES
**Tags:** bim, autodesk, notebook, workflow, skill_ready

**How to ingest:** Ingest notebook cells as a BIM/cloud integration workflow example.
**Use cases:** Notebook study · integration inspiration · workflow decomposition
**Alternates:** B01, B02, B03, B04

---

**ID: B06**
**Name:** geolab
**Type:** GitHub Repo · Source: KHwong12 / GitHub
**URL:** https://github.com/KHwong12/geolab
**Category:** skill_ready | **Difficulty:** intermediate | **Agent-ready:** PARTIAL
**Tags:** mapping, geospatial, hong_kong, shared_toolchain

**How to ingest:** Ingest repo docs and examples as a geospatial learning support resource.
**Use cases:** Mapping · basic spatial workflows · urban data exploration
**Alternates:** C02, B13, B14

---

**ID: B07**
**Name:** mcp-open-data-hk
**Type:** GitHub Repo (MCP server for DATA.GOV.HK)
**URL:** https://github.com/mcp-open-data-hk/mcp-open-data-hk
**Category:** skill_ready | **Difficulty:** advanced | **Agent-ready:** YES
**Tags:** open_data, agent, hong_kong, api, skill_ready

**How to use:**
Clone → read README for MCP tool definitions → use tool schemas as agent skill context. This is a Model Context Protocol server — plug directly into Claude Desktop or any MCP-compatible agent host for full agentic access to DATA.GOV.HK.

```bash
git clone https://github.com/mcp-open-data-hk/mcp-open-data-hk
# Follow README to configure as Claude Desktop MCP server
```

**Use cases:** Agentic open-data workflows · API abstraction · Claude Desktop integration
**Alternates:** C05, C06, C01, C02
**Note:** The most advanced agentic access pattern for HK open data in the catalog.

---

**ID: B10**
**Name:** IBM RAG Notebook (watsonx · Elasticsearch · LangChain)
**Type:** Jupyter Notebook · Source: IBM / GitHub
**URL:** https://github.com/IBM/watsonx-ai-samples/blob/master/cpd5.0/notebooks/python_sdk/deployments/foundation_models/RAG/Use%20watsonx%2C%20Elasticsearch%2C%20and%20LangChain%20to%20answer%20questions%20(RAG).ipynb
**Category:** skill_ready | **Difficulty:** advanced | **Agent-ready:** YES
**Tags:** rag, ai, notebook, skill_ready

**How to ingest:**
Download .ipynb → read cell-by-cell → extract the RAG pipeline pattern: document loading → chunking → embedding → retrieval → generation. Adapt the pattern to your own documents and LLM.

**Use cases:** RAG understanding · workflow comparison · advanced AI prototyping
**Alternates:** B11, A03

---

**ID: B11**
**Name:** RAGFlow
**Type:** GitHub Repo (open-source RAG engine) · Source: InfiniFlow
**URL:** https://github.com/infiniflow/ragflow
**Category:** skill_ready | **Difficulty:** advanced | **Agent-ready:** YES
**Tags:** rag, agent, github_repo, skill_ready

**Quickstart:**
```bash
git clone https://github.com/infiniflow/ragflow && cd ragflow
docker compose up -d
# Open http://localhost → upload HHAI_RAG_SKILL.md
# Create agent → connect knowledge base → ask questions
```

**Use cases:** RAG platform comparison · no-code RAG system · system architecture reference
**Alternates:** B10, A03, B12
**Note:** The easiest way to stand up a full RAG pipeline without writing any code.

---

**ID: B12**
**Name:** myVision-universities-RAG
**Type:** GitHub Repo · Source: Samu01Tech / GitHub
**URL:** https://github.com/Samu01Tech/myVision-universities-RAG
**Category:** context_precedent | **Difficulty:** advanced | **Agent-ready:** PARTIAL
**Tags:** rag, education, agent, context_precedent

**Use cases:** Educational RAG system precedent · RAG architecture comparison
**Alternates:** B10, B11

---

**ID: B13**
**Name:** MCity Digital Twin
**Type:** Article · Source: University of Michigan / Mcity
**URL:** https://mcity.umich.edu/mcity-unveils-digital-twin-making-its-physical-av-testing-facility-available-for-free-in-the-virtual-world/
**Category:** context_precedent | **Difficulty:** beginner | **Agent-ready:** PARTIAL
**Tags:** digital_twin, mobility, context_precedent

**Use cases:** Digital twin case study · platform narrative · smart mobility angle
**Alternates:** A04, A05

---

**ID: B14**
**Name:** VoxCity
**Type:** GitHub Repo · Source: kunifujiwara / GitHub
**URL:** https://github.com/kunifujiwara/VoxCity
**Category:** context_precedent | **Difficulty:** advanced | **Agent-ready:** PARTIAL
**Tags:** digital_twin, mapping, geospatial, architecture

**How to ingest:** Ingest README as a 3D city modeling tool precedent.
**Use cases:** 3D city studies · urban form visualization · mapping precedent
**Alternates:** B06, A04, A05

---

**ID: B15**
**Name:** APA Technology Division — Urban and Regional Planning Resources
**Type:** GitHub Repo · Source: APA Technology Division
**URL:** https://github.com/APA-Technology-Division/urban-and-regional-planning-resources
**Category:** context_precedent | **Difficulty:** intermediate | **Agent-ready:** PARTIAL
**Tags:** planning, smart_city, context_precedent, github_repo

**Use cases:** Topic discovery · planning tech reference · tool comparison
**Alternates:** A03, C06

---

**ID: B16**
**Name:** Open Data for Smart Sustainable City: The Case of Hong Kong
**Type:** Article · Source: HKUST Library
**URL:** https://library.hkust.edu.hk/news-events/news/open-data-smart-sustainable-city-case-hong-kong
**Category:** context_precedent | **Difficulty:** beginner | **Agent-ready:** PARTIAL
**Tags:** hong_kong, open_data, smart_city, article

**Use cases:** Policy context · open-data reflection · Hong Kong framing
**Alternates:** A01, C06

---

**ID: B17**
**Name:** LLM-based RAG Implementation Reference
**Type:** Academic PDF · Source: ICSD / University of the Aegean
**URL:** https://www.icsd.aegean.gr/publication_files/0652d01ecf00a3c6139111da8db24d91.pdf
**Category:** context_precedent | **Difficulty:** advanced | **Agent-ready:** YES
**Tags:** rag, ai, pdf, context_precedent

**How to ingest:** Download PDF → ingest whole document as advanced RAG theory reference.
**Use cases:** RAG concept grounding · advanced reading · pipeline design reference
**Alternates:** B10, B11, A03

---

## SECTION C — LIVE HK OPEN DATA APIs

> All C-series resources are live Hong Kong government endpoints.
> No registration required. No API key needed. Call directly from Python, JS, curl, or any agent tool.

---

**ID: C01**
**Name:** Buildings Department — Building Information
**Type:** Dataset / JSON endpoint
**Source:** Buildings Department / DATA.GOV.HK
**URL:** https://data.gov.hk/en-data/dataset/hk-bd-opendata-building-information
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** open_data, hong_kong, bim, mapping, architecture

**Live endpoint:**
```python
import requests

url = "https://www.bd.gov.hk/en/resources/others/opendata/BUILDING_DETAILS.json"
r = requests.get(url)
data = r.json()

# Returns per-building records:
# building_id, street_address, district, year_of_completion, GFA, usage_class
for building in data[:5]:
    print(building)
```

```javascript
// JavaScript / fetch
const res = await fetch("https://www.bd.gov.hk/en/resources/others/opendata/BUILDING_DETAILS.json");
const data = await res.json();
console.log(data.slice(0, 5));
```

```bash
# curl
curl "https://www.bd.gov.hk/en/resources/others/opendata/BUILDING_DETAILS.json" | python3 -m json.tool | head -60
```

**Returns:** building ID · street address · district · year of completion · GFA · usage class
**Use cases:** Building analysis · urban mapping · site research · building age queries · BIM context
**Alternates:** C02, C06

---

**ID: C02**
**Name:** Lands Department — Location Search API
**Type:** Live REST API
**Source:** Lands Department / DATA.GOV.HK
**URL:** https://data.gov.hk/en-data/dataset/hk-landsd-openmap-development-location-search-api
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** api, mapping, geospatial, hong_kong

**Live endpoint:**
```python
import requests

query = "Causeway Bay"
url = f"https://geodata.gov.hk/gs/api/v1.0.0/locationSearch?q={query}"
r = requests.get(url)
results = r.json()

# Returns: place name, x/y coordinates (HK1980 grid), address string
for result in results.get("SpatialBase", [])[:3]:
    print(result["Address"], result["X"], result["Y"])
```

```javascript
const query = "Causeway Bay";
const res = await fetch(`https://geodata.gov.hk/gs/api/v1.0.0/locationSearch?q=${encodeURIComponent(query)}`);
const data = await res.json();
console.log(data.SpatialBase?.slice(0, 3));
```

```bash
curl "https://geodata.gov.hk/gs/api/v1.0.0/locationSearch?q=Causeway+Bay"
```

**Returns:** Place name · X/Y coordinates (HK1980 projection) · address string
**Use cases:** Location search · site lookup · map-based projects · service geography
**Alternates:** B06, B14

---

**ID: C03**
**Name:** Transport Department — Real-time Parking Vacancy
**Type:** Live REST API + CSV
**Source:** Transport Department / DATA.GOV.HK
**URL:** https://data.gov.hk/en-data/dataset/hk-td-tis_5-real-time-parking-vacancy-data
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** api, smart_city, mobility, hong_kong

**Live endpoint:**
```python
import requests

# Step 1: get the car park list (CSV)
csv_url = "https://static.data.gov.hk/td/parking-vacancy/dataspec/car_park_list.csv"
carparks_csv = requests.get(csv_url).text
# Parse with csv module or pandas to get car park IDs

# Step 2: get live vacancy for a specific car park
park_id = "CP001"  # replace with a real ID from Step 1
vacancy_url = f"https://api.data.gov.hk/v1/carpark-info-vacancy?carparkIds={park_id}&lang=en"
r = requests.get(vacancy_url)
data = r.json()

# Returns: car park name, vacancy count, total spaces, last updated timestamp
print(data)
```

**Returns:** Car park name · vacancy count · total spaces · last updated timestamp
**Use cases:** Dashboard prototype · API practice · mobility project · smart city service design
**Alternates:** C04, C02

---

**ID: C04**
**Name:** EPD — Current AQHI (Air Quality Health Index)
**Type:** Live REST API
**Source:** Environmental Protection Department / DATA.GOV.HK
**URL:** https://data.gov.hk/en-data/dataset/hk-epd-airteam-current-aqhi-of-individual-air-quality-monitoring-stations
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** api_ready, environment, hong_kong, smart_city, ecology

**Live endpoint:**
```python
import requests

url = "https://api.data.gov.hk/v1/historical-air-quality?lang=en"
r = requests.get(url)
data = r.json()

# Returns per-station readings:
# station name, AQHI (1–10+), health risk category, timestamp
for station in data.get("results", [])[:5]:
    print(station["StationName"], station["AQHI"], station["HealthRisk"])

# Stations include:
# Central, Causeway Bay, Mong Kok, Tuen Mun, Tung Chung,
# Kwun Tong, Sham Shui Po, Kwai Chung, Sha Tin, Tai Po, Yuen Long
```

```javascript
const res = await fetch("https://api.data.gov.hk/v1/historical-air-quality?lang=en");
const data = await res.json();
console.log(data.results?.slice(0, 5));
```

```bash
curl "https://api.data.gov.hk/v1/historical-air-quality?lang=en" | python3 -m json.tool
```

**Returns:** Station name · AQHI value (1–10+) · health risk category · timestamp
**Use cases:** AQ dashboard · environmental mapping · sensor data concepts · ecology data
**Alternates:** C03, C02

---

**ID: C05**
**Name:** DATA.GOV.HK API Specification
**Type:** Guide / Documentation
**Source:** DATA.GOV.HK
**URL:** https://data.gov.hk/en/help/api-spec
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** YES
**Tags:** open_data, api, hong_kong, guide

**General filter API pattern (works for any DATA.GOV.HK dataset):**
```bash
# Replace RESOURCE_ID with any dataset ID from data.gov.hk
curl 'https://api.data.gov.hk/v2/filter?q={"resource":"RESOURCE_ID","section":1,"format":"json"}'
```

```python
import requests, json

resource_id = "hk-bd-opendata-building-information"  # example
url = "https://api.data.gov.hk/v2/filter"
params = {"q": json.dumps({"resource": resource_id, "section": 1, "format": "json"})}
r = requests.get(url, params=params)
print(r.json())
```

**Use cases:** API onboarding · developer reference · data portal orientation · finding any HK dataset
**Alternates:** B07

---

**ID: C06**
**Name:** DATA.GOV.HK Open Data Portal
**Type:** Portal (800+ datasets)
**Source:** Hong Kong Government
**URL:** https://data.gov.hk/en/
**Category:** api_ready | **Difficulty:** beginner | **Agent-ready:** PARTIAL
**Tags:** portal, open_data, hong_kong

**How to use:**
Browse by theme → find dataset → each dataset page has an "API" tab with the endpoint URL → call with requests or curl using the C05 filter pattern.

Key providers:
- Buildings Department: https://data.gov.hk/en-datasets/provider/hk-bd
- EPD: https://data.gov.hk/en-datasets/provider/hk-epd
- Transport Dept: https://data.gov.hk/en-datasets/provider/hk-td
- Lands Dept: https://data.gov.hk/en-datasets/provider/hk-landsd

**Use cases:** Finding datasets · comparing providers · topic exploration · 800+ datasets discoverable
**Alternates:** A02, C05

---

## SECTION S — ACCREDITATION & POLICY

---

**ID: S01**
**Name:** HKIA/ARB Accreditation of Architectural Programmes — Criteria and Procedures (Jan 2026)
**Type:** PDF
**Source:** HKIA / ARB (Architects Registration Board)
**URL:** https://www.hkia.net/upload/finder/file/BEdA/HKIA-ARB%20Accreditation%20of%20Architectural%20Programmes%20Criteria%20and%20Procedures%20%282026%20January%29.pdf?v10.1106620472
**Category:** accreditation_policy | **Difficulty:** intermediate | **Agent-ready:** YES
**Tags:** hkia, accreditation, architecture, policy, rag

**How to ingest:**
Download PDF → split by criteria section:
- Design Ability
- Technical Knowledge
- Communication
- Professional Context
- Research & Inquiry

Each section becomes one chunk. Embed for studio outcome mapping and review preparation.

```python
from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

loader = PyPDFLoader("HKIA_ARB_Criteria_2026.pdf")
pages = loader.load()
# Split at ~1200 tokens — criteria sections are medium-length
splitter = RecursiveCharacterTextSplitter(chunk_size=1200, chunk_overlap=200)
chunks = splitter.split_documents(pages)
```

**Use cases:** Criteria awareness · studio outcome mapping · learning outcome alignment · review preparation
**Alternates:** S02, S03

---

**ID: S02**
**Name:** HKIA Professional Assessment Eligibility
**Type:** Webpage · Source: HKIA
**URL:** https://www.hkia.net/en/professional-assessment/eligibility
**Category:** accreditation_policy | **Difficulty:** beginner | **Agent-ready:** PARTIAL
**Tags:** hkia, accreditation, architecture, webpage

**Use cases:** Professional context · accreditation awareness · eligibility reference
**Alternates:** S01

---

**ID: S03**
**Name:** ARB and HKIA Memorandum of Understanding
**Type:** Webpage · Source: ARB (UK)
**URL:** https://arb.org.uk/international-routes/arb-and-hkia-mou/
**Category:** accreditation_policy | **Difficulty:** beginner | **Agent-ready:** PARTIAL
**Tags:** hkia, arb, architecture, accreditation

**Use cases:** Professional context · international recognition · accreditation landscape
**Alternates:** S01, S02

---

## RESOURCE STACKS — RECOMMENDED COMBINATIONS

---

### Stack 1 — HK Smart City Open Data (Beginner)
**Best for:** First project, API literacy, dashboard prototype, URSP3600 framing
**Resources:** A01 · A02 · C05 · C06 · C01 · C02 · C03 · C04
**Alternates:** B07, B16
**Agent use:** Ingest A01 + A02 PDFs for RAG policy Q&A; use C01–C04 as live tool-call APIs
**Strengths:** All authoritative HK government sources · no registration · beginner-friendly · directly downloadable
**Limitations:** Limited GBA cross-border data · some APIs have limited filtering

---

### Stack 2 — BIM / Autodesk / Cloud Workflow (Intermediate)
**Best for:** BIM concepts, cloud API patterns, construction workflow, Postman skill files
**Resources:** B01 · B02 · B03 · B04 · B05
**Alternates:** C01
**Agent use:** Feed B03 / B04 Postman collections as structured skill files; ingest B05 notebook as workflow example
**Strengths:** Real production-grade API patterns · Postman collections as agent skill artifacts
**Limitations:** Requires Autodesk account for live API · not beginner-ready without framing

---

### Stack 3 — Urban AI / RAG / Agent (Advanced)
**Best for:** Building a RAG pipeline, agent system design, LLM urban research
**Resources:** A03 · B10 · B11 · B17
**Alternates:** B12, B07
**Agent use:** Ingest A03 README as a meta-reading-list RAG source; use B17 as domain theory; B11 to run a live RAG system
**Strengths:** Multiple RAG implementation patterns · open-source tools · all downloadable

---

### Stack 4 — Digital Twin / Systems (Advanced)
**Best for:** Platform precedent, city-scale thinking, territorial system architecture
**Resources:** A04 · A05 · B13 · B14
**Alternates:** B06
**Agent use:** Ingest repo READMEs as precedent context; use B14 for 3D city visualization reference

---

### Stack 5 — Architecture Studio / HKIA (Studio)
**Best for:** Studio alignment, accreditation mapping, review preparation, Archipelago Studio
**Resources:** S01 · S02 · S03
**Agent use:** Ingest S01 PDF by criteria section → query for outcome mapping → cross-reference student work with criteria headings

---

### Stack 6 — Mapping / Geospatial / Site (All levels)
**Best for:** Site lookup, territorial mapping, geospatial analysis
**Resources:** C02 · B06 · B14
**Alternates:** A05
**Agent use:** Use C02 API as a live location lookup tool; ingest B06 examples as spatial workflow reference

---

## HOW TO BUILD YOUR OWN RAG

### Route A — Paste into Claude · 2 minutes · no code

1. Download this file → open in any text editor
2. Open Claude → New Project → Project Instructions → paste everything
3. Or upload the `.md` file directly (Claude supports file uploads)
4. Ask: *"Which resource gives me real-time air quality data for Hong Kong?"*

Works with: **Claude Projects · OpenAI Custom GPTs · Gemini Gems · any system-prompt field**

---

### Route B — Vector RAG with Python · LangChain + Chroma

```python
from langchain.document_loaders import TextLoader
from langchain.text_splitter import MarkdownTextSplitter
from langchain.vectorstores import Chroma
from langchain.embeddings import OpenAIEmbeddings
from langchain.chains import RetrievalQA
from langchain.chat_models import ChatOpenAI

# 1. Load this skill file
docs = TextLoader("HHAI_RAG_SKILL.md").load()

# 2. Split by Markdown sections (--- delimiters work well)
splitter = MarkdownTextSplitter(chunk_size=800, chunk_overlap=100)
chunks = splitter.split_documents(docs)

# 3. Embed and store locally
vectorstore = Chroma.from_documents(chunks, OpenAIEmbeddings())
retriever = vectorstore.as_retriever(search_kwargs={"k": 4})

# 4. Create a RAG chain
llm = ChatOpenAI(model="gpt-4o-mini")
chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)

# 5. Query
result = chain.run("How do I get live parking vacancy data for Hong Kong?")
print(result)
```

---

### Route C — RAGFlow · no code · Docker

```bash
# 1. Clone and start RAGFlow
git clone https://github.com/infiniflow/ragflow && cd ragflow
docker compose up -d

# 2. Open http://localhost in your browser
# 3. Create a knowledge base → upload HHAI_RAG_SKILL.md
# 4. Create an agent → connect it to the knowledge base
# 5. Ask questions in natural language
```

Works with: any documents you add alongside this file (A01, A02, S01 PDFs, etc.)

---

### Route D — Cursor / Copilot / Editor Rules

Save as `.cursorrules` in your project root (or use `hhai.cursorrules` from the Skills & APIs page):

```
You are an HHAI-aware coding assistant.

When the user asks about Hong Kong urban data, open data APIs, or RAG workflows:
- Consult the HHAI resource catalog in HHAI_RAG_SKILL.md
- For HK data: first check if a C-series API (C01–C06) covers it — no auth needed
- For RAG pipelines: recommend B10 (IBM notebook) or B11 (RAGFlow) as starting points
- For policy/planning context: recommend A01 (Blueprint) or A02 (DEVB plan)
- For accreditation: S01 (HKIA/ARB criteria)

Always generate working Python code for API calls.
Always cite resource IDs (e.g. C01, A02) in your answers.
```

---

## INGESTION CHEAT SHEET

| ID | Resource | Format | Ingest method | Chunk by |
|----|----------|--------|---------------|----------|
| A01 | HK Smart City Blueprint | PDF | PyPDFLoader | Section headings |
| A02 | DEVB Open Data Plan | PDF | PyPDFLoader | Bureau / chapter |
| A03 | Awesome Urban LLM Agents | GitHub README | TextLoader | H2 headings |
| A04 | Urban Model Platform | GitHub README | TextLoader | Section |
| B03 | APS Model Properties Postman | JSON collection | JSONLoader | Per request |
| B05 | ACC Download Notebook | .ipynb | NotebookLoader | Cell |
| B10 | IBM RAG Notebook | .ipynb | NotebookLoader | Cell |
| B11 | RAGFlow | GitHub README | TextLoader | Feature section |
| B17 | RAG Reference PDF | PDF | PyPDFLoader | Page |
| S01 | HKIA Criteria (2026) | PDF | PyPDFLoader | Criteria section |
| This file | HHAI Skill File | .md | MarkdownTextSplitter | --- delimiters |

---

## ABOUT HHAI

**HHAI** (Hybrid Human Agentic Intelligence) is an open research platform where human expertise and AI agents work together. Human = director. Agent = executor. This catalog = the shared knowledge substrate.

- Platform: **https://www.hhai.city**
- Skills & APIs: **https://www.hhai.city** → Skills & APIs
- Live AI Sandbox (6 agents on Grok-4.5): **https://www.hhai.city** → Sandbox
- Full RAG Catalog: **https://www.hhai.city** → API / Skill / RAG

*This file is open. Adapt, extend, and share freely. HHAI · 2026*