google / google/adk-python-community

Integration Proposal: SDC Agents — Deterministic Semantic Data Artifacts from Enterprise Datastores

Đang mở
#97 0 bình luận 0 reaction 0 người được giao Xem trên GitHub
Ngôn ngữ chính
Python
Star
182
Fork
75
Chỉ số merge pull request
Không có pull request nào được merge trong 30 ngày

Mô tả

## 🔴 Required Information

### Is your feature request related to a specific problem?

ADK developers working with enterprise data sources (SQL databases, CSV, JSON, MongoDB, BigQuery) lack tooling to produce deterministic, W3C-compliant semantic data artifacts. Current integrations focus on retrieval and generation — none produce validated schemas (XSD, SHACL, JSON-LD) or signed XML instances from structured data.

### Describe the Solution You'd Like

We maintain [SDC Agents](https://github.com/SemanticDataCharter/SDC_Agents) (`sdc-agents` on PyPI), a suite of **9 purpose-scoped ADK agents with 32 tools** that transform enterprise data into validated, multi-format semantic artifacts.

**What it does:**

- **Introspect** legacy datastores (SQL, CSV, JSON, MongoDB, BigQuery — read-only) and extract structure
- **Discover** published schemas from a catalog of 6,400+ components (FHIR, NIEM, NIH CDEs, X12, SUS, CIHI)
- **Map** source columns to semantic components with ontology links
- **Generate** XML instances from mapped data
- **Validate and sign** instances via a validation-as-a-service API
- **Distribute** artifact packages to triplestores (Fuseki, Neo4j, GraphDB), REST APIs, and filesystems
- **Assemble** new data models from component libraries

**Architecture:**
Each agent is an `LlmAgent` with a single `BaseToolset`. No agent has both datasource access and network access (security isolation by design). All tools are async, audited (JSONL), and cache-aware.

**Usage:**
```python
from sdc_agents.agents import create_introspection_agent, create_catalog_agent

# Introspect a PostgreSQL database
introspect = create_introspection_agent()

# Search the published catalog of 6,400+ semantic components
catalog = create_catalog_agent()
```

**Already supports:**
- ADK `BaseToolset` / `FunctionTool` / `LlmAgent` patterns
- MCP export via `adk_to_mcp_tool_type()`
- Docker image, CLI, PyPI package (`pip install sdc-agents`)
- 184 tests, 82% coverage

**We propose contributing:**
1. A thin wrapper module under `contributing/samples/` with usage examples
2. A documentation page for the `adk-docs` integrations directory

### Impact on your work

Enables ADK agents to produce deterministic, standards-compliant data schemas and validated instances from enterprise data sources — closing the gap between agentic AI and formal data governance. Targets healthcare, government, financial, and research domains where data provenance and schema validation are mandatory.

### Willingness to contribute

Yes — we have the integration ready. Corporate CLA for Axius SDC, Inc. is signed (2026-03-11).

---

## 🟡 Recommended Information

### Additional Context

- **PyPI**: https://pypi.org/project/sdc-agents/
- **GitHub**: https://github.com/SemanticDataCharter/SDC_Agents
- **User Docs**: https://github.com/SemanticDataCharter/SDC_Agents/tree/main/docs/user
- **Standards**: W3C (XSD 1.1, RDF, OWL 2, SHACL, SPARQL), ISO 21090, ISO/IEC 21838-2 (BFO 2020)
- **Production platform**: SDCStudio (deployed on Google Cloud Run)

Hướng dẫn đóng góp

Mở hướng dẫn đóng góp

Đánh giá

Issue này chưa được đánh giá.

Nhận issue mới trong hộp thư của bạn

Bản tóm tắt ngắn những issue GitHub phù hợp với người mới.