Knowledge-Centric AI is an open research hub bringing together scientific publications, executable specifications, reference implementations, engineering artifacts and field guides exploring architectures where knowledge, evidence and runtime capabilities become first-class building blocks of intelligent systems.
The repository integrates research and engineering across Knowledge Architecture, Evidence-Centric AI, Retrieval-Augmented Generation (RAG), Agent Runtime Architecture, and Institutional Capability Engineering.
Knowledge-Centric AI investigates how organizational knowledge evolves from static information into governed, executable, and continuously evolving capabilities.
The long-term objective is to establish architectural foundations where knowledge is not merely retrieved, but preserved, executed, qualified, and institutionally evolved across changing people, AI agents, models, workflows, and software systems.
Peer-reviewed and preprint research introducing new architectural concepts, reference models, and engineering foundations for Knowledge-Centric AI.
| Paper | Focus | Contribution |
|---|---|---|
| 📄 Knowledge-Centric Information Systems | Knowledge Architecture | Introduces Knowledge Architecture and Knowledge-Centric Information Systems as an architectural discipline for operational knowledge. |
| 📄 Governed Evolution of Agent Runtimes through Executable Operational Cognition | Runtime Governance | Proposes a governance model for the controlled evolution of agent runtimes through Executable Operational Cognition. |
| 📄 From Task-Guided Conversational Graphs to Goal-Oriented Dialogue Runtimes | Goal-Oriented Agent Runtimes | Introduces the Goal-Oriented Dialogue Runtime (GODR) design pattern for long-lived conversational objectives. |
| 📄 Institutional Capability Lineages (ICLA) | Institutional Capability Engineering | Proposes a registry-centered reference architecture that transforms distributed organizational knowledge into governed institutional capabilities through canonical contracts, contextual assemblies, evidence-governed evolution, and persistent capability lineages. |
Applied perspectives connecting research contributions with engineering practice and emerging AI architectures.
| Resource | Description |
|---|---|
| 📄 From Uncertainty to Confidence | An evidence-centric interpretation of modern Retrieval-Augmented Generation architectures. |
Engineering handbooks distilling research and industrial experience into practical guidance for production AI systems.
| Resource | Description |
|---|---|
| 📘 RAG Field Guide | Practical handbook for designing, evaluating and operating Retrieval-Augmented Generation systems. |
The repository currently explores five complementary research areas:
- Knowledge Architecture
- Evidence-Centric AI
- Retrieval-Augmented Generation (RAG)
- Agent Runtime Architecture
- Institutional Capability Engineering
Together, these areas investigate how knowledge evolves from static information into executable capabilities that can be governed, retrieved, validated, executed, and continuously improved across intelligent systems.
Rather than treating knowledge as passive information, Knowledge-Centric AI explores architectures where knowledge becomes an operational asset for humans, AI models, agents, workflows, and software systems.
The common objective across these resources is to understand how intelligent systems transform uncertainty into trustworthy decisions through structured knowledge and governed evidence.
Trust is not assumed. Confidence is constructed from governed evidence.
If you use these resources in academic or industrial work, please cite the corresponding publication or artifact.
Repository-wide citation metadata is provided through CITATION.cff.
Each publication also provides its own landing page, companion artifacts, and citation metadata where applicable.
Unless otherwise specified:
- 📚 Scientific papers, documentation, figures, and research artifacts are licensed under CC BY-NC-SA 4.0.
- 💻 Source code, schemas, reference implementations, and software components are licensed under the MIT License.
See the corresponding LICENSE files for details.