Agentic Knowledge Engineering

Open-code skill and UI project for building and orchestrating Claude skills.

• UI: Zero-dependency HTML/JS served via Python stdlib
• Server: Python stdlib HTTP server (no npm, no pip)

Getting Started

Option A: Q&A Agent

cd features/qna-agent/
cp config/ake.example.json config/ake.json    # edit with LLM credentials
python3 build-index.py                         # initial knowledge index
python3 serve.py                               # http://127.0.0.1:3100/ui_lite/
# See features/qna-agent/ARCHITECTURE.md for deployment to other platforms

Option B: Chat UI

cp ake.example.json ake.json                  # edit with LLM credentials (in project root)
python3 features/init-chat/serve.py            # http://127.0.0.1:3000/ui_lite/

Project Structure

├── ake.example.json   # LLM API config template (copy to ake.json)
├── ake.schema.json    # Config JSON Schema
├── features/          # feature modules (each with ui_lite, skill, specs)
│   ├── init-chat/     # chatbot for project initialization
│   │   ├── serve.py   #   zero-dependency dev server
│   │   ├── ui_lite/   #   zero-dependency chat UI
│   │   ├── skill/     #   project-initializer skill
│   │   └── specs/     #   feature specifications (001-004)
│   └── qna-agent/     # offline knowledge Q&A agent
│       ├── agent/     #   portable agent core (skills + knowledge + memory)
│       ├── ui_lite/   #   zero-dependency browser UI
│       ├── serve.py   #   local dev server
│       └── ARCHITECTURE.md  # design doc + deployment guide
├── specs/             # SDD workflow + template
│   ├── SDD.md
│   └── _template/
├── MEMORY.md          # accumulated project knowledge
└── AGENTS.md          # AI workflow conventions

Configuration

Copy the template and fill in your LLM API details:

{
  "llm": {
    "provider": "openai-compatible",
    "apiKey": "sk-...",
    "baseURL": "https://api.openai.com/v1",
    "model": "gpt-4o",
    "options": {
      "temperature": 0.7,
      "maxTokens": 4096,
      "stream": true
    }
  },
  "skills": {
    "paths": ["features/init-chat/skill/"],
    "autoload": ["project-initializer"]
  }
}

• provider: "openai-compatible" or "anthropic"
• skills.paths: one or more skill directories. For qna-agent use "features/qna-agent/agent/skill/"
• skills.autoload: skill names loaded into context on startup. qna-agent uses ["qna-orchestrator", "knowledge-retrieval", "memory-curation"]
• ake.json is gitignored — never commit your API keys

Each feature has its own config/ake.example.json template. See features/qna-agent/ARCHITECTURE.md for the full deployment guide.

Workflow

All features follow spec-driven development. See specs/SDD.md for the full workflow.