nardial 0.1.8

Nardial Dialog System

NarDialPy

NarDialPy is a Python framework for building and running narrative dialogs.

It combines:

• Dialog blocks written as JSON (functional, narrative, and chitchat dialogs)
• A session manager that executes a chosen dialog agenda
• Voice / NLU / LLM services (Dialogflow, Google TTS, OpenAI GPT via social-interaction-cloud)

The repository includes a complete demo conversation in the examples/ folder that shows how these parts work together.

Repository layout

• src/nardial/ – core runtime (dialog loading, move execution, session flow)
• examples/demo_general_conversation.py – runnable demo entrypoint
• examples/dialogs.json – demo dialogs (conversation content)
• web/authoring/ – browser-based dialog authoring UI
• tests/ – unit tests for move handling and dialog behavior

How the demo conversation works

The demo script (examples/demo_general_conversation.py) creates a ConversationAgent, then runs a fixed agenda through SessionManager:

session_agenda = ["greeting", "hero_can_dream_1", "dream12", "goodbye"]

Those dialog IDs are resolved from examples/dialogs.json, which contains declarative move sequences such as:

• say
• ask_open
• ask_yesno
• ask_options
• branch (outcome-based branching)

This makes it easy to edit conversation behavior without changing Python code.

Quick start (demo)

1) Install

From the repository root:

pip install -e .

Install/update cloud integrations used by the demo:

pip install --upgrade social-interaction-cloud[dialogflow,google-tts,openai-gpt]

2) Configure credentials

1. Create Dialogflow credentials and save to: conf/dialogflow/google_keyfile.json
   • Setup guide: https://socialrobotics.atlassian.net/wiki/spaces/CBSR/pages/2205155343/Getting+a+google+dialogflow+key
2. Create an OpenAI API key and save to: conf/openai/.openai_env
   • API keys: https://platform.openai.com/api-keys

Example .openai_env:

OPENAI_API_KEY="your key"

Never commit or share credential files.

3) Configure the Dialogflow agent

In your Dialogflow agent:

• Remove default intents.
• Go to Settings → Import and Export.
• Import resources/droomrobot_dialogflow_agent.zip to load the intents/entities used by the demo.

4) Start required services

In separate terminals:

conf/redis/redis-server.exe conf/redis/redis.conf
run-dialogflow
run-google-tts
run-gpt

5) Run the demo conversation

From examples/:

python demo_general_conversation.py

Use a supported device configuration in the script (desktop by default).

Customizing conversations

• Edit examples/dialogs.json to add or modify dialog blocks.
• Update the session_agenda in examples/demo_general_conversation.py to change the order/content of a run.
• Use web/authoring/ if you prefer editing dialogs through the browser UI.

Development

Run tests from the repository root:

python -m pytest -q

---