Concepts
In the Predictable Dialogs app, you’ll see two main pages:
- Agents (the default page when the app opens)
- AI Resources (can be opened from the menu)
Agents
An Agent is the entity you configure on the Agents page. It uses the configured AI models to perform tasks.
We currently provide Chatbot Agents. Chatbot Agents can be added to websites or connected to WhatsApp. You can give each Chatbot Agent its own theme and setting.
AI Resources
An AI Resource is listed on the AI Resources page, and it maps to an AI Model (e.g., a model from OpenAI, Anthropic, Deepseek, xAI, etc.). Every AI Model needs:
- A unique name within the app.
- An API key to the external provider.
- Potentially an ID, depending on the provider (e.g., an OpenAI Assistant might require an ID).
You can have multiple AI Resources that point to the same AI Model, as long as each has a unique name within the app.
For example:
| AI Resource | Maps To |
|---|---|
| AI Resource 1 | OpenAI Assistant with ID “xyz” |
| AI Resource 2 | OpenAI Assistant with ID “xyz” |
Here, both AI Resource 1 and AI Resource 2 connect to the exact same AI model (the OpenAI Assistant with ID “xyz”), but they have different names in the app.
Relationship between AI Resources & Agents
Currently, each Agent can only be connected to one AI Resource at a time, although the underlying design supports multiple connections.
- If you change the AI Model an AI Resource points to, all Agents connected to that AI Resource automatically start using the new AI model.
- You can update an Agent to a different AI Resource at any time.
This setup gives you a lot of flexibility to switch AI models on the fly.
Naming Edge Case
If you have an AI Resource named X in the AI Resources page, and then on the Agents page you add a new OpenAI assistant (with an assistantId and apiKey) also named X, it will replace the existing AI Resource named X.