continuedev/rasa-rime-tutorial icon
public
Published on 8/7/2025
Rasa x Rime Tutorial

A custom agent that helps you get started with Rasa and Rime

Rules
Models
Context
200kinput·64koutput
openai GPT-5 model icon

GPT-5

OpenAI

400kinput·128koutput

MCP Servers

Learn more

No MCP Servers configured

Purpose: 
- When engaged, help a user set up a basic conversational AI assistant for a specific use case to show Rasa Pro and Rime working together

Overview:
- Use Rasa Pro to create a conversational AI assistant
- Use Rime’s API endpoints for Text-to-Speech

Before you get started:
1. Search this Rasa Pro documentation link to understand how it to install it: https://rasa.com/docs/pro/installation/python
2. Search this Rasa Pro documentation link to understand how to build an assistant: https://rasa.com/docs/pro/tutorial
2. Search this Rime documentation link to understand how it works: https://docs.rime.ai/api-reference/quickstart
3. Ask the user to describe their use case

General Instructions:
- Install uv and then create a virtual environment with `uv venv --python 3.11`
- Create the project in the current directory; don't create a new one
- The `RASA_PRO_LICENSE` and `RIME_API_KEY` can be found in .env file and should be used from there

Rasa Pro Instructions:
- Use Rasa Pro 3.13
- Ask the user to run `rasa init --template tutorial` and tell them to confirm when they have done so
- For Action code, just mock the API endpoints
- Ensure there is an `assistant_id` in the configuration
- Ask the user to run `rasa train` and tell them to confirm when they have done so
- Ask the user to run `rasa inspect` to test it out themselves and tell them to confirm when they have done so

Rime Instructions:
- Connect to the Rasa assistant, adding Text-to-Speech in front of it with a Python script
- Automatically play audio using macOS's built-in afplay command
- Search this Rime documentation link to understand the structure of the response: https://docs.rime.ai/api-reference/endpoint/json-wav
- The API returns the audio directly as base64 in the JSON response
- Use `arcana` as the model
- Use `audioContent` parameter, not `audio` parameter
- Use `model` parameter, not `modelID` parameter
- Use `speaker` parameter, not `voice` parameter
- Use `allison` as the `speaker`
- Have the user test the integration by running the script

Prompts

Learn more

No Prompts configured

Context

Learn more
Reference all of the changes you've made to your current branch
Reference the most relevant snippets from your codebase
Reference the markdown converted contents of a given URL
Uses the same retrieval mechanism as @Codebase, but only on a single folder
Reference the last command you ran in your IDE's terminal and its output
Reference specific functions or classes from throughout your project
Reference any file in your current workspace