Àkàndé is a self-hosted, provider-agnostic voice assistant that delivers structured executive briefings from any of 10 supported LLM providers. Ask a question by voice or text; receive a concise Overview / Solution / Conclusion / Recommendations briefing with automatic PDF and CSV artefact generation. Run it against OpenAI, or keep your data entirely private with Ollama or LM Studio on your own hardware.
The name "Àkàndé" (Yoruba: "firstborn") signifies a pioneering approach to voice-driven intelligence briefings.
Ask a question. Get a structured briefing. Keep your data.
That is the entire workflow. Voice, text, or web -- Àkàndé listens, queries the LLM of your choice, and delivers a four-section executive briefing: Overview, Solution, Conclusion, Recommendations. PDF and CSV artefacts are generated automatically.
No framework to learn. No output parsing to build. No provider lock-in to accept.
Ten LLM providers -- including fully private local inference through Ollama and LM Studio -- all behind one environment variable. Your hardware. Your models. Your data.
Every other tool gives you the engine and expects you to build the car. Àkàndé hands you the keys. This is how AI assistants should have always worked.
Python developers who deploy it: clone, configure a provider, and run. Àkàndé is a standard Python package with a CLI entry point, a CherryPy web server mode, and 160 tests.
End users and executives who interact with it: speak or type a question and receive a structured briefing read aloud and saved as PDF/CSV. No technical knowledge required once deployed.
- Python 3.9+
- PortAudio (for microphone input):
Ubuntu/Debian
sudo apt install portaudio19-dev, macOSbrew install portaudio - ffmpeg (for audio conversion):
Ubuntu/Debian
sudo apt install ffmpeg, macOSbrew install ffmpeg
# Install system dependencies (required for PyAudio and audio conversion)
# Ubuntu/Debian:
sudo apt install portaudio19-dev ffmpeg
# macOS:
brew install portaudio ffmpeg
git clone https://github.com/sebastienrousseau/akande
cd akande
python -m venv .venv && source .venv/bin/activate
pip install -e .
cp .env.example .env
# Edit .env: set OPENAI_API_KEY=sk-your-key-here
python -m akandeSelect option 2 ("Ask a question"), type your query, and receive a structured briefing.
Set LLM_PROVIDER in your .env file. Install extras for non-OpenAI
providers.
| Provider | LLM_PROVIDER |
Required env vars | Install extras | Default model |
|---|---|---|---|---|
| OpenAI | openai |
OPENAI_API_KEY |
(included) | gpt-3.5-turbo |
| Anthropic | anthropic |
ANTHROPIC_API_KEY |
pip install akande[anthropic] |
claude-3-haiku-20240307 |
| Google Gemini | google |
GOOGLE_API_KEY |
pip install akande[google] |
gemini-pro |
| Ollama | ollama |
OLLAMA_HOST |
(included) | llama3 |
| Azure OpenAI | azure_openai |
AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT |
(included) | gpt-35-turbo |
| Mistral | mistral |
MISTRAL_API_KEY |
pip install akande[mistral] |
mistral-small-latest |
| Cohere | cohere |
COHERE_API_KEY |
pip install akande[cohere] |
command-r |
| Hugging Face | huggingface |
HUGGINGFACE_API_KEY |
pip install akande[huggingface] |
mistralai/Mistral-7B-Instruct-v0.2 |
| Groq | groq |
GROQ_API_KEY |
(included) | llama3-8b-8192 |
| LM Studio | lmstudio |
LMSTUDIO_HOST |
(included) | local-model |
To install all optional provider SDKs: pip install akande[all]
Àkàndé presents a CLI menu with four options:
- Use voice -- speak your question via microphone; response is read aloud
- Ask a question -- type your question; response is printed and read aloud
- Start server -- launches a web UI at
http://127.0.0.1:8080with voice and text input - Stop -- exits the assistant
Question: "What is quantitative easing?"
Overview Quantitative easing (QE) is a monetary policy tool used by central banks to stimulate economic activity when conventional interest rate cuts are insufficient.
Solution
- Central banks purchase government bonds and other securities to inject money into the economy
- This lowers long-term interest rates, encouraging borrowing and investment
- QE increases the money supply without printing physical currency
Conclusion QE can stabilise financial markets during crises but carries risks of asset bubbles and currency devaluation.
Recommendations
- Monitor inflation indicators alongside QE announcements
- Consider the impact on bond yields for portfolio decisions
Each response is automatically saved as a timestamped PDF and CSV in the output directory.
| Problem | Cause | Fix |
|---|---|---|
Could not find PyAudio |
PortAudio missing | Ubuntu: sudo apt install portaudio19-dev. macOS: brew install portaudio. Then pip install pyaudio. |
ffmpeg not found |
ffmpeg not installed | Ubuntu: sudo apt install ffmpeg. macOS: brew install ffmpeg. |
| Microphone not detected | OS permissions | Grant microphone access in system settings. |
ModuleNotFoundError: No module named 'anthropic' |
Provider SDK not installed | pip install akande[anthropic]. See the provider table above. |
Invalid or missing OPENAI_API_KEY |
Key not set or malformed | Ensure .env contains a valid sk- prefixed key. |
- Streaming responses for reduced time-to-first-token
- Conversation memory with multi-turn context
- Plugin system for domain-specific briefing templates
- Docker image for single-command deployment
- 160 tests covering providers, caching, services, server, and utilities
- Apache 2.0 licence -- permissive, enterprise-friendly
- Euxis-audited architecture and security posture
Pull requests are welcome.
This project is dual-licensed under the Apache License 2.0 and the MIT License -- see LICENSE-APACHE and LICENSE-MIT for details.