Warning
This project is still in development and undergoing rigorous testing. It will be released through a beta (when it's that stable) or through iterations that will be available for people to download and fully use. For the beta version, we would appreciate if people are willing to test for bugs as it will help further refine the project and anything that is broken before a suitable release candidate.
This project is also the foundational structure that is created and used for Lotus, which is another project within Hikarian forge that goes above and beyond with SLMs (Small/Specific Language Models) and training your own agent orchestration framework.
Link: Hikarian-Forge/Lotus
Leon, Your digital assistant, on Discord.
Leon is a secure, autonomous AI assistant built for one-on-one use across multiple social platforms. Starting on Discord and expanding towards BlueSky, Twitter, UpScrolled, and more. Leon is designed to replace a specific fragile, "vibecoded" bot with a more principled, production-grade, assistant that you can trust not to provide malicious users with your credit card information.
She operates within her own controlled domain, powered by local and configuration LLMs (via Ollama or OpenRouter, alongside Gemini), and an entire STT-LLM-TTS voice pipeline for real-time voice channel communication.
Here are some of the following features that are planned for Leon:
- Cross-Platform: Starts on Discord, with planned support for Bluesky, Twitter, UpScrolled, and additional platforms as integrations mature.
- Configurable LLM Backend: Supports local inference via Ollama or cloud models through OpenRouter — BYOK supported, your data stays yours.
- Voice Pipeline: Real-time voice interaction using a STT → LLM → TTS setup.
- STT: High-accuracy transcription model (TBD).
- LLM: Ollama or OpenRouter-routed models.
- TTS: Qwen3-TTS-1.7B or Kokoro-TTS for expressive, human-like audio output.
- Persistent Memory & RAG: PostgreSQL-backed storage with pgvector or Weaviate for retrieval-augmented generation across conversations.
- Skill System: A
SKILL.md-driven capability framework that defines and expands what Leon can autonomously execute on your behalf. - Security-First: Scoped permissions, controlled execution environments, and no third-party data exposure — purpose-built to be the opposite of insecure, poorly-audited bots.
As for its architecture, the mono repo structure is something I prefer
when making projects of this nature. The architecture is better explained
through the /docs/arch.README.md file.
| Command | Description |
|---|---|
/ask |
Ask Leon anything such as conversational queries, tasks, or instructions |
/help |
Display available commands and usage guidance |
/stats |
View session statistics and system metrics |
/system prompt |
Set or view the active system prompt |
/system context |
Manage context window state |
/system temperature |
Adjust LLM temperature |
/system top-p |
Adjust nucleus sampling (top-p) |
/system max-completion-tokenization |
Set max completion token limit |
/ping |
Check Leon's responsiveness and connection status |
/join |
Invite Leon to a voice channel to begin a voice session |
/leave |
Remove Leon from the active voice channel |
/docs |
Access inline documentation |
/faq |
View frequently asked questions about Leon |
Note
You will need to have Python 3.11+ and PostgresSQL installed to be able to run this project, or Docker if intending to use the docker compose file for Postgres. As for LLMs, you can use Ollama.
Additionally, if you are testing for a specific feature or looking for a specific mechanic to test out, you may need to search around to find it, or if it doesn't exist yet you may need to further check the documentation details to see the status for some of our systems.
To get started with Leon, follow these steps to set up your local environment:
- Copy the environment using
git clone:
git clone https://github.com/hikues-forge/leon.git .- Once cloned, install dependencies using your preferred package manager:
# Using uv (recommended)
uv sync
# Using pip
pip install -r requirements.txt
# Using poetry
poetry install
# Using pipenv
pipenv install- From here, copy the
.env.samplefile into a.envfile for Leon to use:
Warning
Make sure to add your keys for Openrouter or url for Ollama, if you do
not provide any key for OpenRouter, it will always default to Ollama and
and use the default localhost url which is usually http://localhost:11434/api.
cp .env.sample .env - After this, if using Docker, set up the docker compose file:
Warning
You can ignore this if you have a local PostgresSQL instance over using
Docker to set up your instance. If you have both Postgres and Docker, it
would be best you choose one over the other as you will need to provide a
proper DATABASE_URL in your environment variables (.env).
docker compose up -d- Run database migrations after you have provided the DATABASE_URL:
alembic upgrade head- Now you should be able to run Leon!
bash scripts/enable.sh && ./scripts/run.shLeon's behavior is configurable through a .env alongside /system commands
that you can run through discord (or preferred platform of choice). Here are
the environment variables you NEED to have:
| Env | Description |
|---|---|
DISCORD_BOT_TOKEN |
Your Discord bot token |
DATABASE_URL |
PostgreSQL connection string |
LLM_PROVIDER |
ollama, openrouter or gemini |
OPENROUTER_API_KEY |
Required if using OpenRouter |
DEFAULT_MODEL |
Default model to route completions to, this is heavily dependent on the platform. |
TTS_ENGINE |
qwen3 or kokoro. |
API_URL |
The API url to connect to a backend service for Leon. |
API_ENV |
This can be either dev, staging or prod |
Warning
A quick disclaimer, I intend to store documentation within a wiki hosted usually hosted on my own site. If it's hosted, you can find it at the following link:
If it's not hosted, then consider checking out the docs directory for more detail regarding the development status of Backstage, alongside some links to socials here and there.
The documentation for Leon is extensively organized into categories
within the /docs directory. If you are looking for information regarding
the systems or mechanics of Leon or its internals, your best bet is checking
out /docs. Additionally, if you have inquiries about Leon, check out the FAQ page
hosted on our site: leon.hikue.dev or email me!
TODO: Write proper information about the system requirements for Leo.
TODO: Write proper information about the contact/socials for Leon.
Contributors are always welcome for Leon. However, we recommend properly reading through our CONTRIBUTING.md file to get a better understanding of the best practices and guidelines that you should follow as a contributor.
With that being said, all the best, and to the moon 🚀!