Yallma is a WIP platform for quickly designing and deploying AI assistants available through API or as Telegram bots. It is designed to help developers creating feature-rich code-free AI assistant with support for tooling and RAG.
The platform itself is based on Directus. The API endpoint and telegram bots are designed as Directus extensions. Qdrant database is used for storing embeddings for RAG and past conversation search. Embedding models are self-hosted using Ollama. All services run simultaneously using docker compose.