RagLeapShort answer: install RagLeap Core with one command, choose your AI (Gemini or a local Ollama model), upload your documents, pick one of the 46 built-in AI employee roles or define your own, connect a channel such as WhatsApp or Telegram, and keep autonomy at its default, off, until you trust the assistant. RagLeap Core is open-source (MIT) software that you run yourself, so the same steps apply to a clinic, a law office, a restaurant, an agency or a warehouse. This page follows the v0.11.0 README and release notes and says what is not verified.
What do you need?
- A computer or server with Docker. The installer checks for it, and checks free disk space (4 GB, or 14 GB if you bundle the local AI).
- Either a Gemini API key (the default) or a local Ollama setup with no key.
How do you install it?
curl -fsSL https://raw.githubusercontent.com/ragleap/ragleap-core/main/install.sh | bashThe installer asks which AI to use, generates your API key and secrets, and starts everything. Afterwards the ragleap command handles launch, stop, status, logs, update and key. The self-hosting guide covers the stack and its limits.
How do you give it your knowledge?
The Documents tab (since v0.11.0) lists, uploads and deletes documents. It takes 28 file types (PDF, Word, Excel, PowerPoint, CSV, HTML, EPUB and more) or a public web page by address. Answers cite the source document.
How do you choose what it does?
RagLeap Core ships 46 role-based AI employees, from manager, support and sales to real estate, hospitality, healthcare intake, legal intake, logistics, tutoring and IT helpdesk. You select a role with role=<role> on the chat endpoint. If none fits, PATCH /employees/{role} defines a custom role with its own name, personality, channels and skill tags, without forking the code. See the full role list.
How do customers reach it?
Channel adapters exist for WhatsApp, Telegram, Discord and voice (through Twilio), configured in .env. A web chat widget for your own website is planned but not built yet. See the channels documentation.
How do you keep it under control?
Autonomy has three modes: off (the default), semi (the assistant asks before acting) and full (opt-in). The README says sensitive roles always need approval. Optional daily and monthly token budgets can be set globally or per role.
What should you check before real customers use it?
- Your documents and each conversation go to the AI provider you configure. A local Ollama model keeps those steps on your machine, but local models on a CPU are slower and less accurate than Gemini.
- One deployment serves one business: single-tenant, one document set.
- No named users or roles yet, and no independent security review yet.
- Tested by the maintainer on one Windows laptop and on Linux servers. macOS, Linux desktops and Ollama on Windows are not verified.
- No industry-specific testing or results are claimed.
Sources
The v0.11.0 README, the v0.11.0 release notes and the documentation.