RagLeap
Blog

Open-source AI chatbots for businesses in India: what RagLeap Core does today and what is still research

RagLeap Core is open-source software (MIT) that you run yourself. The hosted RagLeap service that used to be reachable at this address has been retired. This page says what the open-source project does today for businesses that work in more than one language, and what is still research. It makes no claim about how many languages work, and it is not legal advice.

What exists today

  • Channels. WhatsApp, Telegram, Discord and a voice channel (through Twilio) are available. The channels documentation says what each one needs.
  • Language detection. RagLeap detects the language of a question with FastText (through the fast-langdetect package), a character n-gram fallback for hard short queries, and script-based checks for Chinese, Japanese and Korean text. Short queries in closely related languages can still be misdetected, because there is little text to work with. Since v0.11.0, LANGUAGE_DETECTION_MODEL=lite lets an install without internet use the bundled model. Details are in the language documentation.
  • Your choice of AI. You bring your own key and can use a local model through Ollama. Documents and conversations are stored on your own infrastructure, and the text sent for an answer goes to the provider you configure.

What is research, not a feature

The ragleap-voice package (version 0.0.2, pre-alpha) is a model manager for open, local speech-to-text and text-to-speech models, with Indian languages as a priority. It has no speech engines: you cannot transcribe or synthesize audio with it yet. Its published benchmarks use small samples of read speech on one GPU. In them an open Indian-language recogniser (IndicConformer) had clearly lower word error than Whisper large-v3 on Hindi, Tamil, Telugu, Malayalam and Kannada. Naturalness, emotion quality and comparisons with commercial services have not been measured. See the library page.

Limits you should know

  • Tested by the maintainer on one Windows laptop and on Linux servers. macOS, Linux desktops and the Ollama option on Windows are not verified yet.
  • No named users or roles yet, and no independent security review yet.

See the roadmap page for what is planned (no dates).

Back to the blog