Live Local LLM Generation

A small frontend for sending a prompt to your local Ollama server and streaming the generated response as it arrives.

Ollama URL

Model name

Prompt

Ready. Ollama must be running locally and the selected model must be installed.

Streaming output

Ollama streams generated text chunks. This page displays chunks live; it does not expose internal token IDs, embeddings, or next-token probabilities through the standard generate endpoint.

Words / whitespace chunks (visual aid)

These are display chunks, not the model's true tokenizer tokens. For exact token boundaries and IDs, use the tokenizer playground page.

Local setup note: Open this file in a browser. If the browser blocks requests due to CORS, configure Ollama's OLLAMA_ORIGINS to allow your local file/origin, restart Ollama, and retry. Keep this endpoint local; do not expose it publicly without authentication and appropriate security controls.