Ollama with Memory

Conversational context via historyGet() — the LLM remembers what was discussed. Uses Ollama with token streaming.

Requires Ollama running locally on port 11434 with the llama3.1 model. Will not work on GitHub Pages.

Code

const chat = new quikchat('#chat', ollamaStreamingCallback, {
  theme: 'quikchat-theme-light',
  titleArea: { title: 'Memory Chat', show: true, align: 'left' }
});

function ollamaStreamingCallback(chatInstance, userInput) {
  const startPrompt = {
    content: 'You are a skilled assistant.',
    role: 'system'
  };
  let start = true, id;
  chatInstance.messageAddNew(userInput, 'user', 'right');

  fetch('http://localhost:11434/api/chat', {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
      model: 'llama3.1',
      // historyGet() provides full conversation context
      messages: [startPrompt, ...chatInstance.historyGet()],
      stream: true
    })
  })
  .then(r => r.body.getReader())
  .then(reader => {
    return reader.read().then(function process(result) {
      if (result.done) return;
      const data = JSON.parse(new TextDecoder().decode(result.value).trim());
      const content = data.message.content;
      if (start) {
        id = chatInstance.messageAddNew(content, 'bot', 'left');
        start = false;
      } else {
        chatInstance.messageAppendContent(id, content);
      }
      return reader.read().then(process);
    });
  });
}