Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.lmsa.app/llms.txt

Use this file to discover all available pages before exploring further.

LMSA’s chat interface is the heart of the app. Every conversation you have — whether with a local model on your PC or a cloud model via OpenRouter — runs through the same clean, full-featured chat view. Messages are stored locally on your device, and the interface renders markdown, code, and math natively so responses look exactly as the model intended.

​
Starting a conversation

1

Open LMSA and tap the compose icon

Tap the new chat icon in the top bar (or open the sidebar and tap New Chat) to start a fresh conversation.
2

Select a model

Tap the model selector at the top of the screen. LMSA loads the list of available models from your connected provider. Tap any model to activate it.
3

Type your message and send

Type in the input bar at the bottom of the screen, then tap the Send button. By default, the Enter key inserts a new line. You can change this in Settings — toggle Enter sends message to have Enter send and Shift+Enter insert a line break instead.

​
Rich message rendering

LMSA renders AI responses with full formatting support — you are not reading raw text.
Bold, italic, strikethrough, tables, ordered and unordered lists, block quotes, and horizontal rules all render natively inside message bubbles.
Fenced code blocks are highlighted for their language automatically. Tap the copy icon on any code block to copy the contents to your clipboard.
Inline math uses single dollar signs: $E = mc^2$. Display math uses double dollar signs: $$\int_0^\infty e^{-x^2} dx = \frac{\sqrt{\pi}}{2}$$. LMSA renders both using KaTeX with no extra setup required.

​
Thinking and reasoning mode

Some models — such as DeepSeek-R1 — produce a chain-of-thought before their final answer. LMSA exposes this as a collapsible Reasoning Process section inside the message bubble.
  • Show or hide thinking: Tap the [Hide] / [Show] toggle inside any reasoning block to collapse or expand it for that message. You can also set a global default in Settings → Hide thinking to collapse all reasoning blocks automatically.
  • Reasoning level: For models that accept an explicit effort hint, go to Settings → Reasoning level and choose from Default, Low, Medium, High, or Disabled. This controls how much compute the model spends on internal reasoning before replying.
Reasoning mode only appears when your selected model produces <think> blocks. If you do not see it, your current model does not support chain-of-thought output.

​
Smart Reply (Beta)

After each AI response, LMSA can generate a row of quick-reply suggestions tailored to the conversation so far. Tap any suggestion to send it immediately. Enable Smart Reply in Settings → Smart Reply (Beta). Because generating suggestions requires an extra model call, this feature has usage limits on the free tier.

​
Auto-Titles

When you start a new chat, LMSA can ask the model to generate a short, descriptive title based on your first exchange — so your chat history stays organized without manual renaming. Toggle this in Settings → Auto-generate titles.

​
Chat history

Every conversation is saved locally as it progresses. Open the sidebar (swipe right or tap the menu icon) to browse all past chats, tap any entry to resume it, or swipe to delete one.
Tap and hold any message bubble to reveal quick actions: copy the message text, delete the message, or regenerate the response.