Use Optic

Select a Chat Model

Based on the previous section, you can use the model selector to choose your preferred model for the conversation.

In the text box on the Optic main interface, there is a button with a settings icon. Click it to adjust various parameters, including Temperature and Top P.

Hover your mouse over the button to see its intent.

Temperature and Top P control the randomness and creativity of the generated text. The higher the temperature, the more random the model's output and the greater its creativity. The Top P mechanism dynamically selects a set of words whose cumulative probability reaches a threshold, then randomly samples from this set, thereby filtering out low-probability long-tail vocabulary and balancing creativity with fluency.

In short, if you're unsure, you can leave these two options disabled; the LLM service will automatically use appropriate values.

Prompts and the Pure Mode

To ensure that the LLM functions properly, we have pre-configured some system prompts. You can toggle these prompts on or off by clicking a button inside the text box.

If you'd like to receive a clean response without any extra prompts, locate the second button to the right of the model selector at the top of the Optic main interface. Clicking it will enable Clean Mode. In this mode, no extra prompts are provided.

Changing the prompt may result in additional charges

Some LLM services offer a feature called “caching.” When you engage in multi-turn conversations with an LLM, previous dialogue content is retrieved from the cache, resulting in lower costs. However, changing the system prompt during a conversation will invalidate the cache, which will incur the full cost.

Therefore, we do not recommend modifying these settings during a conversation.

For LLM services that do not offer caching, the above recommendation does not apply.

Manage Chats

You can create a folder in the sidebar, then right-click any chat and move it into that folder. If a conversation is located within a Workspace, you cannot place it into a specific folder.

You may have noticed that below the text box on the main interface, there is a selector labeled “Playground” or “Explore.” Changing this will alter the folder or workspace where new conversations take place.

You can also right-click a chat entry in the sidebar to delete or export it.

Minimap

To the left of the “Send” button in the text box, there is a “Collapse” button. Clicking it opens a view of all your sent messages, helping you quickly navigate through long conversations.

Branch messages

When discussing a topic with an LLM, you may come up with multiple ideas and want to ask about each one separately.

In such cases, if the ideas aren't closely related, you don't necessarily need to ask them in sequence. Below each LLM message (except the last one), there is a “Branch” button. Clicking it creates a new branch, allowing you to discuss a different topic within that branch.

The advantage of this approach is that the topics discussed within each branch remain free from interference by other topics. The downside is that if there is a connection between the different topics being discussed—and that connection could lead to new insights—this approach stifles such inspiration.

Message branching is currently recommended for general conversations only and is not recommended for Agentic work scenarios.