#Smart Chat

Smart Chat uses the local model downloaded through Herdsman to answer questions, hold conversations, and help with everyday tasks.

Smart Chat requires a local LLM. Download one before using this feature. See the Quick Start chapter for details.

#Check the Model

If no model name is shown in the bottom-left of the chat box, no model is installed. Sending a message will trigger a "No chat model installed on this device" notice — install a model first.

#Start a Conversation

#Using Suggested Prompts

Smart Chat ships with Suggested Prompts to help you start quickly.

1. On the chat home screen, the Suggested Prompts area presents shortcuts such as:

   • "How should I prioritize my work today?"
   • "Turn this idea into an actionable plan."
   • "Explain an unfamiliar concept with examples."
   • "Analyze the pros, cons, and risks of a decision."
   • "Rewrite this passage to be clearer and easier to read."

2. Click any prompt to send it as a new message.

   

#Sending a Custom Prompt

1. Type your question into the chat box at the bottom of the screen.

2. Optionally toggle any of the following modes (described in detail in later sections):

   • Web Search — search the live internet for up-to-date information
   • Deep Thinking — produce a more detailed reasoning trace
   • Image Generation — generate an image based on your description

3. Click the send button or press Enter to submit. Wait for the model's reply.

   

   

#Switch Models

When you have multiple models downloaded and running, you can switch between them per conversation.

1. Click the model selector at the bottom-right of the chat box to expand the model list.
2. The list contains the following options:
   • Auto matching — Herdsman picks the most appropriate locally installed model:
     • Auto (Recommended) — best overall fit
     • Speed Priority — uses a smaller model for the fastest response
     • Balanced — best trade-off between speed and quality
     • Quality Priority — uses a larger model for higher quality at the cost of speed
   • Recent — the model used most recently
   • Model Management — lists every downloaded model; click one to switch

#Upload Images

Herdsman supports image uploads in chat, but only models tagged as multimodal can interpret image content.

1. Select a multimodal model such as Qwen3.5_4B.
2. Click the + button on the left side of the chat box and pick an image from your computer.
3. Type any accompanying instructions, then send the message.

#Web Search

Enable Web Search to let the model retrieve live information from the internet.

1. Click the Web Search dropdown at the bottom of the chat box.
2. Choose a search mode or keep the default:
   • Fast — quicker, lighter searches
   • Balanced — default mode
   • Deep — broader, more comprehensive results
3. Submit your question. The reply will incorporate retrieved web results.

#Deep Thinking

Deep Thinking mode produces a more detailed reasoning trace, useful for complex problems.

1. Click the Deep Thinking button at the bottom of the chat box to enable it.
2. Enter a question that benefits from step-by-step analysis.
3. The model shows its reasoning steps in addition to the final answer.

#Image Generation

Generate high-quality images from text prompts in a variety of styles.

1. Click the Image Generation button at the bottom of the chat box.
2. Describe the image you want (the prompt).
3. Wait for the model to produce the result.

#Create a New Chat

Each conversation is an independent "workspace." Starting a new chat is like clearing your desk — the model won't be distracted by leftover context from the previous task, so responses stay accurate.

1. Click New Chat at the top of the left sidebar.

   

#Chat History

All conversations are saved in the left sidebar, grouped by date. Click any entry to resume it.

1. In the left sidebar under Chat History, click any conversation to switch to it.
2. Continue chatting where you left off.
3. Conversations are grouped automatically by date (e.g., 2026-05-21, 2026-05-19).

#Delete a Conversation

1. Hover over a conversation in Chat History and click the three-dot menu on the right.
2. Click Delete.
3. In the Confirm Delete dialog, click Delete again to remove it permanently.

#System Resource Monitor

The bottom-left corner shows the live status of the local service and system resource usage.

• Local service running — a green dot indicates the service is healthy.
• CPU — current CPU usage.
• Memory — current RAM usage.
• Disk — current disk usage.

#WeChat Connection

The bottom-right corner shows the WeChat connection status for messaging integration.

• WeChat Not Connected — click to link a WeChat account and enable chat via WeChat.

#FAQ

Q: Can I upload Word or PDF files to ask questions, or have the model produce Word/PDF output?

• Reading files requires engineering integration that Smart Chat does not yet provide. For file-based workflows, use FlowyAIPC with a local model — FlowyAIPC supports reading and generating files.

Q: How does Web Search work?

• Click the Web Search dropdown at the bottom of the chat box to enable it. The model automatically incorporates live web results into its answer.

Q: What is the difference between Deep Thinking and the normal mode?

• Deep Thinking exposes the model's reasoning trace, making it useful for complex problems. The normal mode gives a direct answer and responds faster.

Q: What styles does Image Generation support?

• Artistic, photorealistic, cartoon, and many others. Describe the style in your prompt, e.g., "cyberpunk cityscape" or "watercolor flower."