KCG LM Studio User Guide

Use this app to chat with a local language model and ask questions about your own documents.

Install and Set Up LM Studio

1. Download and install LM Studio from the official website. (https://lmstudio.ai/)
2. Open LM Studio and download at least one chat-capable model. (Tested with qwen3-4b-instruct-2507)
3. Load the model in LM Studio.
4. Start the local server in LM Studio.
5. Confirm the server is listening on http://localhost:1234.
6. In this app, open the model dropdown and select the loaded model.

If you changed LM Studio to another port, update this app’s server URL setting to match.

Quick Start

1. Start your local LM Studio server.
2. Open the app and wait for models to load in the model dropdown.
3. Select the model you want to use.
4. Type your message in the bottom input box and press Enter to send.
5. Use Shift+Enter for a new line without sending.

The WORKING indicator is shown while a response is in progress.

Chat Features

• Responses are appended in the conversation area.
• Long replies are automatically continued and appended when needed.
• The conversation view auto-scrolls to the latest output.

Ask Questions About Your Documents (RAG)

1. Click Attach Files.
2. Select up to 5 files at a time.
3. Total file size per attach must be under 30 MB.
4. Supported formats: PDF, DOCX, TXT, CSV.
5. Ask a specific question that includes names, dates, topics, or key phrases from your documents.

When relevant content is found, the app injects retrieved excerpts into the model prompt so answers are grounded in your files.

Click Clear Docs to remove all currently attached/indexed documents.

Tips for Better Results

• Be explicit in your question: include section titles, terms, and context.
• Ask one focused question at a time for best retrieval quality.
• If no relevant excerpts are found, rephrase with more specific wording from the document.

Troubleshooting

• No models shown: verify the local LM Studio server is running and reachable.
• No files could be indexed: try a small .txt file first to verify file indexing is working.
• PDF/DOCX not loading: your environment may be missing required text extraction tools.
• Slow large responses: this is normal for local generation on larger outputs.