Skip to main content

Quick Start Guide

💡TL;DR

Start with one disposable Markdown note and a provider you can already access. Check the generated file before processing a folder. Cloud calls send task content to your provider; a local model still needs a running server and downloaded model.

Step 1: Configure Your First LLM Provider​

  1. Install and enable Notemd, then open Settings → Notemd.
  2. Select your provider, enter its endpoint and credentials where required, and choose a model available to your account. Use Fetch models where the provider supports discovery; a listed model still needs a successful chat test.
  3. Run Test connection. Resolve authentication, model or endpoint errors before starting a workflow.
  4. Search settings for Concept note folder path. Choose a dedicated folder, such as concepts, if you want concept files. The initial folder setting is empty; concepts/ is an example you choose, not a preconfigured destination.

For Ollama, pull a model first and set its exact local name. The preset endpoint is http://localhost:11434/api. For LM Studio, start its local server and use the LMStudio preset. See local model setup. Web research still needs network access even when the summarization model is local.

Step 2: Try Your First Task​

Create Notemd trial.md with a short paragraph you can safely send to the selected provider:

Machine learning models use neural networks to learn patterns from examples.

Run Process file (add links) from the Notemd sidebar or command palette. By default, this produces a separate Notemd trial_processed.md in the source folder and preserves the original. Open the output and inspect the actual links; model output varies. A plausible result is:

[[Machine learning]] models use [[neural networks]] to learn patterns from examples.

A selected concept folder can also receive concept notes. These files and links require review; processing is not an assertion that the model identified every important concept correctly.

Step 3: Explore Features​

Extract Concept Notes​

Use Extract concepts (current file) to extract concepts without rewriting the source. A configured concept folder is required. The standalone extraction defaults are title-only notes and no source backlinks. Enable backlinks explicitly when you need them. Concept notes explains how this differs from add-links output.

Research & Summarize​

Select a topic in an open Markdown editor, or use the note title, then run Research & summarize. Choose Tavily with its API key or DuckDuckGo explicitly in settings. The summary is appended to the current note. Open and verify its sources before relying on a claim.

One-Click Workflows​

The default One-Click Extract button runs this exact definition:

One-Click Extract::process-current-add-links>batch-generate-from-titles>batch-mermaid-fix

It adds links, generates content for eligible notes in the concept or selected folder, then repairs Mermaid blocks in the generated complete folder. It can therefore affect existing eligible concept notes as well as newly created ones. Use a dedicated test folder for the first run. It does not automatically add a web research or diagram-generation step.

Cancellation And Output Review​

Use Cancel in the progress surface to stop further work. Completed outputs remain; cancellation is not an undo transaction and cannot guarantee that a provider stops generating or charging. Wait for the task to settle before starting another run. Inspect the reported output paths and any recovery conflicts rather than assuming a cancelled run produced nothing.

Troubleshooting​

  • Authentication or missing model: check the exact provider profile and rerun Test connection.
  • No concept files: verify that a concept folder is configured and enabled.
  • No useful links: inspect the response and custom prompt. There is no documented minimum word count that guarantees extraction.
  • Rate limits: keep batch concurrency at its default of 1 while investigating account quotas.

Next Steps​