# Troubleshooting

Most Curator problems come down to two things. Either **Ollama** is not running or has no model assigned, or background processing has not finished. Find the section that matches what you are seeing.

## AI features are greyed out or do nothing

**Problem.** **Search** by meaning, **Chat**, or **Organize** is disabled, empty, or
nothing happens when you use it.

**Cause.** AI features depend on Ollama. They stay off until Ollama is reachable and a
model is assigned to the task you are trying to use. If you skipped model downloads
during setup, no models are assigned yet.

**Fix.** Open **Settings** and go to **AI & Models**.

1. Check the connection badge next to the **Ollama Base URL**. It reads **Not tested**,
   **Testing**, **Connected**, or **Offline**. You want **Connected**. If it shows
   **Offline**, see the next section.
2. Check the model assignments. Each task needs its own model. Chat needs a Chat model, and Search by meaning needs an Embeddings model.
3. If a task has no model, pull a recommended one and assign it.

[Models and assignments](/reference/models/)
  [Settings reference](/reference/settings/)
## Curator can't find Ollama, or shows "Offline"

**Problem.** The connection badge in **AI & Models** stays on **Offline**, and no AI
feature works.

**Cause.** Ollama is not installed, it is not running, or the **Ollama Base URL** points
somewhere Curator cannot reach.

**Fix.** Work through these in order.

1. Make sure Ollama is installed. If you skipped it during setup, Curator can install it for you in one click.
2. Make sure Ollama is actually running. On a Mac it runs in the background once started.
3. Check the **Ollama Base URL** in **AI & Models**. For Ollama on the same Mac it should
   be `http://localhost:11434`. If you connected to Ollama on another computer, confirm
   that address is correct and that the other machine is on and reachable.
4. Use the connection test next to the field. When it reads **Connected**, AI features will work.

[Understanding Ollama and models](/concepts/ollama-and-models/)

## A model is too large, or responses are very slow

**Problem.** Answers take a long time, the Mac fan spins up, or the whole app feels
sluggish while AI runs.

**Cause.** Local models run in your Mac's memory. A model larger than your available RAM will run slowly and make the whole system sluggish. Running many jobs at once makes it worse.

**Fix.**

1. In **AI & Models**, switch to a smaller recommended model. Curator's suggestions are sized to fit your Mac's RAM.
2. In **Settings**, open **Processing** and lower **Concurrent jobs**. Fewer jobs at
   once means less memory pressure.

> [!TIP]
> Smaller models answer faster and use less memory. The difference in quality is often
> small for everyday search and chat.

## Files don't appear after adding a source

**Problem.** You added a folder or bucket, but the files are not showing up, or not all
of them are.

**Cause.** Curator indexes a source in the background. A sync may still be running, or files may still be processing.

**Fix.** Open **Processing** and **Jobs** to check status. A sync detects files, then processing makes them searchable. Both finish on their own. If nothing is running, start a sync from **Settings**.

[Processing and Jobs](/using/processing-jobs/)

## Search by meaning returns nothing useful

**Problem.** **Semantic** search comes back empty or with results that feel unrelated.

**Cause.** Search by meaning relies on embeddings. Files that have not been processed yet have no embeddings, so they cannot be matched. If no **Embeddings** model is assigned, search by meaning cannot work.

**Fix.**

1. Wait for processing to finish. Check **Processing** for progress. Results improve as more files complete.
2. In **AI & Models**, confirm an **Embeddings** model is assigned. Without one, search by meaning has nothing to work with.

## S3 or cloud storage won't connect

**Problem.** Connecting a **Cloud object storage** source fails the connection test.

**Cause.** One of the connection details is wrong, or the SSL setting does not match the
provider.

**Fix.** Recheck each field when you add or edit the source.

- **Display name** is just a label and does not affect the connection.
- **Endpoint** must match your provider exactly.
- **Access Key** and **Secret Key** must be correct and still valid.
- The bucket name must exist and be reachable with those keys.
- The **SSL** toggle must match what your provider expects.

[Manage your sources](/using/manage-sources/)

## The app won't open after download

**Problem.** Curator does not launch after you download it.

**Cause.** Curator is Apple-signed and notarized. macOS should open it without a Gatekeeper warning. If it silently fails to open, your macOS security settings may be the cause.

> [!NOTE]
> PLACEHOLDER: Confirm whether any specific macOS version shows an edge-case prompt on
> first launch, and document the exact wording and the click-through if so.

## Where to find logs for support

If you need to report a problem, the logs help pin it down.

- In the app, open **Settings** and go to **Logs** to view application logs and
  per-run organization traces.
- On disk, the log file is at `~/Library/Logs/com.bleural.curator/Curator.log`.

> [!NOTE]
> PLACEHOLDER: Add the support contact or channel where users should send logs.