> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fim.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Artifacts

> Browse, preview, and download all AI-generated files across your conversations.

Artifacts are the files that AI agents produce during conversations — images, HTML pages, code files, documents, and other outputs. The Artifacts page is a cross-conversation gallery that lets you find, preview, and download everything your agents have created.

## How Artifacts Are Created

Any tool that returns a `ToolResult` with an `artifacts` list generates downloadable files. The agent doesn't need special instructions — artifact creation is a natural side effect of tool execution.

Common producers:

| Tool                 | Artifact Type           | Example                         |
| -------------------- | ----------------------- | ------------------------------- |
| `python_exec`        | Images, HTML, CSV, JSON | A matplotlib chart saved as PNG |
| `template_render`    | HTML                    | A formatted report              |
| Code execution tools | Any file                | Generated code, exported data   |
| Image generation     | Images                  | AI-generated illustrations      |

When a tool produces an artifact, it appears inline in the conversation as a downloadable attachment. All artifacts are also collected in the Artifacts gallery for cross-conversation browsing.

## The Gallery

The Artifacts page (`/artifacts`) provides a unified view of all files you've generated across all conversations.

### Filtering

Four type filters narrow the view:

| Filter     | File Types                                                                         |
| ---------- | ---------------------------------------------------------------------------------- |
| **All**    | Everything                                                                         |
| **Images** | PNG, JPG, GIF, SVG, WebP                                                           |
| **HTML**   | HTML pages and reports                                                             |
| **Code**   | Python, JavaScript, TypeScript, JSON, YAML, XML, CSS, SQL, Markdown, shell scripts |
| **Files**  | Everything else — CSV, PDF, plain text, binary files                               |

### Preview

Click any artifact to open an inline preview:

* **Images** — rendered at full resolution
* **HTML** — displayed in a sandboxed iframe (scripts isolated)
* **Code** — syntax-highlighted with language detection
* **Markdown** — rendered as formatted text
* **Other** — shown as plain text or download prompt

Use arrow keys to navigate between artifacts; press Escape to close the preview.

### Metadata

Each artifact shows:

* File name and size
* Creation timestamp
* Source conversation (click to navigate back to the conversation where it was generated)

## Storage

Artifacts are stored on the server filesystem, one directory per conversation. Each artifact gets a UUID-based filename, making URLs unguessable without access to the conversation. Files are served directly — no database records for individual artifacts, keeping the storage layer simple and fast.