# Features

URL: https://docs.hanzo.ai/docs/chat/features

> Hanzo Chat features — presets, file uploads, RAG, search, vision, conversation import/export, bookmarks, and more.

Hanzo Chat provides a full-featured AI chat experience. This page covers all major features available in both the hosted and self-hosted versions.

## Multi-Model Chat

Switch between models mid-conversation or start new chats with any available model:

- **Enso** — `enso-auto`, and the paid tiers `enso-flash`, `enso-pro`, `enso-ultra`: [which for what, and prices](https://docs.hanzo.ai/docs/models/enso)
- **Zen** — `zen5`, `zen5-pro`, `zen5-coder`, `zen5-flash`, plus `zen-vl` for vision and `zen-guard` for safety
- **Carried models** — models from outside labs are available too, picked from the same model menu
- **Model comparison** — run the same prompt against multiple models side-by-side
- **Streaming** — real-time token streaming for all models

## Presets

Save and reuse conversation settings as presets:

- **Create presets** from model settings (model, temperature, system prompt, top\_p, penalties)
- **Import/Export** presets as JSON files for sharing
- **Default preset** — set a preset to load automatically for new conversations
- **Custom names** — give presets display names for quick identification

### Creating a Preset

1. Open model settings in a conversation
2. Configure model, temperature, system prompt, and parameters
3. Click **Save As Preset**
4. Name it and save — it appears in the preset menu

### Preset Parameters

| Parameter             | Description                                               |
| --------------------- | --------------------------------------------------------- |
| **Model**             | Which model to use (e.g., `zen5`, `zen5-coder`)           |
| **Temperature**       | Controls randomness (0.0 = deterministic, 2.0 = creative) |
| **Top P**             | Nucleus sampling — controls diversity of output           |
| **Frequency Penalty** | Penalizes repeated tokens                                 |
| **Presence Penalty**  | Encourages new topics                                     |
| **System Prompt**     | Custom instructions prepended to every message            |

## File Uploads & RAG

Upload files to conversations for context-aware responses using Retrieval-Augmented Generation:

- **Supported formats** — PDF, DOCX, TXT, MD, CSV, JSON, code files, images
- **Document indexing** — files are chunked, embedded, and stored in PGVector
- **Semantic search** — relevant chunks are retrieved based on your messages
- **Reusable files** — upload once, attach to any conversation via the side panel
- **Configurable** — chunk size, overlap, and embedding model are all adjustable

### How RAG Works

1. **Upload** a file to a conversation
2. **Indexing** — the RAG API chunks the document and creates embeddings
3. **Retrieval** — when you send a message, relevant chunks are found via semantic search
4. **Generation** — retrieved context is appended to your prompt for accurate responses

### Configuration

```bash
# Embeddings come from Hanzo. The RAG API's key/base-URL variables are named
# for the wire format they speak; point them at api.hanzo.ai and use a Hanzo key.
RAG_OPENAI_API_KEY=sk-your-api-key
RAG_OPENAI_BASEURL=https://api.hanzo.ai/v1
EMBEDDINGS_MODEL=zen-embedding

# Optional tuning
CHUNK_SIZE=1500
CHUNK_OVERLAP=100
```

### Resend Files Toggle

By default, files are queried on every message. Toggle **Resend Files** off in conversation settings for targeted queries — retrieval only happens when files are explicitly attached to a message.

## Vision & Multimodal

Analyze images directly in conversations:

- **Drag and drop** images into the chat
- **Paste from clipboard** — screenshots, diagrams, photos
- **Supported models** — `zen-vl` for images in, text out; carried vision models work too

## Conversation Management

- **Persistent history** — all conversations stored in DocumentDB
- **Search** — full-text search across all conversations via Meilisearch
- **Folders** — organize conversations into folders
- **Bookmarks** — bookmark important conversations for quick access
- **Import** — import conversations from ChatGPT, other LibreChat instances, or JSON
- **Export** — export conversations as JSON or Markdown

### Conversation Import

Import conversations from:

- **ChatGPT** — export from OpenAI and import the JSON
- **LibreChat** — import from other instances
- **JSON format** — standard conversation JSON

## Search

Full-text search powered by Meilisearch:

- Search across all conversation messages
- Instant results with typo tolerance
- Requires Meilisearch service (included in Docker setup)

```bash
# Enable search
SEARCH=true
MEILI_HOST=http://search:7700
MEILI_MASTER_KEY=${SEARCH_MASTER_KEY}
```

## Code Artifacts

When models generate code, Chat provides:

- **Syntax highlighting** for all major languages
- **Copy button** for quick clipboard access
- **Code execution** (when enabled) for Python, JavaScript
- **Diff view** for code modifications

## Keyboard Shortcuts

| Shortcut       | Action           |
| -------------- | ---------------- |
| `Enter`        | Send message     |
| `Shift+Enter`  | New line         |
| `Ctrl+/`       | Toggle sidebar   |
| `Ctrl+Shift+S` | Save as preset   |
| `Ctrl+Shift+N` | New conversation |

## Multi-Language Support

Chat supports 50+ languages for the UI. Set the default language:

```bash
APP_LANG=en  # en, es, fr, de, ja, zh, ko, pt, ru, ar, ...
```

## Theming

- **Dark mode** — default, optimized for extended use
- **Light mode** — available via settings
- **Custom CSS** — override styles for self-hosted instances