Skip to main content

Designer Features

This guide walks through each section of the Designer web UI, explaining what you can do and how to use the various features.

Projects​

The Projects section is your starting point in the Designer. From the home page, you can:

Creating a New Project​

  1. Click the "Create new project" button on the home page
  2. Enter a project name (alphanumeric and hyphens only)
  3. Optionally fill in the project brief:
    • What: Describe what your project does
    • Goals: Define what you want to achieve
    • Audience: Specify who will use it
  4. Click "Create" to initialize the project

The Designer will automatically create a llamafarm.yaml configuration file with sensible defaults.

Switching Projects​

Click on any project card on the home page to switch to that project. Your current project is highlighted in the header.

Editing Project Details​

From any project page, click the project name dropdown in the header and select "Edit project" to update the project brief or rename the project.

Dashboard​

The Dashboard provides an at-a-glance view of your active project.

Designer Dashboard The Dashboard showing project overview and key metrics

What You'll See​

  • Project Configuration Summary: Key settings like runtime provider, model, and enabled features
  • Dataset Statistics: Number of datasets, total files, and processing status
  • Quick Actions: Common tasks like creating datasets or editing prompts
  • Recent Activity: Latest changes to your project configuration

Using the Dashboard​

The Dashboard is read-only and designed for quick reference. Use the action buttons to jump directly to relevant sections when you need to make changes.

Data Management​

The Data section is where you manage datasets and configure how files are processed.

Data Management Interface Managing datasets and processing strategies

Processing Strategies​

Processing strategies define how your data is parsed, chunked, and embedded. You can:

  • View existing strategies: See all configured data processing strategies for your project
  • Create new strategies: Click "Create new" to define a strategy from scratch
  • Copy strategies: Duplicate an existing strategy as a starting point
  • Edit strategies: Click on any strategy card to view and modify its configuration

Each strategy includes:

  • Parsers: How to extract text from different file types (PDF, DOCX, TXT, CSV, Markdown)
  • Chunking: How to split documents into manageable pieces
  • Embedding: Which model to use for creating vector embeddings
  • Storage: Which vector database to use

Datasets​

Below the processing strategies, you'll see your datasets:

  • Create Dataset: Click "+ Create dataset" to add a new dataset
    • Choose a name
    • Select a processing strategy
    • Select a database
  • Upload Files: Drag and drop files onto a dataset card or click to browse
  • Process Dataset: After uploading, click "Process" to start ingestion
  • View Details: Click a dataset name to see files, metadata, and processing status
  • Delete Dataset: Use the dropdown menu on each card

Supported File Types​

The Designer supports uploading and processing:

  • PDF documents (.pdf)
  • Word documents (.docx)
  • Text files (.txt)
  • CSV files (.csv)
  • Markdown files (.md, .markdown)

Config Editor Mode​

Toggle to Config Editor mode to see the raw YAML configuration for datasets and strategies. This is useful for:

  • Copying configuration between projects
  • Making bulk changes
  • Understanding the underlying structure

Models​

The Models section lets you configure which AI models your project uses.

Runtime Configuration​

Select your runtime provider:

  • Ollama (default): Local models on your machine
  • OpenAI: GPT models via API
  • OpenAI-compatible: Custom endpoints (vLLM, LM Studio, etc.)

Model Selection​

Depending on your provider:

  • Ollama: Choose from downloaded models or pull new ones
  • OpenAI: Enter your API key and select a model
  • Custom: Configure the base URL and model name

Model Parameters​

Adjust runtime behavior:

  • Temperature: Control randomness (0.0 = deterministic, 1.0+ = creative)
  • Max Tokens: Limit response length
  • Top P: Nucleus sampling threshold
  • Frequency Penalty: Discourage repetition
  • Presence Penalty: Encourage topic diversity

Testing Connections​

Use the "Test Connection" button to verify your model configuration works before saving.

Databases (RAG Configuration)​

The Databases section is where you configure Retrieval-Augmented Generation.

RAG Configuration Configuring vector databases and retrieval strategies

Vector Databases​

Create and manage vector databases:

  • ChromaDB (default): Local vector storage
  • Pinecone, Weaviate, etc.: Cloud-hosted options

Each database needs:

  • A unique name
  • Storage configuration (connection details for hosted options)

Embedding Strategies​

Configure how text is converted to vectors:

  • Embedding Model: Choose from available models (e.g., nomic-embed-text)
  • Chunk Size: How large each embedded piece should be
  • Overlap: How much chunks should overlap to preserve context

Retrieval Methods​

Define how relevant documents are found:

  • Similarity Search: Pure vector similarity (default)
  • Hybrid Search: Combines vector and keyword matching
  • MMR (Maximal Marginal Relevance): Balances relevance with diversity
  • Metadata Filtering: Add filters based on document properties

Testing Queries​

Use the built-in query tester to:

  1. Enter a test question
  2. See which documents are retrieved
  3. Verify relevance and ranking
  4. Adjust retrieval parameters if needed

Prompts​

The Prompts section helps you design and test system prompts.

Creating Prompts​

  1. Click "Add Prompt" to create a new prompt template
  2. Give it a descriptive name
  3. Write your system prompt (supports variables like {context}, {question})
  4. Add any user message templates

Prompt Sets​

Organize related prompts into sets:

  • Default prompts that always run
  • Conditional prompts based on context
  • Specialized prompts for different tasks

Testing Prompts​

Use the test panel to:

  • Enter sample inputs
  • See the rendered prompt (with variables filled in)
  • Get actual model responses
  • Compare multiple prompt versions side-by-side

Prompt Library​

Browse example prompts for common use cases:

  • RAG-enhanced Q&A
  • Document summarization
  • Code generation
  • Data extraction

Chat Interface​

The Chat section provides an interactive way to test your AI project.

Starting a Chat​

Simply type a message and press Enter or click Send. The chat uses:

  • Your configured runtime and model
  • Active system prompts
  • RAG context if enabled

RAG Toggle​

Use the RAG toggle to:

  • On: Retrieve relevant documents before answering (default for RAG projects)
  • Off: Use only the model's training data

When RAG is on, you'll see retrieved document snippets alongside responses.

Session Management​

Each project maintains its own chat session:

  • History persists between visits
  • Clear chat to start fresh
  • Sessions are stored locally in your project directory

Debugging Responses​

The chat interface shows:

  • Retrieved context (when RAG is enabled)
  • Retrieval scores
  • Processing time
  • Token usage (if available)

Use this information to tune your RAG configuration for better results.

Config Editor​

Every section in the Designer includes a toggle between Designer mode (visual) and Config Editor mode (raw YAML).

When to Use Config Editor​

  • Making precise changes to nested configuration
  • Copying configuration between projects
  • Bulk editing (e.g., updating all chunk sizes)
  • Learning the YAML structure
  • Troubleshooting validation errors

Editor Features​

The Config Editor is powered by CodeMirror and includes:

  • Syntax highlighting: YAML with schema awareness
  • Real-time validation: Errors highlighted as you type
  • Auto-completion: Suggestions based on schema
  • Search and replace: Find and modify across the entire config
  • Line numbers: Easy navigation in large files

Validation​

The editor validates your changes against the LlamaFarm schema:

  • Red underlines indicate errors
  • Hover over errors for explanations
  • Changes won't save until validation passes

Formatting​

The editor automatically:

  • Indents correctly
  • Maintains YAML syntax
  • Preserves comments (if you add them manually)

Switching Between Modes​

In any section, look for the toggle button in the top-right corner to switch between Designer and Config Editor modes. Your place in the document is preserved when switching.

Tips and Best Practices​

For New Users​

  • Start with Designer mode to understand available options
  • Switch to Config Editor once you're comfortable with the structure
  • Use the Dashboard to verify your changes took effect

For Advanced Users​

  • Config Editor is faster for repetitive changes
  • Copy example configurations and modify them
  • Use comments in YAML mode to document your choices

For Troubleshooting​

  • Check Config Editor for validation errors that might not be obvious in Designer mode
  • Compare your config with working examples
  • Use the browser console (F12) for detailed error messages

Next Steps​