Slash Commands Reference

All commands in DSPy Code start with /. This page lists every command with examples.

Project Commands

/init

Initialize or scan your DSPy project.

/init

What it does:

Creates dspy_config.yaml if it doesn't exist
Scans your project code
Indexes your installed DSPy version
Shows entertaining jokes while indexing!

When to use:

Starting a new project
After installing new packages
When you want to refresh the code index

Example Output

📖 Building Codebase Knowledge Base

What's being indexed:
  • Your installed DSPy
  • Your project code
  • GEPA optimizer (if installed)

🤖 Teaching the AI to read your code...
✓ Index built successfully!

💡 You can now ask questions about:
  • DSPy concepts
  • Your project code
  • Code examples

/status

Show your current session status.

/status

What it shows:

Generated code in context
Training data (if any)
Conversation message count
Connected model status
Available commands

Use This When

You're not sure if code was generated
You want to see what's in your session
You forgot if you saved your code

Example:

/status

Output:

📊 Session Status

✓ Generated Code: module
  Lines: 45
  Characters: 1234
  Available commands: /save, /validate, /run

💬 Conversation: 6 messages
🤖 Connected Model: llama3.1:8b (ollama)

💡 Next Steps:
  /save <filename>.py - Save your code
  /validate - Check for issues
  /run - Test execution

/clear

Clear conversation history and start fresh.

/clear

What it does:

Clears all conversation messages
Removes generated code from context
Gives you a clean slate

This Cannot Be Undone

Make sure to save any code you want to keep before clearing!

Code Generation Commands

/create

Interactive wizard for creating DSPy components.

/create

What it does:

Asks you questions about what you want to build
Guides you through the process
Generates complete code

Or just describe what you want:

Create a sentiment analyzer

Build a question answering system

Make a text classifier for emails

Natural Language Works!

You don't need to use /create. Just describe what you want in plain English!

/demo

See example DSPy programs in action.

/demo

What it shows:

Complete working examples
Different DSPy patterns
Code you can learn from

Run a specific demo:

/demo sentiment

/demo qa

/demo rag

/examples

Browse DSPy templates and patterns.

/examples

What you'll see:

Signature templates
Module patterns
Complete program templates
Industry-specific examples

Generate from a template:

/examples generate rag

Model Connection Commands

/model

Interactively select and connect to a model.

Basic usage:

/model

You will be prompted to choose:

Local models via Ollama (auto-detected from your Ollama installation)
Cloud providers: OpenAI, Anthropic, Gemini (you then type the model name, e.g. gpt-5-nano, claude-sonnet-4.5, gemini-2.5-flash)

Shortcuts:

/model ollama   # Only show local Ollama models and pick one by number
/model cloud    # Pick a cloud provider, then type the model name

/connect

Directly connect to a specific model if you already know the name.

Ollama (Local):

/connect ollama gpt-oss:120b

OpenAI:

/connect openai gpt-5-nano

Anthropic Claude:

/connect anthropic claude-sonnet-4.5

Google Gemini:

/connect gemini gemini-2.5-flash

API Keys

For cloud providers, set your API key first:

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GEMINI_API_KEY=AIza...

/disconnect

Disconnect from the current model.

/disconnect

When to use:

Switching to a different model
Working offline with templates
Troubleshooting connection issues

Code Management Commands

/save

Save generated code to a file.

/save sentiment.py

/save my_module.py

Where it saves:

Default: generated/ directory
Creates directory if it doesn't exist
Won't overwrite without confirmation

Always Save!

After generating code you like, save it immediately. Use /status to check if you have code to save.

/validate

Check your code for issues and best practices.

/validate

Validates last generated code

/validate my_program.py

Validates a specific file

What it checks:

Signature correctness
Module structure
Best practices
Common mistakes
Security issues

Quality Assurance

Always validate before running. It catches issues early!

/run

Execute your DSPy program safely.

/run

Runs last generated code

/run my_program.py

Runs a specific file

Safety features:

Runs in a sandbox
Timeout protection
Error handling
Safe execution

Sandbox Execution

Your code runs in a safe sandbox. Some operations (like file I/O) may be restricted.

/test

Run tests on your code.

/test

/test my_program.py

What it does:

Runs unit tests
Checks functionality
Reports results

Optimization Commands

/optimize

Generate GEPA optimization code.

/optimize

Be Aware of Optimization Cost

GEPA optimization can issue many LLM calls, especially when using cloud providers (OpenAI, Anthropic, Gemini). Make sure you understand your API pricing and quotas before running /optimize, and prefer small budgets when starting out. For local models, we recommend at least 32 GB RAM for comfortable optimization runs.

What it creates:

Complete GEPA optimization script
Metric with feedback
Training data loaders
Configuration

Options:

/optimize --budget light

/optimize --budget medium

/optimize --budget heavy

Real GEPA

This generates code that uses REAL GEPA optimization from your installed DSPy version!

/eval

Generate evaluation code for your program.

/eval

What it creates:

Evaluation metrics
Test harness
Reporting code

Help Commands

/help

Show all available commands.

/help

What you'll see:

Complete command list
Usage examples
Tips and tricks

/intro

Complete guide to DSPy Code.

/intro

Perfect for:

New users
Learning all features
Understanding workflows

/explain

Get explanations of DSPy concepts.

/explain signatures

/explain modules

/explain optimizers

Topics you can ask about:

Signatures
Modules
Predictors
Optimizers
GEPA
Metrics

Session Commands

/session save

Save your current session.

/session save my-work

What it saves:

Conversation history
Generated code
Session context

/session load

Load a saved session.

/session load my-work

Restores:

All conversation messages
Generated code
Session state

/sessions

List all saved sessions.

/sessions

/session delete

Delete a saved session.

/session delete my-work

Data Commands

/save-data

Save generated training data.

/save-data examples.jsonl

/save-data training.json

When to use:

After generating training examples
Before optimization
For backup

MCP Commands

/mcp-add

Add an MCP server.

/mcp-add my-server --transport stdio --command "python server.py"

/mcp-connect

Connect to an MCP server.

/mcp-connect my-server

/mcp-disconnect

Disconnect from an MCP server.

/mcp-disconnect my-server

/mcp-list

List all MCP servers.

/mcp-list

/mcp-tools

List tools from an MCP server.

/mcp-tools my-server

Utility Commands

/history

Show conversation history.

/history

Show all messages:

/history all

/refresh-index

Rebuild the codebase index.

/refresh-index

When to use:

After updating DSPy
After adding new code
If index seems stale

/index-status

Check the status of your codebase index.

/index-status

Shows:

Index age
Number of files indexed
Cache size
Staleness warning

/exit

Exit DSPy Code.

/exit

Or use:

exit

quit

bye

Command Tips

Tab Completion

Many terminals support tab completion. Type / and press Tab to see available commands!

Command History

Use the up arrow to recall previous commands.

Partial Commands

You can often use shortcuts: - /h for /help - /s for /status - /v for /validate

Case Insensitive

Commands are case insensitive: - /HELP works - /Help works - /help works

Next Steps

Now that you know all the commands, learn how to use them effectively:

Generating Code → Model Connection →