LLM Workflow Engine¶
Are you a process nerd who loves the command line and LLMs?
Welcome home 😃
LLM Workflow Engine (LWE) is a Power CLI and Workflow manager for LLMs:
Design reusable, templated prompts
Stress-test prompts across different LLM configurations
Build multi-step workflows that involve LLMs.
At its heart, LWE has four main components:
- Conversation management:
Start new conversations
Review/extend/remove old conversations
- Model configuration:
Configure LLMs across different providers
Customize model attributes (temperature, etc.)
Save/reload existing configurations
- Prompt templates:
Design/save new prompts
Include/pass variables to prompts during execution
Connect a specific prompt to a specific model configuration
- Workflows:
Design multi-step workflows using YAML
Integrate prompt templates
Integrate model configurations
Save LLM interactions to conversations
When combined, these four components provide a lot of flexibility and power for working with LLMs.
Other LWE nicities¶
- Plugins:
Command plugins: Write a command for LWE that accomplishes some new task
Provider plugins: Easily add new LLM providers into the LWE ecosystem
Custom system messages: Easily create and use different system messages for supported providers
Command completion: Tab completion for most commands and arguments
Managed database updates: Automatic database upgrades for new releases
- Examples: To help jump start your productivity
Prompt templates
Workflows
Model presets
OpenAI function definitions
Ansible-compatible playbooks: Re-use LWE workflows inside a larger Ansible ecosystem
Automatic conversation titles: The LLM generates short titles for your conversations
Token tracking: For supported providers, see the number of tokens the current conversation would send in a request, and auto-prune messages from long conversations
Customizable user prompt
Multi-user management
Streaming responses: For supported providers
Command line help
System clipboard integration
Edit prompts in your system editor (Vim, etc.)