Skip to content

binex start

Synopsis

binex start [OPTIONS]

Description

Interactive wizard that guides you through creating a complete agent workflow. Asks questions step-by-step — choose a pipeline pattern, select an LLM provider, configure API keys — and generates a ready-to-run workflow.yaml.

The wizard uses DSL patterns from binex scaffold and the provider registry to produce valid workflow files without manual YAML editing.

Workflow

  1. Choose a pipeline pattern — select from 9 built-in patterns (simple, diamond, fan-out, fan-in, map-reduce, etc.) or enter a custom DSL expression
  2. Select an LLM provider — pick from 8 supported providers (OpenAI, Anthropic, Gemini, Ollama, Groq, Mistral, DeepSeek, Together)
  3. Configure the model — choose a specific model from the provider's catalog
  4. Set API key — enter or confirm the API key (reads from environment if available)
  5. Name your workflow — give it a descriptive name
  6. Generate — creates workflow.yaml in the current directory

Options

Option Type Default Description
--output / -o Path workflow.yaml Output file path

Example

# Launch the interactive wizard
binex start

# Then run the generated workflow
binex run workflow.yaml --var topic="quantum computing"

Available Patterns

Pattern DSL Description
simple A -> B Linear two-node pipeline
chain A -> B -> C Three-node chain
diamond A -> B, C -> D Diamond dependency
fan-out A -> B, C, D One-to-many parallel
fan-in A, B, C -> D Many-to-one aggregation
fan-out-fan-in A -> B, C, D -> E Parallel with aggregation
map-reduce A -> B, C -> D MapReduce pattern
pipeline-5 A -> B -> C -> D -> E Five-stage pipeline
review A -> B -> C -> D Draft-review-approve chain

Supported Providers

Provider Prefix Example Models
OpenAI openai/ gpt-4, gpt-4o, gpt-3.5-turbo
Anthropic anthropic/ claude-sonnet-4-20250514, claude-haiku-4-5-20251001
Google Gemini gemini/ gemini-pro, gemini-1.5-pro
Ollama ollama/ llama3, mistral, codellama
Groq groq/ llama3-70b, mixtral-8x7b
Mistral mistral/ mistral-large, mistral-medium
DeepSeek deepseek/ deepseek-chat, deepseek-coder
Together together_ai/ llama3-70b, mixtral-8x7b

See Also