1. Flow Information Block
Configures flow metadata used by the LLM agent to determine execution context.
- Flow Name: Identifier for the UI and LLM.
- Flow Description: Functional specification for the LLM. Recommended schema:
- Functional purpose of the flow.
- Execution triggers/use-case examples.
- Variable definitions and operational constraints.
2. Flow Variables Block
Defines global input/output variables accessible across all flow blocks.
- Variable Name: Unique identifier (e.g.,
HackerNewsPostsPath). - Variable Default Value: Fallback value. Overridden if a value is supplied at runtime by the LLM or upstream blocks.
- JSON Object Traversal: Extract nested data using dot notation and zero-indexed arrays within any block field accepting variables.
- Syntax Example: For variable
apiResponse:{ "data": { "users": [ { "name": "John", "details": { "age": 30, "city": "New York" } } ] } }- Retrieve “John”:
apiResponse.data.users[0].name - Retrieve “New York”:
apiResponse.data.users[0].details.city
- Retrieve “John”:
- Syntax Example: For variable
3. Flow Complete Block
- Visual execution termination indicator. Contains no configurable parameters.