Weaver I/O
TL;DR
Data Flows in Agent Weaver
Learning how inputs and outputs work in Weaver is essential to building reliable agents.
1. What Each I/O Port Typically Means
| Symbol | Colour | Meaning |
|---|---|---|
| ● Solid | Red input | Required — the agent pauses until this is filled |
| ○ Hollow | Red input | Optional — executes only if input is passed |
| ● Solid | Blue output | Main output — passed to the next block |
| ○ Hollow | Blue output | Derived or secondary output |
2. Input Types
Agent Weaver supports several drag-and-drop AI tool input types:
| Category | Description |
|---|---|
| Text | Freeform entry, LLM prompts |
| Image | Upload or pass image files (JPG, PNG) |
| Audio | Voice/audio inputs |
| Video | Short video clips or references |
| Others | Advanced structured types (JSON, boolean, etc.) |
Subtypes (under “Others”)
| Subtype | Example | Notes |
|---|---|---|
string | "Write a haiku" | Common for prompt-based agent creation |
number | 3.14 | Float or decimal input |
integer | 42 | Whole numbers only |
boolean | true, false | Used for toggles |
array | ["one", "two"] | Great for batching LLM calls |
object | { "key": "value" } | Structured payloads |
binary | PNG, CSV | Stored in object store; returns signed URL |
date | 2025-01-01 | ISO 8601 format |
any | Flexible fallback (not validated) |
Inputs can also be styled using color labels for better workflow clarity.
Goal: Summarise weekly_sales.csv then email report
Skills:
1. Read file from S3 (file input)
2. Summarise table (json output)
3. Email summary (string input)
3. Chaining Outputs to Inputs
Rule of thumb: blue → red only.
- Drag from blue output ports to red input ports
4. When Components Run (or Don’t)
- Runs → all mandatory inputs connected
- Waits → mandatory inputs unresolved
- Skips → mandatory input missing → Weaver bypasses
Knowing this will help you keep the no-code AI agent creation smooth.