> ## Documentation Index
> Fetch the complete documentation index at: https://docs.keloa.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Conditions

> Branch your flow on channel, tag, AI intent, sentiment and more.

A **Condition** node splits the flow into two paths — one for matches, one for non-matches.

## Fields you can match on

| Field             | Example value                                                       |
| ----------------- | ------------------------------------------------------------------- |
| `channel`         | `webchat`, `email`, `whatsapp`, `instagram`, `messenger`, `shopify` |
| `subject`         | Email subject line                                                  |
| `message.body`    | The incoming message text                                           |
| `contact.email`   | Contact's email                                                     |
| `contact.name`    | Full name                                                           |
| `contact.company` | Company name                                                        |
| `contact.tags`    | Any tag on the contact                                              |
| `intent`          | AI-inferred intent (e.g. `shipping`, `refund`, `pre_sales`)         |
| `sentiment`       | `positive`, `neutral`, `negative`                                   |

## Operators

| Operator      | Meaning                            |
| ------------- | ---------------------------------- |
| `equals`      | exact match                        |
| `not_equals`  | not this value                     |
| `contains`    | substring match (case-insensitive) |
| `starts_with` | beginning of the string            |
| `ends_with`   | end of the string                  |
| `in`          | value is one of a list             |

## Combining conditions

One Condition node can have multiple rules. All rules in a condition are **AND-ed** — the branch matches only if every rule passes.

For OR logic, use **two Condition nodes** back to back, each feeding the same next node from their `yes` outputs.

## Multiple branches

Advanced: add more than two branches on a single Condition node. Each branch has its own rule set — the first matching branch wins. This is like `if/else if/else` instead of just `if/else`.

Use multiple branches when you're classifying into more than two buckets — e.g. *intent = refund* → returns team, *intent = sales* → sales team, *default* → AI.

## Intent and sentiment

`intent` and `sentiment` are classified by the AI on every inbound message. They're available as fields for conditions and as analytics dimensions.

Common intents surface automatically — shipping, refund, pricing, technical, cancellation, pre-sales. If the model isn't confident, `intent` is `unknown`.

## Example

Match a high-value customer in distress:

```
Condition:
  contact.tags     contains   "vip"
  AND sentiment    equals     "negative"
→ yes: Handoff (senior team, reason "VIP in distress")
→ no:  let AI continue
```

## Tips

* **Start with one rule.** Add more only when the first one isn't specific enough.
* **Prefer tags over string matching.** A tag is deliberate; a keyword match fires on anything.
* **Test with the Try run.** The panel shows every condition's evaluation with real values.