---
title: "How do I use Candela AI?"
slug: "how-do-i-use-candela-ai"
updated: 2026-06-15T11:44:21Z
published: 2026-06-15T11:44:21Z
canonical: "support.lusid.com/how-do-i-use-candela-ai"
---

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

# How do I use Candela AI?

To get started with [Candela](/v1/docs/candela-ai) - your AI assistant in LUSID - simply click on the **AI assistant** icon in the top right of the LUSID web app:

![](https://cdn.document360.io/d575ad81-c0ed-4980-bbd1-d59ac5c3de82/Images/Documentation/image-WBKP5I4F.png)

## Your first conversation

When you open Candela, you can enter a prompt that Candela’s Welcome Agent answers directly or routes you to a specialist agent that has the right tool for your request. [Read more about what agents are and how they work.](/v1/docs/understanding-candela-s-system-agents)

**Ask a question directly:**

> [!NOTE]
> ☝️ *“Show me all portfolios in the* `Finbourne-Examples` *scope“*

**Ask what Candela can do:**

> [!NOTE]
> ❔ *“What can you help me with?”*

**Request a specific task:**

> [!NOTE]
> 🪄 *“Run a valuation on my Global Equities fund“*

For example:

![](https://cdn.document360.io/d575ad81-c0ed-4980-bbd1-d59ac5c3de82/Images/Documentation/image-5OG2D9R1.png)

Candela maintains conversation context within a single session. You can start a new conversation at any time, or return to a previous one.

## Tips for effective prompts

### ✅ Do

- [x] **Be specific** - include names, portfolio codes, and date ranges when relevant, for example: ✅ *“Show holdings for portfolio* `UK-Equities` *in* `Finbourne-Examples` *scope."* ❌ *“Show me UK equities.“*
- [x] **Provide context** - if you are looking at a particular fund or date, say so.
- [x] **One task at a time** - break complex requests into steps rather than asking for everything at once.
- [x] **Ask for tables and charts** - say *“return as a table in markdown“* for structured results, or *“visualise the positions in my Global Equities fund“*.
- [x] **Ask follow-up questions** - Candela remembers the context of your conversation.
- [x] **Verify numbers that matter** - Candela shows how it’s processing your prompt; check the logs and cross-reference against the source before making material decisions.

### ❌ Avoid

- [ ] Vague requests with no identifiers - for example *“show me the portfolio“.*
- [ ] Asking for operations you don’t have access to - Candela tells you if a capability is unavailable for your role.
- [ ] Long multi-part requests - break requests into steps for better results.

---

### Troubleshooting

#### `Something went wrong` error

Note the conversation trace ID and contact your LUSID administrator. The trace ID can be used to diagnose the issue.

#### Candela could not complete my request

If Candela tried several approaches but could not finish a task, your request is likely too broad or ambiguous. Try simplifying the request or breaking it into smaller, more specific steps.

#### Candela’s response cut off mid-sentence

The response hit an output length limit. Ask Candela to *"continue"* or *"finish your response"* to get the rest of the response.

#### Candela did not pick the right specialist agent

If Candela did not route your question to the specialist you expected, try rephrasing with more specific terms; for example mention *"portfolio"*, *"valuation"*, *"Luminesce"*, *"Drive"*.

You can ask *"what agents are available?"* to see your agent options.

#### Candela used an unexpected date for *“last 2 weeks“* or a similar relative range

Candela knows the current date and works out relative ranges from it. If a date range looks wrong, restate it with explicit dates which Candela will use directly, for example *"between 2026-05-01 and 2026-06-15"*.