> For the complete documentation index, see [llms.txt](https://docs.dapta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dapta.ai/dapti/tips-and-prompting.md).

# Tips & Prompting

Dapti is good at plain language. You do not need to learn a special syntax. But a few small habits will make every conversation better. This page is a collection of the practical tips that experienced Dapta users have found most useful.

***

## Phrase Questions Like You Would Ask a Coworker

Dapti is built to understand the way people actually talk. Skip jargon, skip prompt-engineering tricks, and just ask what you would ask a colleague.

| Good                                                           | Less good                                                       |
| -------------------------------------------------------------- | --------------------------------------------------------------- |
| "How is Lia doing this week?"                                  | "QUERY: agent\_id=Lia, time\_range=7d, metrics=all"             |
| "Improve her prompt so she handles pricing objections better." | "Modify prompt section 3.2 with objection-handling block"       |
| "What's blocking me from getting real calls?"                  | "Diagnose configuration blockers preventing outbound execution" |

The closer your prompt sounds to natural speech, the better Dapti understands intent.

***

## Be Specific About the Subject

Dapti has access to a lot. When more than one agent, flow, or meeting matches your question, the answer gets fuzzy. Add a specific name and the response sharpens.

| Vague                        | Sharper                                               |
| ---------------------------- | ----------------------------------------------------- |
| "Improve my agent's prompt." | "Improve Lia's prompt so she handles pricing better." |
| "Pull my recent calls."      | "Pull Lia's calls from the last 7 days."              |
| "Summarize my meeting."      | "Summarize my May 15 demo with Acme Corp."            |

If Dapti finds more than one possible subject, it will ask you to clarify. Naming the subject up front skips that round-trip.

***

## Add "Text Agent" or "Voice Agent" When the Name Is Shared

If you ever have both a voice agent and a text agent with similar names (or even the same name), tell Dapti which one you mean. Dapti's first lookup defaults to voice agents.

| What to write                             | Why                                 |
| ----------------------------------------- | ----------------------------------- |
| "Improve my Sophia **text agent** prompt" | Skips the failed voice-agent lookup |
| "Show me **voice agent** Lia's calls"     | Clarifies which surface to query    |

***

## Use Follow-Ups Instead of Restarting

Dapti remembers the full context of the current conversation. If the first answer was almost what you wanted, follow up instead of starting a new chat:

* "Good, but only show calls longer than 90 seconds."
* "Now group those by patient type."
* "Apply that change to Lia too."
* "Same analysis but for last month."

This is how you get from a generic answer to one that fits your actual situation.

***

## Attach Images for Visual Questions

Click the **paperclip** icon to attach a screenshot or image. Dapti can read:

* Screenshots of agent settings or flow diagrams ("what does this configuration do?")
* Charts and dashboards ("what's the trend here?")
* Error messages ("why is this failing?")
* Hand-written notes or call logs ("convert this into a list of follow-ups")

Attach the image, type your question, send.

***

## Switch Languages Anytime

Dapti understands and replies in whichever language you write in. Mix Spanish and English in the same conversation if you want. You can also be explicit:

* *"Switch to Spanish."*
* *"Responde en español a partir de ahora."*
* *"Translate that last response to Portuguese."*

Voice and text agents Dapti builds for you can also be configured in any language. Just say so when you create the agent.

***

## Always Confirm Before Big Changes

Dapti always asks before making destructive changes (modifying a live agent, deploying a flow, scheduling a recurring task). When you see an **Apply changes** confirmation card, read it. The summary tells you exactly what will change.

If you want to inspect the underlying configuration before approving, ask:

* *"Show me the full prompt change you're about to apply."*
* *"What nodes will be in the flow after this change?"*
* *"What's the cron schedule you're about to set?"*

***

## When the Answer Looks Off

Sometimes Dapti will surprise you with an answer that does not match what you expected. A few moves that almost always help:

* **Click the "Thought for Ns" indicator** to see the step-by-step reasoning. Often the data was right but the question was interpreted differently than you meant.
* **Ask again, more specifically**: name the subject, the time range, the metric.
* **Try a different angle**: instead of "what's wrong?", try "compare today to last week" or "what changed?"
* **Click 👎** to flag a bad answer. Your feedback shapes how Dapti improves.

***

## Bookmark the Prompts That Work

When you find a prompt that gives you a great answer, save it. Use it again next week, next month, next quarter. A few patterns worth bookmarking:

* Your **Monday morning prompt**: "How did last week go? Compare to the week before."
* Your **end-of-quarter prompt**: "Generate a full performance report covering Q\[N], including industry comparison and top 3 wins."
* Your **pre-campaign prompt**: "Estimate the credit cost for \[N] outbound calls and check whether my plan covers it."
* Your **pre-meeting prep prompt**: "Pull everything you know about \[contact name] from our recent meetings and calls."

***

## Don't Overthink It

The single biggest tip: just ask. Dapti is forgiving with phrasing, will ask clarifying questions when needed, and will tell you if it cannot help. Less time crafting the perfect prompt, more time iterating on real answers.

***

## Next Steps

{% content-ref url="/pages/z2MqYMcg82nfTMVD8SSU" %}
[Use Cases](/dapti/use-cases.md)
{% endcontent-ref %}

{% content-ref url="/pages/WlLQa9OkMJ9pDvj852NH" %}
[Privacy & Scope](/dapti/privacy-and-scope.md)
{% endcontent-ref %}

{% content-ref url="/pages/tGdZKRq1oCqpksmxm3rK" %}
[Capabilities](/dapti/capabilities.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.dapta.ai/dapti/tips-and-prompting.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.