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

> Agent Library is HeyMilo’s model picker. For each part of an interviewer: live call, text screen, resume pass, or form—you choose which HeyMilo model runs that step.

# Agent Library

Newer **Frontier** versions are the strongest option; **Active** ones are still supported for production;

<Frame>
  <img src="https://mintcdn.com/heymilo/khmklczV5WPXMedr/images/Screenshot%202026-04-02%20at%202.47.18%E2%80%AFPM.png?fit=max&auto=format&n=khmklczV5WPXMedr&q=85&s=e816d54d67b740c3947a6b9e215cda61" width="1610" height="418" data-path="images/Screenshot 2026-04-02 at 2.47.18 PM.png" />
</Frame>

**Sidebar → Interviewers → Agent Library.**

### Two kinds of models

<Tabs>
  <Tab title="Interactive">
    **Interactive** models talk *with* the candidate: voice/video or SMS/WhatsApp. They can follow up and adapt as the conversation goes.
  </Tab>

  <Tab title="Non-interactive">
    **Non-interactive** models work on a **file or submission**: they score a resume or a completed form and hand you a result. No live back-and-forth.
  </Tab>
</Tabs>

Use the in-app labels and **Compare models** when two options look close—that view lines up reasoning, languages, cheat detection, SMS extras like nudges or call escalation, and latency.

### Voice interviewer

<Frame>
  <img src="https://mintcdn.com/heymilo/khmklczV5WPXMedr/images/Screenshot%202026-04-02%20at%203.13.24%E2%80%AFPM.png?fit=max&auto=format&n=khmklczV5WPXMedr&q=85&s=2421ff0b490630671acf787d83c81211" width="2146" height="1462" data-path="images/Screenshot 2026-04-02 at 3.13.24 PM.png" />
</Frame>

Live **phone or web** interviews: your script, intelligent follow-ups, scoring. **HeyMilo Voice v3** is the current **Frontier** line—broad language support, strong structure and follow-ups, rejoin-friendly, detailed scoring, and cheat detection where you turn it on. **Voice v2** stays **Active** for teams that stay on that generation.

### SMS screener

<Frame>
  <img src="https://mintcdn.com/heymilo/khmklczV5WPXMedr/images/Screenshot%202026-04-02%20at%204.54.55%E2%80%AFPM.png?fit=max&auto=format&n=khmklczV5WPXMedr&q=85&s=c1d1e4c0f900722ef074f8881383f94d" width="2164" height="1444" data-path="images/Screenshot 2026-04-02 at 4.54.55 PM.png" />
</Frame>

Text screening on **SMS or WhatsApp**. **HeyMilo SMS v4** is **Frontier**: more judgment, tools, auto-nudges, and call escalation when your setup uses them. Additionally, V4 supports multiple choice and free text. **SMS v3** is the **Active** workhorse for high-volume, structured eligibility screens.

### Resume screener

<Frame>
  <img src="https://mintcdn.com/heymilo/khmklczV5WPXMedr/images/Screenshot%202026-04-02%20at%204.00.10%E2%80%AFPM.png?fit=max&auto=format&n=khmklczV5WPXMedr&q=85&s=6f90834d91fa6bb16885391a7a325bd1" width="1654" height="1128" data-path="images/Screenshot 2026-04-02 at 4.00.10 PM.png" />
</Frame>

**Non-interactive.** Uploads or ingested resumes are scored against your role: fit, must-haves, criteria, signals for AI-generated content, and a clear verdict—**HeyMilo Resume v1** is the **Active** model here.

### Form collector

<Frame>
  <img src="https://mintcdn.com/heymilo/khmklczV5WPXMedr/images/Screenshot%202026-04-02%20at%203.58.39%E2%80%AFPM.png?fit=max&auto=format&n=khmklczV5WPXMedr&q=85&s=0a31bf2a415e764ffc7571e577c441ce" width="1582" height="832" data-path="images/Screenshot 2026-04-02 at 3.58.39 PM.png" />
</Frame>

**Non-interactive.** Custom forms with validation and AI evaluation of answers—**HeyMilo Form v1** is **Active**.

### How this fits with Interview Templates

**Agent Library** = **which engine** powers each stage. **Interview Templates** = **saved whole interviewer setups** you reuse. Pick models when you care about version and capability; use templates when you want a full configuration copied forward.

### Practical flow

1. Match **stage type** to **family**—voice, SMS, resume, or form.
2. Default to **Frontier** unless your team has standardized on a specific **Active** build.
3. Open **Compare models** if you’re choosing between two versions.
4. Do question and rubric work in the **interviewer editor**—the library is for **choosing the model**, not writing questions.

### Next steps

* [Configuring Your AI Interviewer](../../getting-started/log-in/configuring-your-ai-interviewer)
* [Interview Templates](interview-templates)
* [Languages & Voices](languages-and-voices)
* [AI Settings](.)