Roleplay Agent from a file or resource
You can create AI roleplay agents using uploaded training files and documents during agent creation. Fine-tuning (or refinement) workflows can also use file attachments so behaviour stays aligned with specific source material, not only a free-form prompt.
The upload experience supports multiple common file types (for example PDF, DOCX, TXT, and other types available in your Resources library) so agents can be trained from real documents alongside scenario and roleplay settings.
Why use it?
- Factual grounding — Product specs, FAQs, and policies reduce hallucinations and off-brand answers.
- Persona fidelity — Persona writeups, example dialogues, or transcripts steer tone and objections.
- Faster iteration — Swap or add files when tuning instead of rewriting long prompts.
Description
During agent creation, you complete the usual fields (for example roleplay type, scenario, description). You can then attach documents either by uploading new files or selecting existing entries from the Resources library, depending on the UI.
Fine-tuning or edit flows apply the same pattern: attach or update files so refinements respect the latest materials. Some flows may limit the number of attachments at one time; follow on-screen prompts to remove or replace files if needed.
Use cases
| Use case | Example |
|---|---|
| Product expert agent | Attach FAQs and spec sheets so answers stay within documented facts. |
| Buyer persona | Attach persona briefs or sample dialogue for tone and objection patterns. |
| Compliance | Attach approved messaging summaries so the agent does not contradict policy. |
| Ongoing tuning | Add a new competitive brief when fine-tuning without a full prompt rewrite. |
Quick reference
| Topic | Detail |
|---|---|
| Purpose | Train and tune agents using documents plus prompts. |
| File types | Typically PDF, DOCX, TXT, and others supported in Resources. |
| Typical path | Roleplay → Agents → create or edit → attach files. |