Artifacts and Saved Responses

Create and manage artifacts in PromptOwl's Data Room - upload documents, save AI responses, and build knowledge bases for RAG.

This guide explains how to create, manage, and use artifacts in PromptOwl's Data Room, including saving AI responses for future reference.

---

Table of Contents

1. What Are Artifacts?

2. Accessing the Data Room

3. Creating Artifacts

4. Saving AI Responses

5. Managing Artifacts

6. Organizing with Folders

7. Sharing Artifacts

8. Using Artifacts with RAG

9. Vector Sync and Citations

10. Best Practices

11. Troubleshooting

---

What Are Artifacts?

Artifacts are content pieces stored in your Data Room that serve as your knowledge base. They can be:

• Text documents you write

• Uploaded files (PDF, CSV, DOCX)

• Web links with scraped content

• Saved AI responses from conversations

Why Use Artifacts?

Use Case	Benefit
Knowledge Base	Store company documents, policies, FAQs
RAG Source	AI retrieves relevant content when answering
Response Archive	Save valuable AI outputs for reuse
Team Resource	Share knowledge across your organization
Citation Source	AI references artifacts with proper attribution

Artifact Types

Type	Description	Icon
Text	Manual text content	Document
Link	Web URL with scraped content	Link
File	Uploaded PDF, CSV, DOCX	File
Document	Saved AI response	Document

---

Accessing the Data Room

Navigation

1. Click Data Room in the left sidebar

2. View your artifacts and folders

Screenshot: Data Room Navigation

Data Room Interface

The Data Room displays:

Element	Description
Folders	Organized collections of artifacts
Root Artifacts	Artifacts not in any folder
Create Button	Add new artifacts or folders
Search	Find artifacts by name
Sync Status	Shows if content is indexed for AI

Screenshot: Data Room Overview

---

Creating Artifacts

Creating a Text Artifact

1. Click Create in the Data Room

2. Select Artifact

3. Choose Paragraph Text type

4. Fill in the details:

Field	Description	Required
Name	Artifact title	Yes
Content	Main text content (max 15,000 chars)	Yes
Title for Citation	How it appears when cited by AI	No
Author	Content creator	No
Publish Date	When content was created	No
Tags	Categories for organization	No

1. Click Create

Screenshot: Create Text Artifact

Creating a Link Artifact

1. Click Create → Artifact

2. Select Link type

3. Enter the URL

4. System automatically scrapes and processes content

5. Add metadata (title, author, date)

6. Click Create

What Gets Scraped:

• Main content area of the page

• Headers and paragraphs

• Lists and tables

What Gets Removed:

• Navigation menus

• Footers and sidebars

• Ads and cookie banners

• Modals and popups

Uploading Files

1. Click Create → Artifact

2. Select File type

3. Upload your file:

   • Drag and drop onto the upload area

   • Or click to browse and select

Supported Formats:

Format	Extension	Max Size
PDF	.pdf	500 MB
CSV	.csv	500 MB
Word	.docx	500 MB
Text	.txt	500 MB

1. Add metadata (citation title, author, date)

2. Click Create

Screenshot: File Upload

Bulk Upload

For uploading multiple files at once:

1. Click Bulk Upload button

2. Drag multiple files or select from file browser

3. Files upload simultaneously

4. Each becomes a separate artifact

5. Edit individual artifacts later for metadata

---

Saving AI Responses

Save valuable AI responses directly to your Data Room for future reference.

How to Save a Response

1. Have a conversation with your AI prompt

2. Find a response you want to save

3. Click the Save icon on the response

4. Response is saved as an artifact

Screenshot: Save Response Button

Where Saved Responses Go

Saved responses are automatically organized:

Scenario	Folder Location
Prompt has a name	Folder named after the prompt
Prompt unnamed	"Conversation Artifacts" folder
Folder already exists	Added to existing folder

Saved Response Details

Each saved response includes:

• Name: "Response - [timestamp]"

• Type: Document

• Content: Full AI response text

• Description: Links back to original conversation

• Date: When it was saved

Finding Saved Responses

1. Go to Data Room

2. Look for folder matching your prompt name

3. Or search by "Response" to find all saved responses

---

Managing Artifacts

Viewing Artifact Details

1. Click on an artifact card

2. View full content and metadata

3. See how content is chunked for AI retrieval

Screenshot: Artifact Detail View

Editing Artifacts

1. Open the artifact detail page

2. Click Edit button

3. Modify any fields:

   • Name

   • Content

   • Citation title

   • Author

   • Publish date

   • Tags

4. Click Save

Note: Only artifact Owners and Editors can make changes.

Viewing Content Chunks

For RAG purposes, artifacts are split into chunks. To preview:

1. Open artifact detail page

2. Scroll to Chunk Preview section

3. See how the AI will search and retrieve content

Each chunk shows:

• Chunk number

• Content preview

• Character count

Deleting Artifacts

1. Click the trash icon on the artifact card

2. Confirm deletion in the dialog

3. Artifact is permanently removed

Warning: Deletion also removes the artifact from the vector store. This cannot be undone.

---

Organizing with Folders

Creating Folders

1. In Data Room, click Create

2. Select Folder

3. Enter folder name

4. Click Create

Moving Artifacts to Folders

1. Create artifacts directly in a folder:

   • Open the folder first

   • Then create the artifact

2. Or drag artifacts (if supported in UI)

Folder Structure

Data Room
├── Marketing Materials/
│   ├── Product Overview.pdf
│   ├── Pricing Guide.docx
│   └── FAQ Document
├── Technical Docs/
│   ├── API Reference
│   └── Integration Guide
├── Saved Responses/
│   └── Response - Dec 30, 2025
└── [Unfiled Artifacts]

Folder Sync Status

Folders show sync status indicators:

Status	Meaning	Action
Synced	All artifacts indexed	Ready for RAG
Pending	Changes not yet indexed	Will sync automatically
Modified	Content changed since sync	Needs re-sync

---

Sharing Artifacts

Individual Artifact Sharing

1. Open artifact detail page

2. Click Share button

3. Enter email addresses or select teams

4. Click Share

Screenshot: Share Artifact Dialog

Folder Sharing

Share an entire folder to give access to all artifacts within:

1. Open folder

2. Click Share on the folder

3. Add emails or teams

4. All artifacts in folder become accessible

Permission Levels

Role	Can View	Can Edit	Can Delete	Can Share
Owner	Yes	Yes	Yes	Yes
Editor	Yes	Yes	No	Yes
Viewer	Yes	No	No	No

Shared Artifact Indicators

Shared artifacts show:

• Number of people shared with

• "Shared by: [email]" if shared by someone else

• Team badges if shared with teams

---

Using Artifacts with RAG

Artifacts power RAG (Retrieval Augmented Generation), allowing AI to reference your content when answering questions.

Connecting Artifacts to Prompts

There are three main ways to connect your Data Room content to prompts:

Method 1: Using Variables (Recommended for Direct Content Injection)

1. Add {data} (or any variable name) in your prompt content, e.g., Use the following data: {data}

2. The Variables section will automatically detect the variable

3. Click Connect Data button next to the variable

4. Select an artifact or folder from your Data Room

5. The entire content will be injected into the prompt when used

Screenshot: Variable with Connect Data

Screenshot: Connect Data Dialog

Use this method when: You want the entire content injected directly into the prompt (good for small, specific content).

Method 2: Using Block Settings Dataset Field (For RAG/Semantic Search)

1. Click on the block header (e.g., "Block 1") to expand the block

2. Look for the block's LLM Settings or Tools and Data section

3. Find the Dataset field within the block settings

4. Click Connect Data to select folders or artifacts from your Data Room

5. AI will search and retrieve relevant chunks from the connected dataset when the prompt runs

Screenshot: Block Settings Dataset

Screenshot: Dataset Connection

Use this method when: You want AI to search and retrieve relevant parts from large knowledge bases (RAG - Retrieval Augmented Generation).

Method 3: Using Tools & Data Section in Supervisor Mode

When using Run with Supervisor mode, you can connect datasets directly in the block's Tools & Data section:

1. Click on a block in your Supervisor agent (e.g., "SQA_BOOK_BLOCK")

2. Click Block Settings to expand the block settings

3. Find the Tools & Data section

4. Click the Artifact button (this is the Data field)

5. Select folders or artifacts from your Data Room

6. The connected dataset will be used for RAG when that specific block executes

Screenshot: Supervisor Tools & Data

Use this method when: You're using Supervisor mode and want to connect datasets to specific blocks in your agent workflow. Each block can have its own dataset for specialized knowledge retrieval.

How RAG Works

User asks question →
AI searches connected artifacts →
Finds relevant chunks (top 10) →
Uses context to answer →
Cites sources in response

RAG vs Direct Variable

Approach	When to Use	How It Works
RAG (Dataset)	Large knowledge bases	AI searches and retrieves relevant parts
Variable	Small, specific content	Entire content injected into prompt

For detailed RAG configuration, see Understanding Agents and RAG.

---

Vector Sync and Citations

Understanding Vector Sync

Artifacts must be "synced" to be searchable by AI. Syncing:

1. Splits content into optimized chunks

2. Generates embeddings (semantic vectors)

3. Stores in secure vector database

4. Enables similarity-based retrieval

Sync Status Indicators

Status	Badge	Meaning
Synced	Green	Content indexed and searchable
Unsynced	Yellow	New artifact, not yet indexed
Modified	Orange	Content changed, needs re-index

When Sync Happens

• Automatic: When folder is accessed

• On Save: When artifact is created or edited

• Manual: Via sync button (if available)

Citation Display

When AI uses an artifact, citations appear:

Citation Information:

• Display title (from "Title for Citation" field)

• Author name

• Publish date

• Source URL (for links)

• Relevance score

Screenshot: Citation Display

Improving Citations

For better citations, always fill in:

Field	Impact on Citation
Title for Citation	Shows as citation header
Author	Attribution in citation
Publish Date	Recency indicator
URL	Clickable source link

---

Best Practices

Content Organization

Do:

• Use descriptive folder names

• Group related content together

• Keep artifacts focused (one topic each)

• Update content when it becomes outdated

Don't:

• Create one giant artifact with everything

• Mix unrelated topics in same folder

• Leave artifacts without metadata

• Duplicate content across folders

RAG Optimization

Do:

• Write clear, well-structured content

• Use headers to organize sections

• Include key terms users might search for

• Test retrieval with sample questions

Don't:

• Upload scanned images without OCR

• Rely on heavily formatted tables

• Expect AI to read charts/graphs

• Use abbreviations without definitions

Citation Quality

Do:

• Fill in all metadata fields

• Use meaningful citation titles

• Include author for accountability

• Add publish dates for time-sensitive content

Storage Management

• Review artifacts periodically

• Delete outdated content

• Consolidate duplicate information

• Archive old responses if not needed

---

Troubleshooting

Artifact Won't Sync

1. Check sync status indicator

2. Try editing and saving the artifact

3. Open the containing folder (triggers sync)

4. Verify content isn't empty

5. Check for special characters in content

AI Not Finding Relevant Content

1. Verify artifact is synced (green status)

2. Check artifact is in connected dataset

3. Test with different question phrasing

4. Review chunk preview to see how content is split

5. Ensure content contains relevant keywords

Link Scraping Failed

1. Verify URL is accessible (not behind login)

2. Check URL uses http:// or https://

3. Try the URL in a browser first

4. Some sites block automated scraping

5. Consider copying content manually as text

File Upload Errors

1. Check file size (max 500 MB)

2. Verify file format is supported

3. Ensure file isn't corrupted

4. Try re-saving file and uploading again

5. Check file isn't password protected

Citations Not Appearing

1. Verify artifacts have citation metadata

2. Check "Title for Citation" is filled

3. Ensure RAG is enabled on prompt

4. Confirm artifact is in connected dataset

5. Check citation display is enabled in enterprise settings

Shared Artifact Not Visible

1. Verify sharing was saved

2. Ask owner to check sharing settings

3. Ensure recipient email is correct

4. Check team membership (for team shares)

5. Recipient may need to refresh page

---

Quick Reference

Artifact Creation Checklist

Required:

• Name

• Content or file

Recommended:

• Title for Citation

• Author

• Publish Date

• Folder assignment

Supported File Types

Type	Extensions
Documents	.pdf, .docx, .txt
Data	.csv
Max Size	500 MB

Artifact Limits

Limit	Value
Text content	15,000 characters
File size	500 MB
Chunk size	1,000 characters
Chunk overlap	200 characters
RAG results	Top 10 chunks

---

Related Guides

• Adding and Organizing Documents

• Understanding Agents and RAG

• Monitoring, Evaluation, and Improvement