AI Chat

DatumOS Chat lets you interact with your project data using natural language. Ask questions, get insights, and explore information across all your connected data sources.

Starting a Conversation

Create a New Chat

1. Click the New Chat button in the sidebar or use Cmd/Ctrl + N
2. Type your question or request in the message box
3. Press Enter or click the send button
4. Wait for the AI to process and respond

The chat interface streams responses in real-time, so you see results as they’re generated.

Example Questions

Get started with these types of queries:

Project Information

What's the status of the Civic Center project?

Show me all open RFIs for the Oakland Office Building

Find specifications related to mechanical systems

Document Search

Search for documents about fire protection

What submittals are pending approval?

Show me change orders from last month

Analysis

Summarize the key issues in recent RFI responses

Compare the original schedule to actual completion dates

What are the most common causes of delays?

Content Generation

Create a summary of project progress for the stakeholder meeting

Generate a cost breakdown visualization

Draft a response to RFI #1234

Context from Connected Data Sources

DatumOS automatically pulls context from your connected sources:

Autodesk Construction Cloud

When you ask about a project, DatumOS can access:

• Projects & Folders - Navigate your project structure
• Documents - Specifications, drawings, contracts
• BIM Models - Query 3D model data using the AEC Data Model
• RFIs & Submittals - Track outstanding items
• Issues - Review project issues and resolutions

Microsoft 365

Access data from:

• SharePoint - Project document libraries
• OneDrive - Personal and shared files
• Teams - Conversation history (coming soon)

Google Workspace

Query content from:

• Google Drive - All folders and files
• Docs, Sheets, Slides - Document content
• Shared Drives - Team-wide file access

Uploaded Files

Search through files you’ve uploaded directly to DatumOS:

• PDFs, Word documents, Excel spreadsheets
• Images with OCR text extraction
• CAD files and technical drawings

Chat History and Conversations

Accessing Past Conversations

All your conversations are automatically saved:

1. Click Conversations in the sidebar
2. Browse by date or search by title
3. Click any conversation to resume

Conversation Features

• Auto-naming - Conversations are automatically titled based on content
• Search - Find conversations by keyword or topic
• Favorites - Star important conversations for quick access
• Sharing - Share conversation links with team members (coming soon)

Advanced Chat Features

Tool Calling

DatumOS can automatically use tools to answer your questions:

Web Search - Searches the internet for current information

What are the latest LEED certification requirements?

Search your connected systems - Queries your connected data sources

Show me drawings from the structural folder

Artifact Generation - Creates documents and visualizations

Create a project status report with cost breakdown chart

Attachments

Enhance your queries with file attachments:

1. Click the paperclip icon or drag and drop files
2. Add your question about the attachment
3. The AI analyzes both your question and the file

Supported formats:

• Documents: PDF, DOCX, TXT, MD
• Spreadsheets: XLSX, CSV
• Images: PNG, JPG (with vision models)
• Code: JS, TS, PY, and more

Streaming Responses

DatumOS uses streaming to show results as they’re generated:

1. Immediate feedback - See the response start within 2 seconds
2. Progress indicators - Watch tool execution in real-time
3. Cancellation - Stop generation if needed with the stop button

Image Generation

Create AI-generated images directly within your conversations. Click the image icon in the chat toolbar to enable, then ask the AI to create visuals.

See Image Generation for examples and detailed usage.

Document Generation

Control when AI creates formatted documents with the document generation toggle.

Click the notebook icon in the chat toolbar to enable proactive document generation:

• OFF (default): AI only creates documents when you explicitly ask
• ON: AI proactively creates formatted documents, code files, and diagrams when helpful

A visual indicator banner appears when document generation is active, so you always know the current mode.

See Artifacts for more about generated documents and visualizations.

Prompt Selector

Access saved prompts directly from the chat interface using category chips:

• DatumOS - Pre-built prompts optimized for AEC workflows
• Favorites - Prompts you’ve starred for quick access
• Custom - Prompts you’ve created

Click any category chip above the input area, search or browse, then click a prompt to insert it. Hover over prompts to preview the full content.

See Prompt Library for the full guide to creating and managing prompts.

Chat Capabilities

Explore DatumOS AI Chat features in depth:

Tips for Better Search Results

Be Specific

❌ Vague: “Tell me about documents” ✅ Specific: “Show me approved submittals for HVAC equipment from the last 30 days”

Provide Context

❌ No context: “What’s the status?” ✅ With context: “What’s the status of RFI #0234 regarding the fire alarm system?”

Use Project Names

❌ Generic: “Find specifications” ✅ Project-scoped: “Find specifications for the Oakland Office Building project”

Break Down Complex Questions

Instead of:

What are all the issues with the structural, mechanical, and electrical systems, their impact on schedule, and recommendations for resolution?

Try breaking it into steps:

Follow Up Naturally

DatumOS remembers context within a conversation:

Keyboard Shortcuts

Speed up your workflow with these shortcuts:

Performance Expectations

DatumOS is optimized for speed:

• Simple chat: under 2 seconds to first token (95th percentile)
• With tools: under 5 seconds to first token
• Search: Results within 500ms
• Artifact generation: Complete within 10 seconds

Limitations and Considerations

Context Window

Each AI model has a context window limit:

• Claude models: 200,000 tokens (~150,000 words)
• Grok Vision: 131,072 tokens (~100,000 words)

Long conversations may need to be summarized to stay within limits.

Data Freshness

• Connected sources: Real-time access to latest data
• Embedded data: Updated based on sync schedule (typically hourly)
• Web search: Current information from the internet

Cost Management

Different models have different costs per token:

• Claude Haiku 4.5 (default) - Most cost-effective
• Claude Sonnet 4.5 - Higher cost, better reasoning
• Grok Vision - Multimodal capabilities

See AI Models for detailed pricing.

Next Steps

Troubleshooting

Chat not responding

1. Check your internet connection
2. Verify you have an active subscription
3. Try refreshing the page
4. Check the status page  for outages

Incorrect or irrelevant results

1. Be more specific in your query
2. Mention the project name to scope results
3. Try breaking complex questions into steps
4. Switch to a more capable model like Claude Sonnet 4.5

Tool execution fails

1. Check that your data sources are connected ( Settings > Connectors)
2. Verify you have necessary permissions in the source system
3. Try simplifying your query
4. See Help Center for connector-specific troubleshooting

For more help, visit the Help Center or contact support.