Goal: Upload one approved document and check its use in a reply.
Prerequisites: Have access to the correct Command project, a demo agent, and a document with two known answers.
Expected result: The intended file shows Trained, and replies and Sources agree with its contents.
The Upload Workflow in raia Command
You manage your agent's knowledge base using the raia Command application. The process is a simple, four-step workflow:
Step 01: Create a Project
In raia Command, create a new project — a workspace to organize related documents. Projects allow you to group documents by topic, department, or agent. Each project has its own folder structure, access controls, and sync settings.
Step 02: Upload Documents
Drag and drop your files into the Data Room. You can queue up to 500 files at once; each file can be up to 100 MB. Supported formats include PDF, Word, Markdown, JSON, HTML, and plain text.
Step 03: Convert to Markdown
Select the files and choose Convert to Markdown. This step transforms your documents into the optimized format the AI can best understand. You can set the project to do this automatically for all new uploads.
During conversion:
- PDFs are parsed for text content (images of text are OCR'd)
- Word documents have their formatting mapped to Markdown headers and lists
- Tables are preserved in Markdown table format
- Complex layouts are simplified into linear, readable text
Step 04: Upload to Vector Store
Select the intended documents, choose Send to AI, then Send Documents to Agent for Training. Keep separate files and select the approved agent only when you intend to queue training; selecting the agent queues training immediately, without a separate confirmation step. Wait for Trained on the exact file before testing retrieval. Convert Ready is not trained status; elapsed time alone does not confirm completion.
Keeping Knowledge Current
Your knowledge base is not a "set it and forget it" system. Business information changes — policies update, products launch, pricing shifts. There are two ways to keep your agent's knowledge current:
Manual Updates
If you upload a file with the same name as one already there, raia Command asks what to do: skip it, or replace (overwrite) the existing file. Only a file you choose to replace is updated. Check the converted content, send the updated file to the intended agent for training, and wait for Trained. Ask about a changed fact and compare the reply and Sources with the updated file; conversion alone does not train the agent.
Connectors (Automatic Sync)
raia Command includes Connectors that automatically pull and sync content from external sources directly into your project:
| Connector | What It Syncs |
|---|---|
| Google Drive | Documents and folders from your Google Workspace |
| Zendesk | Help center articles and ticket templates |
| Salesforce | Knowledge articles and case documentation |
| GitHub | Repository documentation and README files |
| Confluence | Wiki pages and team documentation |
Testing Your Upload
Once you upload a document to the Vector Store, immediately test it:
- Open raia Copilot
- Ask the agent a question about the specific document you just uploaded
- Compare the answer with the source document; check another known-answer question before drawing a conclusion
- If it doesn't, check: Is the document well-structured? Are the headings clear? Is the content between 200–500 words per section?
FAQ: Can I connect my existing Zendesk or Google Drive instead of uploading files manually? Yes. raia Command includes Connectors that automatically pull and sync content from external sources like Zendesk, Google Drive, Salesforce, and GitHub directly into your project.
Practice check
Recommended practice: ask both known-answer questions in Copilot and compare each reply with the source. A single correct answer does not prove the entire upload is usable. If a reply is wrong, check the selected agent, document readiness, and source clarity before repeating.
Practice: upload, train and check one fictional document
Practice with a project and agent you are authorized to change. This fictional Northstar Desk document gives you three known facts, so you can check both the upload and the agent’s answer without using customer data.
Download the fictional Markdown practice file- 1
In raia Command, open your approved Knowledge project and its Data Room. Choose Upload Files, select Documents as the destination, then select northstar-docs-training-demo-20261022.md. Confirm the filename before proceeding.
Select the practice file, then review it before uploading. - Highlight 1: Destination folder: Documents.
- Highlight 2: The selected synthetic Markdown file.
- Highlight 3: Upload Files.
- 2
Click Upload Files in the drawer. Wait for All files uploaded; if it does not appear, inspect the drawer error before retrying. Do not assume that selecting a file uploaded it.
Upload drawer confirmation for the fictional practice file. - Highlight 1: All files uploaded.
- 3
Find the exact filename in All files. Check Convert and Train separately: Ready means the text is available for the next step, not that the agent has been trained. AI Search Index is separate from agent training.
The uploaded Markdown is Ready for conversion, while Train is blank; these are different states. - Highlight 1: The new practice filename.
- Highlight 2: Convert: Ready.
- Highlight 3: Train: blank, not trained.
- 4
Open the new document. Compare the rendered text with the downloaded fixture: 23 days, original power adapter and ORCHID-742. If text is missing, correct the source or conversion before training.
The rendered new document contains fictional practice facts, not customer data. - Highlight 1: Rendered synthetic document: compare its facts with the practice file.
- 5
Choose Send to AI, then Send Documents to Agent for Training. Keep separate files. Before selecting the approved target agent, confirm it is the one you intend: selecting the agent queues training immediately, with no extra confirmation button.
Choose training for reusable knowledge; selecting the agent queues it immediately. - Highlight 1: Send Documents to Agent for Training.
- 6
Read the queue confirmation, then return to the file list to monitor progress. Queued is not completed. If an error appears, read its details and check the target before retrying.
Selecting the target agent queues training immediately. Check the queue, then wait for Trained before asking about the document. - Highlight 1: Training Documents queue confirmation.
- 7
Wait for Trained beside the exact filename. If it does not appear, inspect the job status and error before asking the agent to use the document.
The fictional document later showed Trained in the file list. - Highlight 1: The exact new filename.
- Highlight 2: Trained.
- 8
Open the same target agent in Copilot. Ask: “For the fictional Aurora lamp, what is the return window, required accessory and demo reference? Use northstar-docs-training-demo-20261022.md as the source; take no external actions.” Check for 23 days, original power adapter and ORCHID-742, and compare the cited source with the fixture.
Compare the answer’s facts and citation with the source: 23 days, the original power adapter and ORCHID-742. - Highlight 1: 23 days; original power adapter; ORCHID-742.
- Highlight 2: The exact practice filename and citation.
- 9
Ask: “What are the warranty length and shipping cost in this fictional document?” Neither fact is provided, so the agent should say it does not know from this source. If it invents an answer, check the target agent, document and instructions. Revise one issue and repeat both questions.
For facts missing from the document, check that the agent acknowledges the gap rather than inventing warranty or shipping details.