Guides & Documentation

1. Getting Started

Welcome to Nemetron. To open your private sandboxed console and interact with models safely:

1. Navigate to the Authentication Page.
2. Register an account using your email and a strong password, or sign in instantly with a Google account.
3. If registering with an email, check your inbox for a 6-digit verification code and enter it on the OTP verification screen.
4. Once authenticated, click Launch Agent in the navbar.
5. Use the chat composer to start a conversation, upload documents, or search the web.

2. Asking Normal Questions

Standard requests, greetings, and conversational follow-ups are identified by the Deterministic Supervisor and routed to Fast Direct Chat. This bypasses heavier RAG tools, delivering sub-second replies while still automatically maintaining chat context and database memory history.

3. Searching the Web

Ask Nemetron for live, real-time updates or current statistics. The system automatically searches the web to find the latest updates and includes links to the sources.

Note: Live searches require network fetching and may take 3 to 6 seconds to complete. Citations will appear directly under the generated answer.

4. Reading a Webpage URL

To extract content from a specific documentation article, blog post, or page, paste the complete HTTP/HTTPS URL into the chat composer. Nemetron will read the page to summarize its content or answer questions about details.

5. Searching YouTube

You can search for video content directly from the chat interface. The system searches YouTube and displays direct cards containing the video titles, channels, and lengths.

6. Transcribing a Video

Once a video card appears in the chat results, click the Transcribe this video button.

• The input box will pause briefly while the system retrieves the video text.
• Progress alerts appear in the status panel.
• The transcript is saved and automatically added as an active document source in your sidebar.

7. Video-based Q&A

When a video transcript is marked as active in your workspace sidebar, you can ask specific questions about what the speaker said. The agent will search the text and reply with a summary.

8. Uploading a Document

Nemetron accepts PDF, DOCX, TXT, and Markdown documents.

1. Click the attachment icon in the chat console composer.
2. Select the document file from your local storage.
3. Wait for the system to process the document text and save it to your private workspace.
4. The document will automatically display as an active source in the sidebar.

9. Active Sources Panel

An Active Source restricts Nemetron RAG queries to search strictly within that specific target document or transcript.

• Only one document, transcript, or webpage can be active at a time.
• Select a different file from the sidebar to switch active contexts.
• Click the "Close/Clear" button next to the active source badge to return to open conversational search.

10. Conversation History

Your sidebar lists all previous chat conversations.

• The orchestrator automatically generates concise titles from your first question.
• Click a previous title to reload previous messages and RAG context instantly.
• Hover over a conversation card and click the trash icon to delete it.

11. Prompt Limits & Quotas

To prevent system overload, Nemetron applies rolling query limits on a per-user basis.

• The default quota configuration allows 5 prompts within a rolling 15-minute window.
• The sidebar displays a visual limit gauge showing remaining prompts and waiting times.
• If a model query fails due to provider side capacity errors, quota charges are refunded automatically.

12. Account Settings

Access your profile settings from the bottom left of the sidebar:

• Change password securely (requires password validation meeting strength criteria).
• Trigger the email recovery loop if you forgot your credentials.
• Select "Purge My Data" to perform a cascading purge of your user data, including transcripts, files, and chat history.

13. Troubleshooting Guide

14. Responsible Use Policy

When utilizing Nemetron, please follow these guidelines:

• Verify important facts, code snippets, and statistics using the provided citation references.
• Do not upload raw database credentials, API access keys, or highly sensitive corporate financial assets.
• Note that generative models can occasionally make layout errors, output syntax mistakes, or generate incorrect summaries.