10 Alysium Features You Probably Missed

TL;DR: Ten shipped Alysium features most builders discover late: version history snapshots on every publish, custom CSS widget field, file upload during chat (20 files/30MB each), the About Profile creator card, conversation export to CSV, domain restriction, retrieval instructions for the knowledge base, the income projection simulator, the Help & Support contact card, and full-text conversation search with autocomplete. All available now, no upgrade required.

Most people who build their first Alysium agent find the core loop quickly — upload, configure, embed. These 10 features live one or two clicks deeper and often go undiscovered for months. They're worth finding now.

1. Version History Snapshots

Every time you publish an update, Alysium saves a complete snapshot: your instructions, model choice, parameters, widget configuration, and all attached documents at that moment. Snapshots are listed in descending order. Versions can be marked deprecated with a reason.

Buyers who purchased your agent always receive the version from their purchase date — your future updates don't change their installed copy. This lets you iterate aggressively without fear of breaking anything for existing buyers.

Where to find it: the Version History tab in your agent builder.

The practical use: before doing a major knowledge base overhaul or rewriting your instructions, publish your current agent first to lock in a clean snapshot. That way if the new version behaves unexpectedly, you have a documented prior state to reference. Snapshots include your model choice, parameters, and widget configuration — not just instructions — so you can see exactly what changed between versions.

2. Custom CSS for Complete Widget Styling

The Theme & Appearance section has a custom CSS field almost no one opens. If you or a developer knows CSS, you can override any visual element of the widget — font, spacing, border radius, shadow, bubble shape.

The 36 themes plus hex color override cover most brand needs without it. But when brand standards require exact pixel-level fidelity, the CSS field means there's no visual ceiling.

The practical use: if your site uses a specific font or the 36 built-in themes all feel slightly off, custom CSS is your escape hatch. Common adjustments include font-family overrides, border-radius changes to match site card styles, and chat bubble color tweaks that land between two of the preset themes. The live preview panel reflects CSS changes immediately, so you can experiment without touching your live site.

3. File Upload During Chat

Enable this in Widget Appearance → Features to let visitors attach files to their messages — up to 20 files per conversation, 30MB each.

This transforms the interaction from one-directional Q&A to genuinely collaborative. A writing feedback agent where the student uploads their draft. A coaching intake where the client uploads their goal worksheet. A tax question agent where the visitor uploads their documents. The agent analyzes and references the uploaded content directly in its response.

The practical use: tell visitors in your agent's welcome message or conversation starters that they can attach files. Something like 'You can upload a PDF of your lease and I can help you identify the key clauses' turns the upload feature into an active capability rather than an undiscovered button. The agent treats uploaded files as conversation context — the content is available for the duration of that conversation.

4. The About Profile Creator Card

In Widget Appearance → Agent Menu, there's an About Profile slot. Configure it with your name, creator story, bio, credentials (multiple entries), and optional links: LinkedIn, Twitter, website, booking link, space name, space description, value bullets, and a "Learn more" link.

This is the most underused feature for coaches and consultants. It converts an anonymous AI widget into a credentialed expert AI with a face, a story, and a booking link. The difference in user trust is significant.

The practical use: if you're selling agents on AgentHub, the About Profile is how buyers understand who built the agent and why. Fill in credentials, your methodology backstory, and a 'Book a call' link. Even for free-embedded agents, the About screen establishes trust when someone is deciding whether to keep using your agent or close the widget.

5. Conversation Export to CSV

From the analytics view, export individual conversations (three-dot menu) or all conversations in bulk as CSV or plain text. Exports include messages, timestamps, and conversation metadata.

Most builders discover this when they want to analyze usage patterns. It's also useful for compliance documentation, client records, and identifying knowledge base gaps — search for your escalation phrase in the export to find every question your agent handled poorly.

The practical use: run a monthly export and open it in Google Sheets. Filter for low helpfulness ratings and read those conversations in full. Patterns in why conversations end unhelpfully — missing knowledge base content, instruction ambiguity, or scope questions the agent wasn't configured to handle — are usually visible within the first few rows.

6. Domain Restriction

By default, your script tag loads on any website. Configure domain restriction in the Deployment settings to restrict loading to approved domains only.

Useful for private team knowledge bases (employees only), or for preventing others from copying your embed code and loading your agent on their sites. Leave unrestricted for public consumer-facing agents.

The practical use: if you only want your agent accessible on your paid course platform, domain-restrict it to that URL. Students who find the script tag in your page source and try to embed your agent elsewhere will get a silent load failure. For public-facing agents where reach matters more than restriction, leave this unrestricted.

7. Retrieval Instructions

Inside the Knowledge Base section, there's a retrieval instructions field separate from your main instructions. This field guides how the agent uses the knowledge base during a search — how to prioritize certain documents, when to search broadly vs. narrowly, how to handle conflicting information across multiple documents.

Most builders never configure this. For agents with large or complex knowledge bases with overlapping topics, it can meaningfully improve answer quality and consistency.

The practical use: write retrieval instructions that tell the agent which documents to prioritize for which question types. 'If the question is about pricing, always check the pricing FAQ document first' prevents the agent from citing a pricing reference buried in a general services document when a dedicated pricing doc is available. More specific retrieval instructions reduce the chance of an agent synthesizing an answer from the wrong source.

8. Income Projection Simulator

When you set paid pricing on an AgentHub listing, an income projection tool appears. Enter your expected monthly conversations, average messages per conversation, and your per-message credit price. It instantly calculates projected monthly and annual revenue.

It's not connected to live marketplace data — it's a planning tool. Use it before deciding on a per-conversation price: model a few scenarios at different price points to understand what income requires what volume.

The practical use: use it to pressure-test your pricing before listing. Plug in different per-message rates across conservative, moderate, and optimistic conversation volume assumptions. The simulator makes clear what volume you actually need to hit a meaningful monthly income — which helps you decide whether to price for conversion (low per-message cost, high volume) or for value (higher per-message cost, fewer but higher-intent buyers).

9. Help & Support Contact Card

In Widget Appearance → Agent Menu → Help & Support, you can configure an inline contact card rather than just a URL redirect. Fields: support message, email, and optionally phone, website, Discord link, Twitter, physical address, and free-form additional info.

This is the right way to handle escalations. "I don't have that information" is only a useful response if there's a clear next step. The contact card puts that next step directly in the widget interface.

The practical use: for professional agents, include both an email and a Discord or community link. Visitors who can't find an answer through the agent have a clear next step rather than leaving frustrated. This is especially important for paid AgentHub agents — buyers expect a support path when they're paying per conversation.

10. Conversation Search with Autocomplete

The analytics view has full-text search across all conversations — both user messages and AI responses. As you type, autocomplete suggestions appear from your 10 most recent searches for that agent. Combine with date range filtering to narrow to a specific period.

The most useful searches: your escalation phrase (finds all knowledge gaps at once), a competitor's name (finds all comparison mentions), "pricing" or "cost" (reviews all pricing conversations in one query). This is where patterns emerge that scrolling through the conversation list never reveals.

Where Most Builders Stop

Most people configure instructions and upload a document, then publish. That's enough to get a working agent — but it leaves most of the product unused. The ten features above are what separate an agent that answers questions from one that fits into how people actually work, reflects your brand, and compounds value over time through the data it surfaces.

None of these require any new setup — everything is available in your current builder. Open your agent and check which ones you haven't configured yet.