AI Chat

The AI Chat panel lets you have a conversation about your lab results with an AI that already knows everything about your data — your biomarker history, lifestyle context, supplements, notes, and health goals.

Opening the Chat

Click the chat bubble in the bottom-right corner of the screen to slide the chat panel open. You can also click the Ask AI button in the header. Press Escape to close it.

Chat FAB

The floating chat bubble (FAB) is always visible in the bottom-right corner, giving you one-tap access to the AI from any screen.

Setup Guide

If no AI provider is configured, the chat panel shows a setup guide instead of the conversation view. The guide explains your provider options and includes a Connect with OpenRouter button for one-click OAuth setup — no API key needed. Once connected, the chat is ready immediately.

What the AI Knows

Every message you send includes a full snapshot of your data:

• All lab values across every draw date, with reference ranges and trend direction
• Your nine Context Cards (diet, sleep, exercise, environment, etc.)
• Your Interpretive Lens and Health Goals
• Your supplements and their date ranges
• A timestamped change timeline for context cards (diet changes, stress level shifts, etc.) so the AI can correlate lifestyle changes with lab trends
• Your notes
• For female profiles: menstrual cycle data and phase context

You do not need to paste your results into the chat. Just ask your question.

Active Model Display

The chat header shows the name of your currently active AI model, so you always know which model is responding. Each message also includes a cost footnote showing the estimated token cost for that exchange.

Personalities

The AI adopts a personality that shapes its communication style. Choose from three:

Default

A clear, evidence-informed tone. Explains markers plainly, notes trends, and flags concerns without drama.

House

Takes on the style of a sharp, skeptical clinician who asks uncomfortable questions. Pushes back on assumptions and digs for root causes.

Custom

Create your own persona. Type a name in the custom personality field and click Generate — the AI will create a full personality profile for that persona, including communication style, analytical approach, and philosophical lens. You can edit the generated text before saving.

The Enforce evidence-based accuracy option (off by default) adds a strict disclaimer to the AI's instructions, keeping responses grounded in published research rather than speculation.

TIP

Custom personalities are saved per profile and persist across sessions. You can create a persona based on a specific medical philosophy, a fictional doctor character, or any style that makes conversations more useful for you.

Conversation Threads

The chat panel includes a thread rail on the left side — a list of your past conversations. Each thread is named automatically from your first message, and you can rename any thread by clicking its name.

• Start a new conversation at any time
• Switch between threads without losing history
• Up to 50 threads are stored per profile; the oldest are pruned automatically

On mobile, tap the hamburger icon in the chat header to open the thread list, and use the back button to return to the conversation.

Image Attachments

You can attach images to chat messages — photos of lab reports, supplement labels, food logs, skin conditions, or anything else you want the AI to see.

How to attach:

• Click the paperclip button in the chat input area
• Paste an image from your clipboard (Ctrl+V / Cmd+V)
• Drag and drop an image file onto the chat input

Up to 5 images per message. Supported formats: JPEG, PNG, GIF, WebP.

HD Mode

The HD button next to the paperclip toggles between standard (1024px) and high-resolution (2048px) image quality. HD mode preserves more detail but uses more tokens. Standard mode is usually sufficient for lab reports and supplement labels; use HD for fine print or detailed photos.

Quality Warnings

Before sending, getbased analyzes each image and warns you if it detects issues:

• Blurry — try holding steady or tapping to focus
• Too dark — try better lighting
• Overexposed — try less direct light
• Low resolution — the AI may struggle with fine details

These checks save tokens by catching bad photos before they're sent.

Privacy

All image metadata is automatically stripped before sending. EXIF data — GPS location, camera model, timestamps, device serial numbers — is removed during the resize step. Only raw pixel data reaches your AI provider.

TIP

The attach and HD buttons only appear when your active model supports vision (image input). If you don't see them, switch to a vision-capable model in Settings.

Web Search

The Web toggle in the chat header lets the AI search the internet before responding. This is useful for questions about recent studies, drug interactions, supplement research, or anything where up-to-date information matters.

Toggle it on, ask your question, and the AI will pull in current web results alongside your lab context.

Higher cost

Web search injects search results into the AI's context, significantly increasing input tokens. Expect messages to cost 2–4x more than normal. The cost footnote shows a 🌐 web indicator when search was active.

Availability

Web search is available with OpenRouter and Venice only. The toggle is hidden when using Anthropic or Local AI.

Per-Marker AI Explanations

From any marker's detail view (click a marker name in the sidebar or on the dashboard), you will find an Ask AI button. This opens a pre-populated chat asking the AI to explain that specific marker in the context of your results — without you having to type anything.

Markdown Responses

The AI's responses are rendered with full markdown formatting:

• Headings, bold, and italic text
• Bullet and numbered lists
• Code blocks and inline code
• Clickable links

Responses stream in smoothly as the AI generates them, with a typewriter effect that trickles text at a steady rate for a pleasant reading experience.

Token Costs and What to Expect

Every chat message sends your full lab context + conversation history to the AI. Here's what makes up the token count:

Component	Typical size	Notes
System prompt	~1,300 tokens	Fixed — personality instructions, rules
Lab context	2,000–15,000 tokens	Scales with number of draw dates, markers, and filled context cards
Conversation history	0–10,000+ tokens	Last 30 messages (both yours and AI responses)
Image (current message only)	1,000–5,000 tokens per image	Only attached to the message being sent, not stored in history
Total input per message	~3,000–25,000+ tokens

What drives the cost up

• More draw dates — each date adds values to every marker. 2 dates ≈ 3k lab context; 8+ dates ≈ 10k+
• Filled context cards — each of the 9 lifestyle cards adds 50–300 tokens when filled
• Long conversations — AI responses are often 300–800 tokens each. After 10 back-and-forth exchanges, history alone can be 5k–8k tokens
• Images — a single standard-quality image adds ~1,500–3,000 input tokens. HD images cost more. Images are only sent with the current message, never re-sent in history
• Web search — when the Web toggle is on, search results are injected into the context, adding thousands of tokens. Expect 2–4x the normal cost per message

Realistic cost examples

Using Claude Sonnet 4 via OpenRouter (~$3/$15 per 1M input/output tokens):

Scenario	Input tokens	Output tokens	Cost per message
First message, 2 draw dates, no images	~4,000	~500	~$0.02
Mid-conversation (10 messages), 4 draw dates, all cards filled	~15,000	~600	~$0.05
Long conversation (20+ messages), 8 draw dates, 1 image	~25,000	~800	~$0.09
Discuss mode (2 personas, 3 rounds each)	~20,000 × 6	~600 × 6	~$0.40 total

Cost-saving tips

• Start new threads — click "New Chat" to reset conversation history to zero. This is the single biggest cost saver
• Use standard image mode (not HD) unless you need fine detail
• Local AI is free — run Ollama or LM Studio locally for unlimited chat at zero cost
• Venice offers free-tier models with no per-token charges

Each message shows its cost

Every AI response includes a footnote showing the model name, estimated cost, and total token count — so you always know exactly what you're spending.

Choosing a Provider

The AI chat works with any of the four supported providers: Anthropic, OpenRouter, Venice, or Local AI. See AI Providers to configure your key or local server. The chat is not available until a provider is set up.

WARNING

Conversations are stored locally in your browser and encrypted if you have set a passphrase. The last 30 messages from each conversation are sent to the AI provider with every request to maintain context. Your provider's privacy policy applies to that data.