dhaga.docs
Using Dhaga

Capturing people

Add a person to your graph from pasted text, a card photo, voice, an upload, or by typing them in — pick the method from one list, review the fields, and write your first note about them on the same form.

Capture is one click from anywhere. A Capture button in the app nav opens the capture dialog on every /app screen — the fastest path in. The same capture lives behind the + in the middle of the phone's bottom bar, the Capture button on Home, and the dedicated /app/quick-add page, which additionally lets you attach the new person to an existing event (below). All run the same extraction, keep the original input as a receipt, and meter AI usage per account. A card scan or a paste-to-contact costs 1 credit, and the free tier includes 10 AI credits a month — ten card scans, say, or five scans and five notes. Run out and the app stays fully usable for the methods that don't use AI. See AI credits.

Pick how you're adding them

Every one of those entry points opens on the same question — "How do you want to add them?" — over a grid of all seven methods. There's no default surface and nothing hidden a level down: desktop, phone and the Home button all show the same list.

  • Paste text — an email signature, a LinkedIn intro, a DM
  • Scan a card — photograph a visiting card or badge
  • Record a voice note — just talk; Dhaga writes it down
  • Upload a photo — choose a card photo already on your device
  • Add a person — type their name, role and contact details
  • Log a relationship — record who knows whom, and how
  • Note a fact or follow-up — something to remember, or something to do

Pick one and you land straight on it, with a ← All methods link in the header to come back and choose differently. The four AI methods (paste, scan, voice, upload) are marked Needs credits when your AI credits have run out — they stay clickable, and the surface itself explains the gate, so you're never guessing which of the seven still work. A method your browser can't run — voice without WebGPU — is greyed out with a Coming soon note instead.

Under the grid, a quiet line reads Adding lots of people? with a link to Import from Google, LinkedIn or a file. That's the answer to a different question — a whole address book rather than one person — so it sits below the seven rather than competing with them. See Importing contacts.

The methods that don't use AI — Add a person, Log a relationship, Note a fact or follow-up — keep working when you're out of credits, or on a deployment with no AI provider configured.

The Dhaga capture method chooser asking "How do you want to add them?" over all seven methods — Paste text, Scan a card, Record a voice note, Upload a photo, Add a person, Log a relationship, Note a fact or follow-up — with an "Adding lots of people? Import from Google, LinkedIn or a file" line beneath

Dhaga manual add-person form with fields for name, jobs, emails, phones, and links

What each method does

  • Paste text — drop in an email signature, the text off a business card, or a LinkedIn intro, and Dhaga pulls out the name, title, company, and contact methods. A Voice mic sits under the box: tap it and describe who you met; the dictation streams live into the text and is extracted the same way. With Dhaga Voice the transcript comes back as tappable word-chips — tap a mis-heard word to fix it, and Dhaga learns that spelling for next time. Record a voice note lands on this same surface — it just names the mic as the reason you came, rather than making you find it under the box.

The Dhaga quick-add Paste text surface, ready to accept an email signature or card text, with the Voice mic beneath it

Where your dictation is transcribed

Dhaga Voice runs one of two engines, and which one you get is a property of the deployment, not of your plan:

  • On our server. Where the deployment has speech-to-text configured, you record, stop, and the clip is uploaded and transcribed by our provider — so the audio does leave your device, the same way a card scan's photo does. This engine needs no WebGPU, which is what makes dictation work on iOS Safari and every other browser. It costs one AI credit per clip, there is no live word-by-word transcript (the text arrives when you stop), and a single recording is capped at about two minutes.
  • On your own device. Otherwise the speech model runs in the browser and the audio never leaves your machine, streaming words as you talk. That engine needs WebGPU — available today in desktop Chrome and Edge — and it costs no credits.

Where neither can run — no server transcription and no WebGPU — Dhaga says so up front: the mic is greyed out with a "Coming soon" note explaining why, so you never tap a button that can't work, and the same check greys out Record a voice note in the method chooser. It's a live capability check, not a browser blocklist. Typing and manual entry are always available.

  • Scan a card and Upload a photo — take or upload a photo of a business card. Dhaga runs OCR and also reads a LinkedIn QR code if one is on the card. You're not limited to one shot: add several photos of the same card or leaflet — front and back, or multiple pages — in a tray where you can reorder, crop, or remove any of them before scanning. On desktop you can also shoot straight from a live webcam, capturing multiple frames in one sitting. Every image you add merges into a single contact, and each photo is kept as its own source receipt. Where a card marks a number or address — Mobile, Direct, Office, a branch name — that marking is carried into the contact's phone and email labels, so you don't have to retype which is which.

    The two differ only in where they start you: Scan a card opens the live camera straight away (a card is in front of you now), and Upload a photo opens your photo library. Either way you land in the same tray, and Use live camera on the drop area switches to the webcam at any point.

Dhaga quick-add card-photo surface with a "Take or choose card photos" drop area (front, back, or multiple pages all merge into one contact) and a Use live camera button

AI extraction needs a key

The card scan, voice, and text extraction run through Claude. On a self-hosted enterprise deployment whose AI provider hasn't been configured, capture falls back to simpler heuristic parsing. See Self-hosting.

Prefer your phone?

You can also forward a contact card, note, or photo straight to Dhaga on WhatsApp or Telegram — no app to install. Cards still scan best in the app: the review dialog below is where Dhaga shows you what its two readings of the photo disagreed on, so you correct it before the contact is saved, while a forwarded card is saved first and reported back to you afterwards. Linking a chat also needs a Pro or Power plan, where scanning in the app works on every plan. See Messaging capture.

Review, disambiguate, and save

Extraction never saves silently. A review dialog shows the fields it pulled so you can correct them before the person lands in your graph:

  1. Check the name, title, company, emails, phones, and links.
  2. If the details look like someone you already have, Dhaga flags the match so you can merge into the existing contact instead of creating a duplicate.
  3. Optionally attach the person to an event right from the review dialog — handy when you're clearing a stack of cards from one conference. See Groups & events.
  4. Optionally write your first note about them in the same form.
  5. Save.

Write the note as you add them

The reason you're adding someone is usually the thing you'll forget first: where you met them, what they need, what you promised. Every form that creates a person carries a Note box, so you can write that down while you still have it — instead of saving, waiting for their page to load, and hunting for the notes section.

It's on all three create surfaces: the Add a person form, the review dialog after a paste or a card scan, and the full person form at /app/people/new. It is deliberately not on the edit form — a saved contact already has a notes box on its own page, and a second one there would only be a duplicate.

  • Your words are kept exactly as you typed them, and the note goes through the same extraction as one you add a minute later, so facts and follow-ups come out of it the same way. See Notes & facts.
  • Leave it empty and nothing is created — no note, and no AI credit spent.
  • Out of credits, the note is still saved. You lose the extracted facts and follow-ups, never the words. Add facts by hand, or re-run extraction on the note once the counter resets.
  • A scanned card gives you two notes, on purpose. One is the card's own text — Dhaga's record of what it read — and one is your line about the person. They stay separate rows you can edit and delete separately, because folding them together would put words you never wrote inside your own note.

The Dhaga Quick add page on the "Add a person" method, showing the whole form at once — Name and Location filled in, Experience, Education, Emails, Phones and Links sections, an expanded More details section holding an Event field, and a Note box carrying a couple of lines about where the person was met and what they asked for, directly above the Save person button

Attach them to an event

Under More details, both the quick-add page and the review dialog carry an Event field, so someone you met at a conference can be filed against it as you save. It's one searchable box, not a dropdown: start typing and the list narrows on the event's name or its city — "the Pune one" is how most people remember an event — and each row carries where and when it was, Pune · 12 Mar, so three meetups all called Monthly stay tellable apart.

Type a name that matches nothing and the same list offers Create event "…", so starting a new event needs no second field. Type the name of an event you already have and Dhaga attaches to that one rather than quietly minting a duplicate. None sits at the top of the list, because most captures belong to no event at all. On a phone it's an ordinary field you type into, so nothing covers the form while you choose.

The Dhaga Event field on a capture form with "de" typed into it, its dropdown open below showing None, then a Design Week row with the sublabel "Bengaluru · 5 Sept", then a "Create event 'de'" row, with the Save person button beneath

Adding someone manually

Three of the seven methods use no AI and no API key at all: Add a person types someone in by hand, Log a relationship records how two people you already have know each other, and Note a fact or follow-up logs something to remember or something to do against a contact. They used to be sub-tabs nested inside a Manual mode; they're three rows in the same list as everything else now. For the full person form with every field, go straight to /app/people/new.

The quick-add "Add a person" surface: a blank contact form under its "All methods" back link

The person form keeps Experience and Education as two separate sections:

  • Experience — roles with a title, company, dates, and a relationship type (employment, internship, founder, board member, advisor, volunteer).
  • Education — schooling with an institution, degree / programme, field of study, years, and a currently studying toggle. Education is stored as a "studied at" affiliation, so schools become nodes in your graph just like employers.

Add as many of each as you need, along with several labelled emails and phones and links, then save.

The manual person form's Experience and Education sections — Education showing Institution, Degree / programme, Field of study, year fields, a "Currently studying here" toggle, and a relationship-type selector

On this page