Bandhu — UX Reference

The user's journey, screen by screen

This is the companion, plain view to docs/pipeline.html — that one shows what the backend does with a check-in; this one shows what a person actually sees and taps. Screen names match the Stitch prototype 1:1, so you can check this against the real build directly.

Screen
Crisis path
Valid ending — screen feels finished, nothing forced next
Known mismatch with current Stitch build

The core loop

Left column is the spine every check-in follows. Right column is everything that can branch off the Response screen — most check-ins end without ever reaching any of them.

Welcome
"A companion, not a treatment tool."

First open only. Mascot, privacy line, one button: "Say hello." No sign-up, no intake form.

every later visit starts here instead ↓
Home
"Share how you're feeling…"

The text input is the one dominant thing on screen. Breathe / Write Together / Listen / Poem exist, but small and secondary — never competing with the input before the person has said anything.

person types, speaks, or taps a mood
(invisible) Safety check
Every message, silently

If crisis language is detected, everything below is skipped — jumps straight to Crisis Support, right column.

no crisis language →
Response
Acknowledgment, then a real companion conversation

"I hear you. That sounds like a lot to carry." Below the acknowledgment, at most one small, muted, easy-to-ignore line — never a two-button decision the person has to resolve. But the screen itself stays open: there's always a place to keep talking, turn after turn, like texting someone who remembers what you just said — not a single reply that ends the interaction.

nothing forces the person to keep going, but nothing cuts them off either — they leave when they're done, not when the screen decides they are
Crisis Support
Real numbers, zero cleverness

Vandrevala Foundation, iCall, Dial 112. Call Now / WhatsApp. Amber, not alarm-red. The one screen that's deliberately different in tone — on purpose.

hand-off to a real person, not managed by the app
Suggestion (inline, then full screen)
"Want something to try?"

Tapped only when eligible. Appears beneath the acknowledgment, not replacing it. Can lead into Immersive Breathing — full screen, one-tap exit, "stay as long as you need."

exit anytime, no completion required
Thinking Trap
Naming tool, not an AI guess

Opens only after the quiet "want to look at it together?" is accepted. The app doesn't diagnose a single pattern — it offers a small set of plain-language options so the person names what they're feeling. Whatever they pick becomes context for the RAG layer, and the conversation continues from there, tailored to that pattern.

Same Response screen, different content
Medical doubt · disorder question · medication question · life decision · dependency signal

None of these are separate screens. They render as the same acknowledgment + quiet-line pattern — some pull a fixed redirect, life-decisions and dependency signals are generated warm within a guardrail. Visually indistinguishable from any other Response.

"Let's look" / "Maybe later"
Explicit memory recall

"Do you remember what I told you last week?" gets a light reply, not a recap. "Let's look" leads into Looking Back; "Maybe later" is a genuine, accepted deferral.

Always reachable, not part of the check-in sequence

Sit behind the bottom nav / menu. A person can go here without ever checking in, and checking in never requires passing through them.

Looking Back
A summary of the week, first

Opens with one line summarizing the week — "Tuesday felt heavy — Thursday felt lighter" — then the daily timeline underneath as supporting detail, not the other way around. No charts, no streaks. Extends to reference whether a past suggestion actually helped.

Settings / Privacy
Language toggle, delete my data

Plain language, not buried in a menu tree.

Co-Create (Poem)
Additive, scope risk

On-brand, but README flags this as a decision still open: ship in v1, or wait for the core loop to be validated first.

Listen (Music)
Additive, scope risk

Same open question as Co-Create.

What makes this "light"

Every screen above has a valid ending

Not a fallback, not a degraded state — leaving the Response conversation at any point, Crisis Support, a finished breathing session, and Looking Back are all complete stopping points. The person is never left mid-task, and never made to feel like they cut something short by going quiet.

The branch list is longer than what most people ever see

Thinking Trap, the suggestion detail screens, and the special-content variants all exist, but the default path — Home → share → acknowledgment → maybe a few more turns back and forth — is the one most check-ins actually take. The branches are exceptions, not the main event.

Where this doesn't match the current Stitch build yet

The Thinking Trap screen's multi-pattern selection was flagged here in an earlier pass — that was a misread. It's not a mismatch: the selection UI is the intended design, not an AI-guessed single pattern.

Response screen

Currently shows an immediate two-button card ("Not right now" / "Yes, let's try") right under the acknowledgment — a decision the person has to resolve. Should be a single muted, ignorable line instead.

Home (Evolved)

Four suggestion buttons (Breathe / Write Together / Listen / Poem) sit prominent, right below the input, plus a persistent feature nav — reads as a menu before the person has said anything.