# Project export: CareLoop This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: UC Berkeley AI Hackathon 2026 - Tagline: CareLoop is a voice-first healthcare concierge that triages urgency, finds nearby providers, estimates cost, and helps book care from one calm conversation. - Devpost: https://devpost.com/software/careloop-gsiy8w - GitHub: https://github.com/VishalDani1602/AIHACK26.git - Result: winner (Best Use of The Agentverse by Fetch AI) - Team: 3 GitHub contributor(s) — Claude Opus 4.8 (28 commits), VishalDani1602 (28 commits), paria03 (14 commits) ## Devpost submission (written by the team) ### Inspiration CareLoop was inspired by how stressful healthcare navigation can be. Even when someone knows they need care, they still have to figure out urgency, insurance, cost, providers, and booking, often while anxious or helping a loved one. ### What it does CareLoop lets users describe a health concern by voice or text. It triages urgency, routes emergencies to 911, finds nearby providers, estimates out-of-pocket cost, and helps guide appointment booking in one conversation. ### How we built it We built CareLoop with Fetch.ai uAgents, ASI:One, Deepgram voice, FastAPI, Redis, Stripe, and CMS NPPES provider data. A main orchestrator agent coordinates specialist agents for triage, provider search, cost estimation, scheduling, and payment. ### Challenges we ran into The hardest part was making many moving pieces feel simple: voice input, AI reasoning, agent messaging, provider data, payments, and booking. We also had to build strong safety guardrails so emergency symptoms are handled immediately. ### Accomplishments we're proud of We are proud that CareLoop feels like a real healthcare access workflow, not just a chatbot. It combines voice, agents, real provider data, cost estimates, booking flow, fallback logic, and emergency routing into one usable demo. ### What we learned We learned that healthcare AI needs structure as much as intelligence. Clear agent roles, state management, safety rules, and honest limitations are what make the experience trustworthy. ### What's next Next, we would integrate live insurance eligibility, real appointment availability, provider booking APIs, multilingual caregiver workflows, and stronger clinical review so CareLoop can move from demo to dependable care navigation. ## README (from the GitHub repository) # CareLoop — a voice-first, multi-agent healthcare-access concierge > Speak a health problem in plain language and CareLoop's team of AI agents > **triages urgency → finds real in-network providers → estimates your > out-of-pocket cost → books an appointment** — then reads the confirmation > back to you by voice. Built for the **UC Berkeley AI Hackathon 2026**: **Fetch.ai** (uAgents + ASI:One) multi-agent orchestration, **Anthropic Claude** clinical triage, a **Deepgram** voice layer, **Redis** caching/state, and **Stripe** payments. Developed with **Claude Code**. ⚠️ **CareLoop is a navigation aid, not a medical provider.** It does not diagnose. Red-flag symptoms (chest pain, stroke signs, trouble breathing, etc.) are routed straight to **911** and never to a booking. --- ## Why it matters The hardest part of healthcare for many people — elderly patients, caregivers, non-native English speakers, anyone in distress — isn't the medicine, it's the **navigation**: *Who do I see? Is it urgent? Who's nearby and takes my insurance? What will it cost? How do I book it?* CareLoop collapses that maze into one calm conversation you can have **by voice, in your own language**. ## How it works (architecture) ``` ┌──────────────── ASI:One chat ────────────────┐ 🎙 Browser voice app + 📊 dashboard │ (Agent Chat Protocol — no frontend needed) │ (Deepgram STT/TTS, interactive cards) └───────────────────────┬──────────────────────┘ │ POST /voice (REST) ChatMessage │ │ ▼ ▼ ┌───────────────────────────── Orchestrator Agent ──────────────────────────────┐ │ ASI:One LLM: parse intent, manage the conversation, compose replies + cards │ └──┬──────────┬──────────┬──────────┬───────────┬──────────┬───────────────────────┘ │ send_and_receive (agent-to-agent messaging via Agentverse) ▼ ▼ ▼ ▼ ▼ ▼ ┌────────┐┌──────────┐┌────────┐┌──────────┐┌────────┐┌────────────┐ │ Triage ││ Provider ││ Cost ││ Scheduler││Payment ││ Evidence │ │ 911 ││CMS NPPES ││ est. ││ .ics + ││ Stripe ││ trials + │ │ rules ││ ││ ││ confirm ││ deposit││ drug safety│ └────────┘└──────────┘└────────┘└──────────┘└────────┘└────────────┘ ``` - **7 agents**, each registered on **Agentverse** with its own mailbox + profile. - The **Orchestrator** speaks the **Agent Chat Protocol**, so the *entire workflow runs from ASI:One with no custom frontend* — then the Deepgram web app layers a natural voice experience (and interactive cards) on top of the same agent. - **Real data:** providers from **CMS NPPES**, recruiting trials from **ClinicalTrials.gov**, drug-safety from **openFDA**. - **Takes action:** for serious/chronic conditions it surfaces relevant **clinical trials + drug-safety notes**, and it completes a real (test-mode) **Stripe deposit** before booking. - **Resilient:** every agent-to-agent call falls back to in-process logic, so a single agent hiccup never breaks the demo. - **Redis** is shared infrastructure across all agent processes: a provider/triage/ trials cache (huge repeat-latency savings), a session store, a **Streams audit trail**, and live stat counters — all degrading gracefully if Redis is offline. ## The agents | Agent | Address (derived from seed) | Role | |-------|------------------------------|------| | Orchestrator | `agent1qfxka5afzlk2wsp3agw77l7gz426d9vymwr9l4cfhy2gdvvhq2h8kn798uc` | ASI:One-facing brain; chat protocol + voice REST | | Triage | `agent1q29ewatrgfnskvkk020y3p2n7qkl36xw293v4x74kpy49czdjqxrqxl9xpe` | **Claude-powered** urgency + specialty reasoning; hard 911 red-flag rules | | Provider-Finder | `agent1qdjaj2ctxsjs3vpt5j7uxk46vq9znrquhn9q8fkljqhjueqrr93nzwj3whq` | Real providers from CMS NPPES (filtered to the requested city/state) | | Cost-Estimator | `agent1qtrge69qnynvpfwd7duw9pgjeffsmmlemflyz06v6v4r2yndtvcwq5k4sd6` | Plan-, provider- & region-aware out-of-pocket estimate | | Scheduler | `agent1q0d726utvqsgjmctt4etsclcapk9tvx75t7tlrcp0luyvsecyr7yz9ujgyf` | Confirmation + iCalendar invite | | Payment | `agent1qdr7s04hzndeefr2tt085nt29q8jxklf8hne9yhchn63huey9eurczs5ux5` | Stripe Checkout deposit + server-side verify | | Evidence | `agent1qt2nusghdan489nh92w37ax33gz8yxfuklezkph029eyju5dyt7vv7m2eag` | Recruiting clinical trials (ClinicalTrials.gov) + drug safety (openFDA) | Profile URL pattern: `https://agentverse.ai/agents/details/
/profile` (run `./venv/bin/python -m scripts.print_addresses` to reprint). ## Quick start ```bash # 1. Install python3 -m venv venv && ./venv/bin/pip install -r requirements.txt # 2. Configure keys cp .env.example .env # fill in ASI1_API_KEY, AGENTVERSE_API_KEY, DEEPGRAM_API_KEY # optional: STRIPE_SECRET_KEY (deposit), REDIS_URL # 3. One command: starts Redis + all 7 agents + voice app, registers, runs a full demo ./scripts/demo.sh # -> then open the voice app at http://127.0.0.1:8080 # and the live dashboard at http://127.0.0.1:8080/dashboard ```
…or run each step manually ```bash ./venv/bin/python -m scripts.selftest # sanity-check the pipeline (no keys/network needed) redis-server --daemonize yes # or: docker run -d --name careloop-redis -p 6379:6379 redis:7-alpine ./scripts/run_all.sh # start all 7 agents (each in its own process + mailbox) ./venv/bin/python -m scripts.register_agents # auto-register every mailbox on Agentverse (no browser clicks) ./venv/bin/uvicorn voice.backend:app --port 8080 ```
ASI:One promo code: `BERKELEYAI` · Agentverse promo code: `BERKELEYAIAV` ## Try it **Golden path** (books a real provider): ``` my dad has had a bad cough and a low fever for five days, he's on Medicare and we're in Berkeley ``` → "…I'd start with Primary Care … the best match near Berkeley is **Allison Aiken, M.D.** at 2222 Bancroft Way with an opening in 2 days … with a **provider- and plan-specific out-of-pocket estimate** … want me to book it?" → *"yes"* → **pay a refundable $25 deposit** via a real Stripe Checkout link (test card `4242 4242 4242 4242`) → *"done"* → agent verifies payment → ✅ booked + calendar invite. > If `STRIPE_SECRET_KEY` is not set, CareLoop skips the deposit and books directly. **Clinical-evidence path** (serious/chronic condition → real recruiting trials + drug safety): ``` my mom was diagnosed with early-stage breast cancer, she's in San Francisco on Medicare and takes tamoxifen ``` → routes to **Oncology**, names a real SF oncologist, and surfaces **recruiting clinical trials** near her (ClinicalTrials.gov) plus a **tamoxifen drug-interaction note** (openFDA) — options to discuss with her doctor. **Emergency path** (routes to 911, no booking): ``` I'm having crushing chest pain and I can't breathe ``` ## Tech - **Fetch.ai uAgents** — multi-agent framework, Agentverse mailboxes, Almanac discovery - **Agent Chat Protocol** — ASI:One discoverability & chat - **ASI:One** (`asi1-mini`) — orchestrator intent parsing + reply composition - **Anthropic Claude** (`claude-sonnet-4-6`) — triage clinical reasoning (falls back to ASI:One) - **Deepgram** — Nova-3 STT (multilingual) + Aura-2 TTS - **Stripe** — real (test-mode) Checkout for a refundable booking deposit, verified server-side - **Redis** (beyond caching) — shared provider/triage/trials cache (91×–853× faster on repeats), session store, **Streams audit trail** of every clinical/payment/booking decision, live stat counters - **CMS NPPES** — real provider data · **ClinicalTrials.gov** — recruiting trials · **openFDA** — drug safety - **FastAPI** — voice web app + interactive cards + live **/dashboard** ## Deliverables - 🔗 ASI:One shared chat: _add link here_ - 🤖 Agentve [README truncated for size] ## Detected evidence (automated analysis) Indexed codebase: 46 recognized source files, 225 KB. - Anthropic (technology) — detected in the code - CSS (language) — detected in the code - FastAPI (technology) — detected in the code - HTML (language) — detected in the code - JavaScript (language) — detected in the code - OpenAI (technology) — detected in the code - Python (language) — detected in the code - Redis (technology) — detected in the code - AI coding agent: Claude Code — evidence: config files committed to the repository; commit authorship or trailers ## Codebase structure (from repository index) ### Files (52 of 52) ``` .claude/scheduled_tasks.lock .env.example .gitignore agents/__init__.py agents/common/__init__.py agents/common/asi.py agents/common/claude_llm.py agents/common/clinical.py agents/common/config.py agents/common/logic.py agents/common/models.py agents/common/nppes.py agents/common/orchestration.py agents/common/prompts.py agents/common/store.py agents/common/stripe_pay.py agents/cost.py agents/evidence.py agents/orchestrator.py agents/payment.py agents/provider_finder.py agents/readmes/cost.md agents/readmes/evidence.md agents/readmes/orchestrator.md agents/readmes/payment.md agents/readmes/provider_finder.md agents/readmes/scheduler.md agents/readmes/triage.md agents/scheduler.py agents/triage.py data/cost_table.json data/taxonomy_map.json DEVPOST.md docs/UIUX_PLAN.md PRESENTATION.md README.md requirements.txt scripts/__init__.py scripts/chat_client.py scripts/demo.sh scripts/print_addresses.py scripts/register_agents.py scripts/run_all.sh scripts/selftest.py scripts/showcase.py voice/__init__.py voice/backend.py voice/static/app.js voice/static/dashboard.html voice/static/dashboard.js voice/static/index.html voice/static/style.css ``` ### Dependencies - requirements.txt: anthropic@>=0.40.0, deepgram-sdk@>=3.5.0, fastapi@>=0.110.0, openai@>=1.30.0, python-dotenv@>=1.0.0, redis@>=5.0.0, requests@>=2.31.0, stripe@>=9.0.0, uagents@>=0.22.0, uagents-core@>=0.1.0, uvicorn[standard]@>=0.29.0, websockets@>=12.0 ### Recent commits (newest first) - Pitch: full word-for-word 2-person script + 60s version - Add pitch script (3-4 min, 2 presenters) + judge Q&A - Stripe: auto-continue original tab after payment (no dead-end /?paid=1) - Dashboard: live architecture/explanation board + triage-via-Claude indicator - Take insurance plan as explicit input (+ ask for it conversationally) - docs: README reflects Claude triage, location-filtered providers, provider/region-aware cost - Add Claude-powered triage (Anthropic) with graceful fallback to ASI:One - Cost: vary estimate by provider + region (no more flat $108-$216) - Add dashboard and demo polish - Add multilingual reply flourish - docs: update README to 7 agents (+ Evidence) and latest features - Add persisted dark mode - Refresh CareLoop identity polish - Backend: structured card + actions in voice response (interactive cards) - Polish frontend micro interactions - Add frontend accessibility controls - Fix location and provider action handling - Add interactive care cards - Add voice feedback and thinking state - Add first run example prompts ## Key source files (fetched from GitHub, selected and truncated for size) ### DEVPOST.md ```markdown # CareLoop — Devpost submission _Voice-first, multi-agent healthcare-access concierge. Fetch.ai uAgents + ASI:One + Deepgram._ ## Inspiration For millions of people the hardest part of healthcare isn't the medicine — it's the **navigation**. An elderly parent, a caregiver juggling work, someone who speaks English as a second language, or anyone scared in the moment all hit the same wall: *Who do I see? Is this urgent? Who's nearby and takes my insurance? What will it cost? How do I actually book it?* That maze is where people give up, delay care, or end up in the ER. CareLoop turns it into one calm conversation you can have **out loud, in your own language**. ## What it does You describe a health concern in plain language (by **voice** or text). CareLoop: 1. **Triages** urgency and the right kind of clinician — with hard-coded red-flag rules that route emergencies straight to **911**, never to a booking. 2. **Finds real providers** near you via the public **CMS NPPES** registry. 3. **Estimates your out-of-pocket cost** based on your insurance type. 4. **Takes a refundable deposit** via a real **Stripe** Checkout (test mode) and verifies the payment server-side before confirming. 5. **Books the appointment** and produces a calendar invite — then reads the confirmation back to you by voice. ## How we built it - **Fetch.ai uAgents**: six agents — an **Orchestrator** plus **Triage**, **Provider-Finder**, **Cost-Estimator**, **Scheduler**, and **Payment** specialists — each registered on **Agentverse** with its own mailbox and profile. - **Stripe agent transaction**: the Payment agent creates a real (test-mode) Checkout session and verifies `payment_status == "paid"` server-side before the Scheduler confirms — a genuine intent→action transaction, not a mock. - **Redis (beyond caching)**: shared infrastructure across every agent process — a provider/triage cache (91×–853× faster on repeats), a session store, a **Streams audit trail** of every clinical/payment/booking decision (a real healthcare audit log), and live stat counters surfaced in the UI. Degrades gracefully if offline. - **Agent Chat Protocol + ASI:One**: the Orchestrator is discoverable and fully usable from ASI:One, so the **entire workflow runs in ASI:One chat with no custom frontend**. Agents coordinate through real agent-to-agent messaging (`send_and_receive`). - **ASI:One LLM** (`asi1-mini`) parses intent and drives the multi-turn conversation. - **Anthropic Claude** (`claude-sonnet-4-6`) powers the triage agent's clinical reasoning (urgency + specialty), with graceful fallback to ASI:One — so the build is on Claude Code *and* uses the Claude API in the product. - **Deepgram**: Nova-3 multilingual STT + Aura-2 TTS power a click-to-talk web app layered on top of the same Orchestrator — voice is the primary, essential interface. - **CMS NPPES NPI Registry** for real, verifiable provider data. - Built with **Claude Code**. ## Target users Caregivers [truncated — 3104 more characters] ``` ### PRESENTATION.md ```markdown # CareLoop — Pitch script (3–4 min) + Judge Q&A **Roles:** **A = product/host** · **B = technical** (drives the live demo). **Setup:** voice app open at `http://127.0.0.1:8080`; dashboard open in a 2nd tab at `/dashboard`; mic on. Pre-run one golden query so the cache is warm. Lead with the **local voice app** (network can't break it); show ASI:One only if Wi-Fi is solid. Stage directions are in [brackets]; everything else is spoken word-for-word. --- ## FULL SCRIPT (read aloud) **A:** Quick show of hands — has anyone here ever had to find a doctor for a sick parent or grandparent? Right. Now picture doing that at seventy years old, in your second language, while you're scared. Here's the thing we kept coming back to: the hard part of healthcare usually isn't the medicine — it's the **navigation**. Who do I even see? Is this urgent? Who's nearby, who takes my insurance, what will it cost, and how do I actually book it? That maze is where people give up, delay care, or just show up at the ER. **A:** So we built **CareLoop** — a voice-first *team* of AI agents that takes you from a spoken symptom all the way to a booked appointment. It doesn't just chat. It acts. [B] is going to show you, live. **B:** Thanks. Everything here is real and running. I'll just talk to it. [tap mic] *"My dad's had a bad cough and a fever for five days. He's on Medicare, and we're in San Francisco."* [release] **B:** [as the card appears] So in a couple of seconds: Deepgram transcribed that; our triage agent — running on Claude — decided this is primary care and not an emergency; and we pulled a **real** San Francisco doctor from the federal provider registry, with a cost estimate for his Medicare plan. It comes back as a card with a **Book** button. **B:** I'll book it. [click Book] That opens a real **Stripe** checkout for a small refundable deposit. I'll pay… [enter test card 4242…] …and watch the original screen — it just **continued on its own**. We're booked, with a calendar invite. No dead end, no copy-pasting a confirmation. **A:** And it goes deeper when it needs to. **B:** [tap mic] *"My mom was just diagnosed with early-stage breast cancer."* [release] **A:** For a serious condition, CareLoop surfaces **real, currently-recruiting clinical trials** near her from ClinicalTrials.gov, plus **drug-safety** information from the FDA — options she can take to her doctor. A normal booking app doesn't do that. **B:** And safety is built in. [tap mic] *"I'm having crushing chest pain and I can't breathe."* [release; red banner shows] Red-flag symptoms never go to booking — it routes straight to **911**, before any model even runs. **B:** One last thing. [switch to dashboard tab] This is our live system board: **seven agents**, what each one does, who talks to whom, and real-time stats. You can see triage runs on **Claude**, and the orchestrator on **ASI:One**. **B:** Architecturally: seven **uAgents on Fetch.ai's Agentverse**, fully usable from **ASI:One wi [truncated — 5975 more characters] ``` ### requirements.txt ``` # CareLoop — voice-first multi-agent healthcare-access concierge # Fetch.ai uAgents + ASI:One + Deepgram voice uagents>=0.22.0 uagents-core>=0.1.0 openai>=1.30.0 # OpenAI-compatible client, pointed at ASI:One (api.asi1.ai/v1) deepgram-sdk>=3.5.0 # STT (Nova-3) + TTS (Aura-2) fastapi>=0.110.0 # voice web app backend uvicorn[standard]>=0.29.0 requests>=2.31.0 # NPPES provider lookup python-dotenv>=1.0.0 websockets>=12.0 stripe>=9.0.0 # refundable booking-deposit flow (test mode) redis>=5.0.0 # shared cache + session store + audit stream + live stats anthropic>=0.40.0 # optional: Claude powers triage clinical reasoning ``` ### voice/static/app.js ```javascript // CareLoop voice web app — click-to-talk, record-then-send. const chat = document.getElementById("chat"); const micBtn = document.getElementById("mic"); const micLabel = micBtn.querySelector(".mic-label"); const statusEl = document.getElementById("status"); const srUpdates = document.getElementById("srUpdates"); const viaEl = document.getElementById("via"); const langSel = document.getElementById("lang"); const player = document.getElementById("player"); const textInput = document.getElementById("textInput"); const sendBtn = document.getElementById("sendBtn"); const composer = document.querySelector(".composer"); const meterBars = Array.from(document.querySelectorAll(".voice-meter span")); const largeTextToggle = document.getElementById("largeTextToggle"); const contrastToggle = document.getElementById("contrastToggle"); const darkModeToggle = document.getElementById("darkModeToggle"); const reduceMotionQuery = window.matchMedia("(prefers-reduced-motion: reduce)"); let sessionId = "web-" + Math.random().toString(36).slice(2, 10); let mediaRecorder = null; let chunks = []; let recording = false; let busy = false; let emergencyFocusTimer = null; let audioContext = null; let analyser = null; let meterFrame = null; let meterData = null; let lastDetectedLanguage = null; const DISPLAY_PREF_KEY = "careloopDisplayPrefs"; const displayPrefs = readDisplayPrefs(); const LANGUAGE_LABELS = { ar: "Arabic", en: "English", es: "Spanish", fr: "French", he: "Hebrew", hi: "Hindi", zh: "Chinese", }; const RTL_TEXT = /[\u0590-\u08ff]/; const EXAMPLE_PROMPTS = [ "My dad has a bad cough and fever", "Find a cardiologist near me", "My mom was diagnosed with diabetes", "I have chest pain", ]; function addMessage(role, text, opts = {}) { const wrap = document.createElement("div"); const language = opts.language || null; const isRtl = RTL_TEXT.test(text) || (role === "user" && language?.dir === "rtl"); wrap.className = "msg " + role + (opts.emergency ? " emergency" : "") + (isRtl ? " rtl" : ""); const bubble = document.createElement("div"); bubble.className = "bubble" + (opts.rich ? " rich-bubble" : ""); if (language?.code) bubble.setAttribute("lang", language.code); if (isRtl) bubble.setAttribute("dir", "rtl"); bubble.innerHTML = opts.emergency ? renderEmergency(text, language) : renderMessageBody(text, opts); wrap.appendChild(bubble); chat.appendChild(wrap); scrollChatToEnd(); return bubble; } function addThinkingBubble() { const wrap = document.createElement("div"); wrap.className = "msg bot thinking"; wrap.innerHTML = `
CareLoop is thinking
`; chat.appendChild(wrap); scrollChatToEnd(); return wrap; } // Minimal markdown: **bold**, clickable links, and newlines. function render(t) { return escapeHtml(t) .replace(/\*\*(.+?)\*\*/g, "$1") .replace(/(https?:\/\/[^\s<]+)/g, '$1') .replace(/\n/g, "
"); } function renderMessageBody(text, opts = {}) { const languageHtml = opts.language ? renderLanguageChip(opts.language) : ""; const cardsHtml = opts.cards?.length ? `
${opts.cards.map(renderCard).join("")}
` : ""; const actionsHtml = opts.actions?.length ? renderActions(opts.actions) : ""; if (!cardsHtml && !actionsHtml) return `${languageHtml}${render(text)}`; const summary = cardsHtml ? `
Conversation note${render(text)}
` : render(text); return `${languageHtml}${summary}${cardsHtml}${actionsHtml}`; } function setStatus(s) { statusEl.textContent = s; } function scrollChatToEnd() { requestAnimationFrame(() => { chat.scrollTo({ top: chat.scrollHeight, behavior: reduceMotionQuery.matches ? "auto" : "smooth", }); }); } function playAudio(b64) { if (!b64) return; player.src = "data:audio/mpeg;base64," + b64; player.play().catch(() => {}); } async function sendTurn(promise, opts = {}) { if (busy) return; busy = true; micBtn.classList.add("disabled"); if (opts.userText) { lastDetectedLanguage = detectLanguage(opts.userText, langSel.value); addMessage("user", opts.userText, { language: lastDetectedLanguage }); } const thinking = addThinkingBubble(); try { const res = await promise; const data = await res.json(); thinking.remove(); if (data.error) { setStatus("Warning: " + data.error); return; } if (data.transcript && data.transcript !== opts.userText) { lastDetectedLanguage = detectLanguage(data.transcript, langSel.value); addMessage("user", data.transcript, { language: lastDetectedLanguage }); } addBotReply(data); setEmergencyFocus(Boolean(data.emergency)); viaEl.textContent = data.via ? "routed via: " + data.via : ""; playAudio(data.audio); setStatus(""); } catch (e) { thinking.remove(); setStatus("Network error: " + e.message); } finally { busy = false; micBtn.classList.remove("disabled"); } } function addBotReply(data) { announceBotReply(data.reply); const language = data.language ? normalizeLanguage(data.language) : lastDetectedLanguage; if (data.emergency) { addMessage("bot", data.reply, { emergency: true, language }); return; } const cards = collectCards(data); const actions = normalizeActions(data.actions) || inferActions(data, cards); addMessage("bot", data.reply, { cards, actions, rich: Boolean(cards.length || actions.length), language }); } function renderEmergency(text, language) { return `