Wardly Intake Review
A Next.js 16 app where a patient chats or speaks with an AI intake assistant while a live clinical chart assembles itself through slot-filling tool calls.
Each review covers deployability, value versus commercial alternatives, and privacy model. Tools that can run locally were started and exercised; mobile or backend-dependent tools were assessed from published builds, source code, and deploy guides. Ratings reflect what we were able to verify.
Wardly Intake — Pre-Visit Clinical Intake Agent
A Next.js 16 app where a patient chats or speaks with an AI intake assistant while a live clinical chart assembles itself through slot-filling tool calls. Last updated: 2026-06-21.
One-sentence verdict: A well-architected take-home demo with an impressive slot-filling approach and a live clinician chart, but it remains a prototype with in-memory state and no authentication.
What the System Is
Wardly Intake is a two-pane web application for pre-visit clinical intake. The left pane is a chat or voice conversation with an empathetic AI assistant; the right pane shows a clinical chart that fills in real time as the agent records facts via typed tool calls. At the end of the encounter, a second model synthesizes a structured clinical brief containing the chief complaint, HPI in OLDCARTS format, targeted review of systems, red flags, pertinent negatives, and a recommended triage level.
The deploy guide reports the service was started on port 5008.
| Key data | |
|---|---|
| Category | Clinical Intake |
| Language | TypeScript / Next.js |
| License | Proprietary / All rights reserved (take-home submission) |
| Self-hosted | Yes |
| AI provider | Groq (OpenAI GPT-OSS 20B chat, Llama 4 Scout synthesis); Gemini fallback |
| Database | None; in-process globalThis session store |
| Auth | None |
| Tests | Typecheck + eval harness with 3 synthetic patient personas |
How to Install and Deploy
cd /data2/docker/going-global/repos/wardly-intake
pnpm install
cp .env.local.example .env.local
# Add GROQ_API_KEY (or GOOGLE_GENERATIVE_AI_API_KEY / OLLAMA_BASE_URL)
PORT=5008 pnpm dev
Then open http://localhost:5008.
How to Test
The documented test flow is:
- Open
http://localhost:5008and click Begin intake. - Describe a symptom (or use a seeded scenario such as chest pain).
- Watch the right-hand chart populate with CC, HPI / OLDCARTS, ROS, and red flags.
- Continue until the intake-complete indicator appears, then click Generate brief.
- Review the markdown brief and the raw JSON structure.
- In a second terminal, run
pnpm evalto replay the 3 built-in personas and assert on captured slots.
Privacy & Compliance
Wardly Intake is not HIPAA compliant. It has no authentication, no persistent database, and no audit logging. Conversation data is held in an in-memory globalThis Map that survives hot-module reloads but is wiped on server restart. The README lists "PHI handling + HIPAA-ready posture" as a future, non-trivial effort. Do not use it for real patient data.
Wardly Intake vs Commercial Intake Tools
| Dimension | Wardly Intake | Commercial Tools (e.g., Notable, Ambience Healthcare, Nabla) |
|---|---|---|
| Cost | Free to run locally + API usage | Enterprise subscription |
| Output format | Structured markdown + JSON brief | EHR-integrated notes, billing codes |
| Slot-filling | Typed tools mutate live state; brief generated from state | Proprietary NLP pipelines |
| CC-driven ROS | Selects 2–4 systems based on chief complaint | Often broader or template-based |
| Persistence | In-memory only | Persistent cloud records |
| Auth / HIPAA | None | BAA, SSO, audit logs |
| Voice | Browser Web Speech API | Often telephony-integrated |
| Customization | Full source; prompt is plain English | Vendor-controlled |
Who Should Use It
- Developers studying slot-filling tool patterns for clinical intake.
- Teams prototyping a two-pane intake UI with live chart assembly.
- Interview reviewers evaluating the original Wardly take-home submission.
Who Shouldn't Use It
- Clinics needing authenticated, persistent, or HIPAA-compliant intake.
- Users who want phone-based intake without browser speech support.
- Organizations that require EHR integration or billing-code output.
FAQ
Is Wardly Intake open source?
The source is published for review, but the license is all-rights-reserved and prohibits redistribution, modification, commercial use, or training ML models. It is not an open-source license in the OSI sense.
Does it require a paid API key?
The default provider is Groq, which offers a free tier. Google Gemini or a local Ollama instance can be used instead with no code changes.
Is the structured brief reliable?
The brief is generated from the structured intake state rather than the raw transcript, which reduces hallucination risk, but it is still an AI-generated summary without clinical validation.
Verdict
Wardly Intake is a strong prototype for a clinical-intake agent. Its slot-filling architecture, CC-driven ROS selection, and live chart UI show careful product thinking. The main barriers to real-world use are the lack of persistence, authentication, and HIPAA controls. It is best treated as a design reference and scaffold.
Ratings: Deployability 4/5 · Value vs Commercial 3/5 · Privacy Compliance 2/5