BodyStation

BodyStation Devpost submission draft

Inspiration

Health advice piles up fast.

I had PT notes, gait findings, nutrition reminders, recovery cues, and random voice memos sitting in my Health wiki. The advice was useful, but I was not going to open a dashboard ten times a day to remember it.

So BodyStation started with a simpler question: what if my health notes could talk at the exact moment I needed them?

What it does

BodyStation turns a personal Health wiki into a private radio station for your body.

It reads trusted notes, picks conservative micro-actions, and plays them as short ElevenLabs voice cues across the day. Hydration at lunch. Mobility before training. Recovery in the afternoon. A calm-down track at night.

The demo also has a generated station. I can ask, "I want to run today but my calf hurts. What should I do?" BodyStation treats pain as a constraint, checks the notes, lowers the intensity, and creates a safer station for the day.

It does not diagnose. It does not replace a clinician. It turns health advice I already trust into audio I can actually follow.

How we built it

We built BodyStation as a deterministic hackathon demo around one polished loop: Health wiki in, BodyStation out.

The source material comes from local Health wiki notes: PT mobility cues, running gait recommendations, nutrition reminders, and recovery guidance. The interface is a single HTML demo with station controls, a real clock, a demo time machine, Health wiki source cards, and an agent trace for the generated station.

ElevenLabs powers the voice cues and relaxation audio. For the video, the audio is pre-generated so the demo is reliable. The agent loop is still visible: parse the goal, detect the constraint, retrieve relevant notes, generate the daily agenda, then play the station.

Challenges we ran into

The hardest part was making the product feel agentic without making the demo fragile.

Live generation is exciting until it fails during a three-minute video. So we made the demo honest: the audio is pre-generated for reliability, but the screen shows the loop the product would run in production.

The second challenge was health safety. It is easy to accidentally sound like a medical assistant. BodyStation is not that. The demo had to show constraint handling without overclaiming: calf pain means lower intensity, stop conditions, and conservative reminders. Not diagnosis.

Accomplishments that we're proud of

BodyStation feels different from another health dashboard.

The strongest moment is the calf-pain station. The user asks for a running ramp-up, mentions pain, and the system does not push harder. It adapts. It creates a safer audio plan: no hard run, hydration check, mobility work, light hinge drills, and stop conditions.

We are also proud that ElevenLabs is central to the product. Voice is not decoration here. The whole product is built around listening instead of staring at another screen.

What we learned

The best health interface might not be a dashboard.

A lot of health data is already there: notes, plans, reminders, voice memos, checklists. The missing piece is timing. If the advice arrives when I can act on it, it becomes useful again.

We also learned that agent demos need visible honesty. Judges should see what is real, what is simulated for reliability, and what the production loop would automate.

What's next for BodyStation

Next, BodyStation should generate the station live each morning from the latest Health wiki updates.

After that:

• voice queries for new stations
• wearable and calendar context
• better safety boundaries for pain and fatigue
• station history and adherence logs
• exportable summaries for a coach, PT, or doctor

The long-term idea is simple: your body has a station now. You should be able to press play on the plan.

Built With

• elevenlabs
• hermes
• html