Omi X Browser Use

• 

  browser use API running

• 

  Product comparison Skill response displayed on a clean UI

• 

• 

  2 suggested actions found

• 

  Planning a Trip skill response

• 

  College search based on user conversation

• 

  Researching an identity skill

• 

  Continuation of researching an identity skill

problem you tackled • On Omi’s app, a conversation summary is all you get - with the only actionable item being an addition of potential tasks to a to‑do list. Omi gave us insights into our conversation, but I tackled the gap between “insight” and “action.” I allow users to perform various tasks from within the Omi app based on their conversations using Browser Use.

approach and architecture • Summaries → Gemini selects relevant Browser Use skills (or none), filling params + missing fields. • One‑tap execution → Browser Use runs the web workflow; Omi renders templated results + keeps local history.

User Actions made possible through Browser Use

• Wikipedia place summary + attractions/travel info (place overview)
• Price comparison across stores (cheapest products list)
• Job search results (recent roles + salary/links)
• Nearby places finder (cafes/parks/shops with distance + contact)
• Deep research on a topic (summary + key findings + sources)
• Person research (instant bio, news, and web results)
• College research (top programs + deadlines + highlights)

tech stack • Flutter (Omi app) • Gemini API (intent → action selection) • Browser Use Skills API (web workflows as APIs)

Impact

• Omi shifts from “summary app” to “action engine”, turning spoken intent into real web actions on sites without APIs!!

Next steps: • Expand skill catalog (appointments, reservations, job applications, login based tasks). • Add per‑skill confidence gating and safety confirmations.

Built With

• flutter
• python