Baalam

Open Market: Multimodal AI Sales Advisor for Agricultural Businesses Small agricultural input distributors sell products that require technical guidance: the right product for the pest, the right dosage, and the right quantity. Most cannot afford to have an agronomist available for every customer inquiry. I built Open Market as an AI-powered sales advisor using Muse Wonder through the Meta Model API. Farmers can describe crop problems through text, photos, or voice notes using a public chat interface or WhatsApp. The agent searches the distributor's catalog, retrieves technical datasheets, recommends products based on documented evidence, provides the corresponding dosage, and builds a priced order through to payment instructions. When no datasheet supports a recommendation, it asks for clarification or escalates the case to a human agronomist rather than inventing an answer. What I built during the hackathon: I extended an existing AI sales advisor, migrating it to English and integrating Muse Wonder through the Meta Model API. I added multimodal photo and voice input, built the public chat interface, implemented model evaluation and benchmarking, developed an integration route, and created a test Meta integration. Existing foundation: I built on an earlier version of the agent and my catalejo agent library. The WhatsApp channel, chatroom, and human handoff functionality shown in the demonstration were already part of the project before the hackathon. Open Market is part of my broader vision: democratizing access to frontier AI so small businesses can automate customer-facing operations, serve more customers, and compete without needing specialized AI engineering teams.