What We Learned Building a Chatbot for La Silla Vacía's Newsroom Archive (gijn.org)

La Silla Vacía has successfully launched a community chatbot, SillaIA, utilizing its extensive news archive to enhance user engagement and streamline information access. Launched on June 24, 2025, the chatbot's development was profoundly influenced by previous experiences with AI models, particularly the unpredictable nature of differing outputs as models evolve. This early experimentation revealed the challenge of maintaining stable performance, prompting the team to approach the project with a focus on community interaction and iterative improvements. The architecture of SillaIA is modular, incorporating semantic search and context awareness to handle user queries effectively. Initial versions struggled with literal matches and understanding common user input variations, leading to noteworthy user-driven insights on handling spelling errors and alternative phrasing. As users began to ask questions about the organization itself, the team recognized the need for continual adaptation and feature expansion beyond just news-related inquiries. This journey reflects crucial lessons for the AI/ML community, emphasizing the importance of responsiveness to user needs and the necessity of robust frameworks for evolving AI capabilities in journalistic settings.