Virtual Assistant

Virtual Assistant (Chatbot) for Evidence-Based Decision-Making with Citizen Science Data

The Virtual Assistant is an AI-powered multilingual chatbot designed to facilitate the retrieval and processing of biodiversity information. It integrates multiple data sources, including observations from the MINKA citizen observatory, taxonomic databases from the web, and scientific articles stored in a vector-based retrieval system. By leveraging natural language processing and AI-driven querying, the assistant provides users with both general and highly specific insights in an interactive conversational format. The available languages are English, Spanish and Catalan.

Building the multitasking agent in Python, it extracts biodiversity observations directly from MINKA’s API, structuring and filtering the data for easy interpretation. Additionally, it cross-references taxonomic classifications with external sources, ensuring accurate species identification and contextual information. Through its integration with a vector database, it enables efficient semantic search across loaded scientific documents, allowing users to access relevant research findings quickly. Moreover, the chats are saved anonymously in a SQL database, allowing continuous improvement of the tool and the project.

Designed to streamline information retrieval, the Virtual Assistant enhances accessibility to biodiversity knowledge without requiring technical expertise. Users can ask questions and receive structured responses in the form of text, tables, or lists. It also enables exploration of biodiversity records, taxonomic data, and project evolution summaries while facilitating scientific reports and providing AI-driven reasoning.

Basic Technologies Used

As mentioned above Pl@ntNet underwent three main core modifications and improvements as part of the GUARDEN project:

Deployment

Deploy using Docker, Task scheduling using Cron

AI Agent base

OpenAI API, LangChaing & LangGraph Python module (multitask)

Database Management

Milvus (document vector-store), SQLAlchemy Python module (chat memory), LangChain (extra tools)

Backend & Data Processing

Python, Minka API, mecoda-minka Python module

Frontend

Streamlit, CSS, HTML

Data Access & Formats

URL, PDF (documents), XML (chat database)

Usefulness in GUARDEN Case Studies

Spanish case study: BioPlatgesMet a citizen science project for urban beaches management

The Virtual Assistant plays a key role in the Spanish case study of BioPlatgesMet, a citizen science initiative focused on urban beach management. By integrating AI-powered interactions with biodiversity observations from the MINKA citizen observatory, the assistant enhances accessibility to critical environmental data, making it easier for stakeholders to retrieve relevant insights.

As part of the GUARDEN project, the Virtual Assistant is connected to MINKA’s data ecosystem, allowing users—including citizen scientists, researchers, and policymakers—to access taxonomic data, project evolution summaries, and scientific reports through a conversational interface. This tool enables stakeholders to query biodiversity records, extract meaningful information, and interpret findings without requiring technical expertise.

The assistant’s ability to process and retrieve species observations, scientific articles, and taxonomic classifications supports the decision-making process in urban beach management. By providing structured responses in text, tables, and lists, it facilitates rapid information exchange for the Metropolitan Area of Barcelona (AMB), the primary stakeholder. This is particularly valuable for planning activities such as beach cleaning initiatives, invasive species monitoring, and biodiversity conservation efforts.

Moreover, the anonymous logging of user interactions in an SQL database allows for continuous refinement of the assistant’s capabilities, ensuring it evolves to meet project needs. The integration with MINKA’s broader public participation tools, including dashboard and forum, strengthens collaborative engagement by enabling knowledge-sharing and fostering discussions around key environmental challenges.

Links

Credentials: For the moment is private for administrators, ask to the contact below if needed

Virtual Assistant