Large Language Models (LLMs) excel at synthesising globally documented knowledge but lack the finegrained, real-time awareness required for field-level agricultural and rural-planning decisions.This paper introduces Jack-Daw, a spatially enabled chat-agent architecture that couples foundation-model reasoning with multi-modal geospatial data streams and a retrieval-augmented generation (RAG) pipeline.JackDaw implements a tool-prefiltering mechanism that selects only those data connectors whose topical, temporal and spatial metadata match the current query, thereby mitigating the diminishing returns observed when LLMs are exposed to large, flat toolsets.Through LangChain-based orchestration the platform dynamically assembles workflows that range from lightweight natural-language processing models to domain-specific analytic kernels, while a value-engineering strategy allocates computationally intensive models (e.g., GPT-4-class) only to tasks that require broad contextual reasoning.Benchmark experiments on forestry-asset discovery and vineyard-site assessment demonstrate that JackDaw delivers location-specific, traceable answers that outperform a standalone proprietary LLM, which provides only generic or spatially misattributed responses.The results confirm that bridging global language models with local spatial intelligence markedly reduces hallucination rates and enhances the operational readiness of AI for sustainable agriculture and rural development.

Efficient crop monitoring and crop dynamics forecasting leveraging diverse satellite and point data are described.UnCRtainTS neural network architecture is utilized for cloud removal in satellite imagery which overcomes an issue in crop monitoring.Combining optical (Sentinel-2) and radar (Sentinel-1) satellite data improves the robustness and accuracy of the model in terms of satellite image reconstruction and vegetation index estimation.However, available soil-type geographical data and land surface analysis products, do not improve prediction accuracy significantly.

Analyzing massive code bases is a staple of modern software engineering research – a welcome side-effect of the advent of large-scale software repositories such as GitHub. Selecting which projects one should analyze is a labor-intensive process, and a process that can lead to biased results if the selection is not representative of the population of interest. One issue faced by researchers is that the interface exposed by software repositories only allows the most basic of queries. CodeDJ is an infrastructure for querying repositories composed of a persistent datastore, constantly updated with data acquired from GitHub, and an in-memory database with a Rust query interface. CodeDJ supports reproducibility, historical queries are answered deterministically using past states of the datastore; thus researchers can reproduce published results. To illustrate the benefits of CodeDJ, we identify biases in the data of a published study and, by repeating the analysis with new data, we demonstrate that the study’s conclusions were sensitive to the choice of projects.