The advent of the Internet of Things (IoT) enables the development of new applications, context-aware services. However, for applications requiring a real-time consciousness of their environmental conditions, some additional mechanism is necessary. The paradigm of fog (or edge) computing is a promising candidate to meet this requirement by supporting the deployment at the network edge of entities for pre-processing the data produced by the IoT. Thus, the acquisition, filtering, processing (aggregation, fusion...) of contextual data can be performed locally in real-time within software entities deployed on equipments of a fog. As context is central in the targeted applications, qualifying context information becomes essential. Meta-data may therefore be added including some quality criteria such as precision, freshness, completeness, for measuring the Quality of Context (QoC) information. QoC management must take place throughout the whole processing chain of context information, impacting the operations performed within the entities of the fog. Facing the potential physical limitations of the equipments at the network edge, this paper promotes declarative programming of processing entities able to qualify context information, self-reconfiguration. Thus, the parameters controlling the transformation, qualification operations may be adjusted based on the observation of resource usage. A prototype shows that the reconfiguration time does not exceed one second, remains within acceptable limits for the targeted applications. This solution offers an alternative to the principle of offloading code advocated by some works on Fog Computing.