Once partial book drafting is enabled in SF, we would like to track how it is being used. We cannot rely on the chapter filters specified in the build metadata, since the data that a project actually contains might differ from the specified filter. The best way to track partial book drafting is to use the computed results from preprocessing. Currently, the train and inference counts are recorded for each build. A finer granularity of counts should provide sufficient information to track partial book drafting usage. The proposed approach is to record the train/inference counts at book and chapter levels. A simple chart might show the percentage of drafted books that are partial drafts. Another possible chart could show the most common book portions that are drafted, e.g. MAT 1-10.
Once partial book drafting is enabled in SF, we would like to track how it is being used. We cannot rely on the chapter filters specified in the build metadata, since the data that a project actually contains might differ from the specified filter. The best way to track partial book drafting is to use the computed results from preprocessing. Currently, the train and inference counts are recorded for each build. A finer granularity of counts should provide sufficient information to track partial book drafting usage. The proposed approach is to record the train/inference counts at book and chapter levels. A simple chart might show the percentage of drafted books that are partial drafts. Another possible chart could show the most common book portions that are drafted, e.g. MAT 1-10.