prediction: docs++

src/prediction.rs

@@ -197,6 +197,10 @@ impl Session {
                         .parameters(schema_for!(StageEventArgs))
                         .build().unwrap()
                 }),
+                // TODO: There should only be two queries, one against the ship's onboard archive, and another against the relay network, or whatever we call it. Both should be structured with the same arguments schema
+                // TODO: A relay search should try to grab first from beets, then musicbrainz, then from bandcamp.
+                // TODO: A query should specify what parts of metadata are sufficient for the result, so we don't always have to hit all the layers of data. beets can of course, ignore this.
+                // TODO: A query should be hierarchical somehow? eg, "I already know about artist X, but I want to know everything about track Y from album Z" or "I don't know anything about artist X/album Y, please give me an overview"
                 ChatCompletionTools::Function(ChatCompletionTool {
                     function: FunctionObjectArgs::default()
                         .name("archive_query")