Fixed Semantic Search leaving out passages that did not fit into a search (#1167)

app/Tests/Tools/ToolCalling/SemanticSearchToolRequestTests.cs

@@ -11,8 +11,8 @@ namespace AIStudio.Tests.Tools.ToolCalling;
 /// </summary>
 /// <remarks>
 /// The model may only search the data sources offered to it, and it may only turn pages where
-/// that means something: in one data source at a time, and not beyond the window the data sources
-/// fetch at most. Every refusal says what would have been right, so that the model can correct
+/// that means something: in one data source at a time, and not beyond what a data source can
+/// return at all. Every refusal says what would have been right, so that the model can correct
 /// itself with its next call.
 /// </remarks>
 [TestFixture]
@@ -95,17 +95,24 @@ public sealed class SemanticSearchToolRequestTests
     }
 
     [Test]
-    public void APageBeyondTheWindowIsRefusedWithTheLastPage()
+    public void APageBeyondWhatADataSourceCanReturnIsRefused()
     {
-        var message = Refusal(() => SemanticSearchTool.ReadRequest(Arguments($$"""{"query":"travel expenses","data_source_ids":["{{HANDBOOK.Id}}"],"page":10}"""), OFFERED));
-        Assert.That(message, Does.Contain("at most 9").And.Contain("but was 10").And.Contain("Rephrase the query"));
+        // Nine pages of the data source with ten matches each, and at least one passage per page of the search:
+        var message = Refusal(() => SemanticSearchTool.ReadRequest(Arguments($$"""{"query":"travel expenses","data_source_ids":["{{HANDBOOK.Id}}"],"page":91}"""), OFFERED));
+        Assert.That(message, Does.Contain("at most 90").And.Contain("but was 91").And.Contain("Rephrase the query"));
     }
 
     [Test]
-    public void AnERIDataSourceHasNoFurtherPage()
+    public void AnERIDataSourcePagesWithinItsMatches()
     {
-        var message = Refusal(() => SemanticSearchTool.ReadRequest(Arguments($$"""{"query":"travel expenses","data_source_ids":["{{INTRANET.Id}}"],"page":2}"""), OFFERED));
-        Assert.That(message, Does.Contain("at most 1").And.Contain("but was 2").And.Contain("Rephrase the query"), "A server which allows only the configured number of matches would refuse the page, and the data source would count as not searched.");
+        var request = SemanticSearchTool.ReadRequest(Arguments($$"""{"query":"travel expenses","data_source_ids":["{{INTRANET.Id}}"],"page":10}"""), OFFERED);
+        var message = Refusal(() => SemanticSearchTool.ReadRequest(Arguments($$"""{"query":"travel expenses","data_source_ids":["{{INTRANET.Id}}"],"page":11}"""), OFFERED));
+
+        using (Assert.EnterMultipleScope())
+        {
+            Assert.That(request.Page, Is.EqualTo(10), "The matches the server returns may fill several pages, one of them each at the least.");
+            Assert.That(message, Does.Contain("at most 10").And.Contain("but was 11"), "ERI v1 knows no offset, so there is no page beyond the configured number of matches.");
+        }
     }
 
     private static JsonElement Arguments(string json) => JsonSerializer.Deserialize<JsonElement>(json);