Improved the errors models get from web search for invalid arguments

app/MindWork AI Studio/Tools/ToolCallingSystem/ToolCallingImplementations/WebSearch/WebSearchTool.cs

@@ -97,6 +97,21 @@ public sealed class WebSearchTool(IEnumerable<IWebSearchBackend> backends, WebPa
     private const string TIME_RANGE_MONTH = "month";
     private const string TIME_RANGE_YEAR = "year";
 
+    /// <summary>
+    /// The time ranges a search can be restricted to.
+    /// </summary>
+    /// <remarks>
+    /// Only those which every service with a time filter understands: SearXNG documents day, month,
+    /// and year, while Tavily knows a week as well. The schema offers exactly these and the reader
+    /// checks against them, so the two cannot drift apart.
+    /// </remarks>
+    private static readonly string[] TIME_RANGES = [TIME_RANGE_DAY, TIME_RANGE_MONTH, TIME_RANGE_YEAR];
+
+    /// <summary>
+    /// How much of a wrongly passed argument an error message repeats back to the model.
+    /// </summary>
+    private const int MAX_ARGUMENT_ECHO_LENGTH = 40;
+
     public string ImplementationKey => ToolSelectionRules.WEB_SEARCH_TOOL_ID;
 
     /// <inheritdoc />
@@ -118,7 +133,7 @@ public sealed class WebSearchTool(IEnumerable<IWebSearchBackend> backends, WebPa
             Parameters = ToolParameterSchemaBuilder.Create()
                 .RequiredString(QUERY_ARGUMENT, "The search query.")
                 .OptionalString(LANGUAGE_ARGUMENT, "Optional IETF language tag restricting the search to one language, such as 'de-DE', 'en-US', or 'all' for no restriction. Leave it out to search in the language configured for this tool. Do not pass a language name such as 'German': search engines expect the tag and silently return nothing for anything else.")
-                .OptionalEnum(TIME_RANGE_ARGUMENT, "Optional time range filter for the search.", TIME_RANGE_DAY, TIME_RANGE_MONTH, TIME_RANGE_YEAR)
+                .OptionalEnum(TIME_RANGE_ARGUMENT, "Optional time range filter for the search.", TIME_RANGES)
                 .OptionalInteger(PAGE_ARGUMENT, "Optional search result page number starting at 1.")
                 .OptionalInteger(LIMIT_ARGUMENT, $"Optional maximum number of ranked result pages to retrieve and return. The hard maximum is {MAX_RESULTS}.")
                 .Build(),
@@ -482,14 +497,11 @@ public sealed class WebSearchTool(IEnumerable<IWebSearchBackend> backends, WebPa
 
     public async Task<ToolExecutionResult> ExecuteAsync(JsonElement arguments, ToolExecutionContext context, CancellationToken token = default)
     {
-        var query = ReadRequiredString(arguments, QUERY_ARGUMENT);
+        var query = ReadQuery(arguments);
-        var language = ReadOptionalString(arguments, LANGUAGE_ARGUMENT);
+        var language = ReadLanguage(arguments);
-        var timeRange = ReadOptionalString(arguments, TIME_RANGE_ARGUMENT);
+        var timeRange = ReadTimeRange(arguments);
-        var page = ReadOptionalPositiveInt(arguments, PAGE_ARGUMENT);
+        var page = ReadPage(arguments);
-        var requestedLimit = ReadOptionalPositiveInt(arguments, LIMIT_ARGUMENT);
+        var requestedLimit = ReadLimit(arguments);
-
-        if (timeRange is not null && timeRange is not (TIME_RANGE_DAY or TIME_RANGE_MONTH or TIME_RANGE_YEAR))
-            throw new ArgumentException($"Invalid time_range '{timeRange}'.");
 
         language = string.IsNullOrWhiteSpace(language) ? context.SettingsValues.GetValueOrDefault(DEFAULT_LANGUAGE_SETTING) : language;
         var safeSearch = ReadSafeSearchPolicy(context.SettingsValues);
@@ -807,42 +819,99 @@ public sealed class WebSearchTool(IEnumerable<IWebSearchBackend> backends, WebPa
         return result;
     }
 
-    private static string ReadRequiredString(JsonElement arguments, string propertyName)
+    /// <summary>
+    /// Reads the search query, the one argument the model always has to pass.
+    /// </summary>
+    internal static string ReadQuery(JsonElement arguments)
     {
-        var value = ReadOptionalString(arguments, propertyName);
+        var query = ReadOptionalString(arguments, QUERY_ARGUMENT, whenLeftOut: null);
-        if (string.IsNullOrWhiteSpace(value))
+        if (string.IsNullOrWhiteSpace(query))
-            throw new ArgumentException($"Missing required argument '{propertyName}'.");
+            throw new ArgumentException($"Missing required argument '{QUERY_ARGUMENT}'.");
 
-        return value;
+        return query;
     }
 
-    private static string? ReadOptionalString(JsonElement arguments, string propertyName)
+    /// <summary>
+    /// Reads the language tag the model asked for, or null for the configured language.
+    /// </summary>
+    internal static string? ReadLanguage(JsonElement arguments) => ReadOptionalString(arguments, LANGUAGE_ARGUMENT, "to use the configured language");
+
+    /// <summary>
+    /// Reads the time range the model asked for, or null for no restriction.
+    /// </summary>
+    internal static string? ReadTimeRange(JsonElement arguments)
     {
-        if (!arguments.TryGetProperty(propertyName, out var value))
+        if (!TryGetArgument(arguments, TIME_RANGE_ARGUMENT, out var value))
             return null;
 
-        return value.ValueKind switch
+        var timeRange = value.ValueKind is JsonValueKind.String ? value.GetString()?.Trim() : null;
-        {
+        if (timeRange is null || !TIME_RANGES.Contains(timeRange, StringComparer.Ordinal))
-            JsonValueKind.Null => null,
+            throw InvalidArgument(TIME_RANGE_ARGUMENT, value, $"one of {string.Join(", ", TIME_RANGES)}", "to search without a time restriction");
-            JsonValueKind.String => value.GetString()?.Trim(),
+
-            _ => throw new ArgumentException($"Argument '{propertyName}' must be a string."),
+        return timeRange;
-        };
     }
 
-    private static int? ReadOptionalPositiveInt(JsonElement arguments, string propertyName)
+    /// <summary>
+    /// Reads the result page the model asked for, or null for the first one.
+    /// </summary>
+    internal static int? ReadPage(JsonElement arguments) => ReadOptionalPositiveInt(arguments, PAGE_ARGUMENT, "to get the first page");
+
+    /// <summary>
+    /// Reads how many results the model asked for, or null for the configured number.
+    /// </summary>
+    internal static int? ReadLimit(JsonElement arguments) => ReadOptionalPositiveInt(arguments, LIMIT_ARGUMENT, "to get as many results as configured");
+
+    /// <summary>
+    /// Looks up an argument, treating null the same as leaving it out.
+    /// </summary>
+    private static bool TryGetArgument(JsonElement arguments, string propertyName, out JsonElement value) =>
+        arguments.TryGetProperty(propertyName, out value) && value.ValueKind is not JsonValueKind.Null;
+
+    private static string? ReadOptionalString(JsonElement arguments, string propertyName, string? whenLeftOut)
     {
-        if (!arguments.TryGetProperty(propertyName, out var value))
+        if (!TryGetArgument(arguments, propertyName, out var value))
             return null;
 
-        if (value.ValueKind is JsonValueKind.Null)
+        if (value.ValueKind is not JsonValueKind.String)
+            throw InvalidArgument(propertyName, value, "a string", whenLeftOut);
+
+        return value.GetString()?.Trim();
+    }
+
+    private static int? ReadOptionalPositiveInt(JsonElement arguments, string propertyName, string whenLeftOut)
+    {
+        if (!TryGetArgument(arguments, propertyName, out var value))
             return null;
 
         if (value.ValueKind is not JsonValueKind.Number || !value.TryGetInt32(out var intValue) || intValue <= 0)
-            throw new ArgumentException($"Argument '{propertyName}' must be a positive integer.");
+            throw InvalidArgument(propertyName, value, "a positive integer", whenLeftOut);
 
         return intValue;
     }
 
+    /// <summary>
+    /// Builds the error a model gets for an argument it passed wrongly.
+    /// </summary>
+    /// <remarks>
+    /// The model reads this and tries again, so it says what arrived, what would have been right,
+    /// and, for an optional argument, that leaving it out is always an option. A model which
+    /// believes the argument has to be there otherwise keeps trying placeholders, and every attempt
+    /// costs one of the tool calls an answer may make.
+    /// </remarks>
+    /// <param name="propertyName">The argument.</param>
+    /// <param name="value">What the model passed, as it arrived.</param>
+    /// <param name="expectation">What the argument must be, completing "must be ...".</param>
+    /// <param name="whenLeftOut">What happens without the argument, completing "Leave it out ...", or null for a required one.</param>
+    private static ArgumentException InvalidArgument(string propertyName, JsonElement value, string expectation, string? whenLeftOut)
+    {
+        var receivedValue = value.GetRawText();
+        if (receivedValue.Length > MAX_ARGUMENT_ECHO_LENGTH)
+            receivedValue = $"{receivedValue[..MAX_ARGUMENT_ECHO_LENGTH]}...";
+
+        var message = $"Argument '{propertyName}' must be {expectation}, but was {receivedValue}.";
+        return new ArgumentException(whenLeftOut is null ? message : $"{message} Leave it out {whenLeftOut}.");
+    }
+
     private static string FormatQueryForLog(string query)
     {
         var singleLineQuery = query

app/Tests/Tools/ToolCalling/WebSearchToolArgumentTests.cs

@@ -0,0 +1,157 @@
+using System.Text.Json;
+
+using AIStudio.Tools.ToolCallingSystem.ToolCallingImplementations.WebSearch;
+
+namespace AIStudio.Tests.Tools.ToolCalling;
+
+/// <summary>
+/// Checks how the web search reads the arguments a model passes, and what it says about wrong ones.
+/// </summary>
+/// <remarks>
+/// A wrong argument is refused rather than guessed at: a placeholder such as 0 is not a page, and
+/// quietly reading it as "no page" would do something the model did not ask for. What makes the
+/// refusal work is its message. The model reads it and tries again, so it has to say what arrived,
+/// what would have been right, and that an optional argument may simply be left out. A model
+/// which believes it has to pass one otherwise keeps trying placeholders, and every attempt costs
+/// one of the tool calls an answer may make.
+/// </remarks>
+[TestFixture]
+public sealed class WebSearchToolArgumentTests
+{
+    [Test]
+    public void AnArgumentLeftOutIsNotSet()
+    {
+        var arguments = Arguments("""{"query":"weather"}""");
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(WebSearchTool.ReadLanguage(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadTimeRange(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadPage(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadLimit(arguments), Is.Null);
+        });
+    }
+
+    [Test]
+    public void ANullArgumentIsTheSameAsOneLeftOut()
+    {
+        var arguments = Arguments("""{"query":"weather","language":null,"time_range":null,"page":null,"limit":null}""");
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(WebSearchTool.ReadLanguage(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadTimeRange(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadPage(arguments), Is.Null);
+            Assert.That(WebSearchTool.ReadLimit(arguments), Is.Null);
+        });
+    }
+
+    [Test]
+    public void ValidArgumentsComeThrough()
+    {
+        var arguments = Arguments("""{"query":"  weather  ","language":"de-DE","time_range":"day","page":2,"limit":7}""");
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(WebSearchTool.ReadQuery(arguments), Is.EqualTo("weather"));
+            Assert.That(WebSearchTool.ReadLanguage(arguments), Is.EqualTo("de-DE"));
+            Assert.That(WebSearchTool.ReadTimeRange(arguments), Is.EqualTo("day"));
+            Assert.That(WebSearchTool.ReadPage(arguments), Is.EqualTo(2));
+            Assert.That(WebSearchTool.ReadLimit(arguments), Is.EqualTo(7));
+        });
+    }
+
+    [TestCase("0")]
+    [TestCase("-1")]
+    [TestCase("2.5")]
+    [TestCase("\"5\"")]
+    [TestCase("true")]
+    public void AWrongPageIsRefusedWithWhatArrived(string value)
+    {
+        var message = Refusal(() => WebSearchTool.ReadPage(Arguments($$"""{"query":"weather","page":{{value}}}""")));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(message, Does.Contain("'page'").And.Contain("a positive integer"));
+            Assert.That(message, Does.Contain($"but was {value}."), "Without the value, the model cannot tell which of its arguments the tool means.");
+            Assert.That(message, Does.Contain("Leave it out"), "The way out a model needs when it believes the argument is required.");
+        });
+    }
+
+    [TestCase("0")]
+    [TestCase("-1")]
+    [TestCase("\"5\"")]
+    public void AWrongLimitIsRefusedWithWhatArrived(string value)
+    {
+        var message = Refusal(() => WebSearchTool.ReadLimit(Arguments($$"""{"query":"weather","limit":{{value}}}""")));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(message, Does.Contain("'limit'").And.Contain("a positive integer"));
+            Assert.That(message, Does.Contain($"but was {value}."));
+            Assert.That(message, Does.Contain("Leave it out"));
+        });
+    }
+
+    [TestCase("\"\"")]
+    [TestCase("\"Day\"")]
+    [TestCase("\"week\"")]
+    [TestCase("5")]
+    public void AWrongTimeRangeIsRefusedWithTheValuesThatWouldDo(string value)
+    {
+        var message = Refusal(() => WebSearchTool.ReadTimeRange(Arguments($$"""{"query":"weather","time_range":{{value}}}""")));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(message, Does.Contain("'time_range'").And.Contain("one of day, month, year"));
+            Assert.That(message, Does.Contain($"but was {value}."));
+            Assert.That(message, Does.Contain("Leave it out"));
+        });
+    }
+
+    [Test]
+    public void ALanguageWhichIsNoStringIsRefused()
+    {
+        var message = Refusal(() => WebSearchTool.ReadLanguage(Arguments("""{"query":"weather","language":5}""")));
+        Assert.That(message, Does.Contain("'language'").And.Contain("but was 5.").And.Contain("Leave it out"));
+    }
+
+    [Test]
+    public void AMissingQueryIsRefused()
+    {
+        var message = Refusal(() => WebSearchTool.ReadQuery(Arguments("""{"page":1}""")));
+        Assert.That(message, Does.Contain("Missing required argument 'query'"));
+    }
+
+    [Test]
+    public void AQueryWhichIsNoStringIsRefusedWithoutOfferingToLeaveItOut()
+    {
+        var message = Refusal(() => WebSearchTool.ReadQuery(Arguments("""{"query":42}""")));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(message, Does.Contain("'query'").And.Contain("but was 42."));
+            Assert.That(message, Does.Not.Contain("Leave it out"), "The query is required, so leaving it out is no way out.");
+        });
+    }
+
+    [Test]
+    public void ALongValueIsShortenedInTheMessage()
+    {
+        var longValue = new string('x', 500);
+        var message = Refusal(() => WebSearchTool.ReadTimeRange(Arguments($$"""{"query":"weather","time_range":"{{longValue}}"}""")));
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(message, Does.Not.Contain(longValue), "The model sent the value itself; repeating all of it back only costs tokens.");
+            Assert.That(message, Does.Contain("..."));
+        });
+    }
+
+    private static JsonElement Arguments(string json) => JsonSerializer.Deserialize<JsonElement>(json);
+
+    /// <summary>
+    /// Runs a reader which has to refuse its argument and returns what it said.
+    /// </summary>
+    private static string Refusal(TestDelegate read) => Assert.Throws<ArgumentException>(read)!.Message;
+}