Read the context window from four more model lists

app/MindWork AI Studio/Provider/BaseProvider.cs

@@ -6,6 +6,7 @@ using System.Text.Json;
 using System.Text.Json.Serialization;
 
 using AIStudio.Chat;
+using AIStudio.Models.Live;
 using AIStudio.Provider.Anthropic;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Provider.SelfHosted;
@@ -191,6 +192,7 @@ public abstract class BaseProvider : IProvider, ISecretId
         Action<HttpRequestMessage, string>? requestConfigurator = null,
         JsonSerializerOptions? jsonSerializerOptions = null,
         bool isTryingSecret = false,
+        Func<TResponse, IEnumerable<ModelListing>>? listingFactory = null,
         CancellationToken token = default)
     {
         var secretKey = await this.GetModelLoadingSecretKey(storeType, apiKeyProvisional, isTryingSecret);
@@ -220,6 +222,16 @@ public abstract class BaseProvider : IProvider, ISecretId
                 if (parsedResponse is null)
                     return FailedModelLoadResult(ModelLoadFailureReason.INVALID_RESPONSE, "Model list response could not be deserialized.");
 
+                //
+                // What the list stated about the models, read before anything is filtered out of
+                // it: a model left out below as an embedding model is still a model somebody may
+                // have configured this instance with, and a list like this one is the only place
+                // its window is ever stated. Only pass a whole list in here -- reporting a part of
+                // one would tell the app that everything left out has stopped existing.
+                //
+                if (listingFactory is not null)
+                    ListedModels.Shared.Report(this.ConfiguredProviderId, listingFactory(parsedResponse));
+
                 return SuccessfulModelLoadResult(modelFactory(parsedResponse));
             }
             catch (Exception e)

app/MindWork AI Studio/Provider/Groq/GroqModel.cs

@@ -0,0 +1,15 @@
+using System.Text.Json.Serialization;
+
+namespace AIStudio.Provider.Groq;
+
+/// <summary>
+/// One model as Groq lists it.
+/// </summary>
+/// <remarks>
+/// Groq says more about a model than the shared OpenAI-compatible list does, which is why this
+/// provider brings a data model of its own instead of using that one: the shared record is read by
+/// a dozen providers, and a field only one of them sends has no business in it.
+/// </remarks>
+/// <param name="Id">The model's ID.</param>
+/// <param name="ContextWindowTokens">How much the model reads and writes in one conversation, in tokens.</param>
+public readonly record struct GroqModel(string Id, [property: JsonPropertyName("context_window")] int? ContextWindowTokens);

app/MindWork AI Studio/Provider/Groq/GroqModelsResponse.cs

@@ -0,0 +1,7 @@
+namespace AIStudio.Provider.Groq;
+
+/// <summary>
+/// A data model for the response from the Groq models endpoint.
+/// </summary>
+/// <param name="Data">The models Groq serves.</param>
+public readonly record struct GroqModelsResponse(IList<GroqModel> Data);

app/MindWork AI Studio/Provider/Groq/ProviderGroq.cs

@@ -1,6 +1,7 @@
 using System.Runtime.CompilerServices;
 
 using AIStudio.Chat;
+using AIStudio.Models.Live;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Settings;
 
@@ -113,10 +114,12 @@ public class ProviderGroq() : BaseProvider(LLMProviders.GROQ, new Uri("https://a
 
     private Task<ModelLoadResult> LoadModels(SecretStoreType storeType, string? apiKeyProvisional, CancellationToken token)
     {
-        return this.LoadModelsResponse<ModelsResponse>(
+        return this.LoadModelsResponse<GroqModelsResponse>(
             storeType,
             "models",
-            modelResponse => modelResponse.Data,
-            apiKeyProvisional, token: token);
+            modelResponse => modelResponse.Data.Select(n => new Model(n.Id, null)),
+            apiKeyProvisional,
+            listingFactory: modelResponse => modelResponse.Data.Select(n => ModelListing.For(n.Id, n.ContextWindowTokens)),
+            token: token);
     }
 }

app/MindWork AI Studio/Provider/HuggingFace/HFModel.cs

@@ -5,4 +5,30 @@ namespace AIStudio.Provider.HuggingFace;
 /// </summary>
 /// <param name="Id">The ID of the model, written as "org/model".</param>
 /// <param name="Providers">The inference providers serving this model.</param>
-public readonly record struct HFModel(string Id, IList<HFModelProvider>? Providers);
+public readonly record struct HFModel(string Id, IList<HFModelProvider>? Providers)
+{
+    /// <summary>
+    /// The window this model has when it is reached the way this user set things up.
+    /// </summary>
+    /// <remarks>
+    /// A window belongs to an inference provider here, not to the model: the same weights run
+    /// behind several of them, each configured by somebody else. Where the user named one, its
+    /// number is the answer. Where they let the router choose, the smallest window among the
+    /// providers currently serving the model is -- nobody knows which one the router will take, and
+    /// a number promising more than the chosen provider delivers would walk a conversation into an
+    /// error the user could not see coming.
+    /// </remarks>
+    /// <param name="providerSlug">The inference provider the user chose, or empty when the router chooses.</param>
+    /// <returns>The window in tokens, or null where nobody stated one.</returns>
+    public int? ContextWindowTokens(string providerSlug)
+    {
+        if (this.Providers is null)
+            return null;
+
+        var serving = this.Providers.Where(provider => provider.IsLive);
+        if (!string.IsNullOrEmpty(providerSlug))
+            serving = serving.Where(provider => string.Equals(provider.Provider, providerSlug, StringComparison.OrdinalIgnoreCase));
+
+        return serving.Where(provider => provider.ContextWindowTokens is > 0).Min(provider => provider.ContextWindowTokens);
+    }
+}

app/MindWork AI Studio/Provider/HuggingFace/HFModelProvider.cs

@@ -1,3 +1,5 @@
+using System.Text.Json.Serialization;
+
 namespace AIStudio.Provider.HuggingFace;
 
 /// <summary>
@@ -5,4 +7,13 @@ namespace AIStudio.Provider.HuggingFace;
 /// </summary>
 /// <param name="Provider">The slug of the inference provider, e.g. "novita".</param>
 /// <param name="Status">Whether the provider currently serves the model. Known value: "live".</param>
-public readonly record struct HFModelProvider(string Provider, string Status);
+/// <param name="ContextWindowTokens">How much this provider reads and writes in one conversation, in tokens.</param>
+public readonly record struct HFModelProvider(string Provider, string Status, [property: JsonPropertyName("context_length")] int? ContextWindowTokens)
+{
+    private const string LIVE = "live";
+
+    /// <summary>
+    /// Whether this provider serves the model right now.
+    /// </summary>
+    public bool IsLive => string.Equals(this.Status, LIVE, StringComparison.OrdinalIgnoreCase);
+}

app/MindWork AI Studio/Provider/HuggingFace/ProviderHuggingFace.cs

@@ -2,6 +2,7 @@
 using System.Runtime.CompilerServices;
 
 using AIStudio.Chat;
+using AIStudio.Models.Live;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Settings;
 using AIStudio.Tools.PluginSystem;
@@ -221,7 +222,23 @@ public sealed class ProviderHuggingFace : BaseProvider
     /// <inheritdoc />
     public override Task<ModelLoadResult> GetTextModels(string? apiKeyProvisional = null, CancellationToken token = default)
     {
-        return this.LoadModelsResponse<ModelsResponse>(SecretStoreType.LLM_PROVIDER, "models", this.SelectChatModels, apiKeyProvisional, token: token);
+        return this.LoadModelsResponse<ModelsResponse>(SecretStoreType.LLM_PROVIDER, "models", this.SelectChatModels, apiKeyProvisional, listingFactory: this.ListingsOf, token: token);
+    }
+
+    /// <summary>
+    /// What the router stated about the models it knows.
+    /// </summary>
+    /// <remarks>
+    /// Every model the router reports, not only the ones offered for chatting below: which models
+    /// are offered depends on the chosen inference provider, while a window belongs to whoever is
+    /// configured here, and both questions are asked of the same list.
+    /// </remarks>
+    /// <param name="response">The response of the model endpoint.</param>
+    /// <returns>One listing per model, which says nothing for the models nobody stated a window for.</returns>
+    private IEnumerable<ModelListing> ListingsOf(ModelsResponse response)
+    {
+        var providerSlug = this.hfProvider.EndpointsId();
+        return response.Data.Select(hfModel => ModelListing.For(hfModel.Id, hfModel.ContextWindowTokens(providerSlug)));
     }
 
     /// <summary>
@@ -253,8 +270,8 @@ public sealed class ProviderHuggingFace : BaseProvider
             return false;
 
         return hfModel.Providers.Any(provider =>
-            string.Equals(provider.Provider, providerSlug, StringComparison.OrdinalIgnoreCase) &&
-            string.Equals(provider.Status, "live", StringComparison.OrdinalIgnoreCase));
+            provider.IsLive &&
+            string.Equals(provider.Provider, providerSlug, StringComparison.OrdinalIgnoreCase));
     }
 
     /// <inheritdoc />

app/MindWork AI Studio/Provider/Mistral/Model.cs

@@ -1,3 +1,13 @@
+using System.Text.Json.Serialization;
+
 namespace AIStudio.Provider.Mistral;
 
-public readonly record struct Model(string Id, string Object, int Created, string OwnedBy);
+/// <summary>
+/// One model as Mistral lists it.
+/// </summary>
+/// <param name="Id">The model's ID.</param>
+/// <param name="Object">What kind of thing the entry is. Known value: "model".</param>
+/// <param name="Created">When the model was published, as seconds since the epoch.</param>
+/// <param name="OwnedBy">Who Mistral names as the owner of the model.</param>
+/// <param name="ContextWindowTokens">How much the model reads and writes in one conversation, in tokens.</param>
+public readonly record struct Model(string Id, string Object, int Created, string OwnedBy, [property: JsonPropertyName("max_context_length")] int? ContextWindowTokens);

app/MindWork AI Studio/Provider/Mistral/ProviderMistral.cs

@@ -1,6 +1,7 @@
 using System.Runtime.CompilerServices;
 
 using AIStudio.Chat;
+using AIStudio.Models.Live;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Settings;
 
@@ -139,6 +140,8 @@ public sealed class ProviderMistral() : BaseProvider(LLMProviders.MISTRAL, new U
             storeType,
             "models",
             modelResponse => modelResponse.Data.Select(n => new Provider.Model(n.Id, null)),
-            apiKeyProvisional, token: token);
+            apiKeyProvisional,
+            listingFactory: modelResponse => modelResponse.Data.Select(n => ModelListing.For(n.Id, n.ContextWindowTokens)),
+            token: token);
     }
 }

app/MindWork AI Studio/Provider/OpenRouter/OpenRouterModel.cs

@@ -1,8 +1,16 @@
+using System.Text.Json.Serialization;
+
 namespace AIStudio.Provider.OpenRouter;
 
 /// <summary>
 /// A data model for an OpenRouter model from the API.
 /// </summary>
+/// <remarks>
+/// The window is the model's, not that of any one provider behind it. OpenRouter also states a
+/// window per provider it currently prefers, but it picks one per request, so a number taken from
+/// there would describe a choice nobody has made yet.
+/// </remarks>
 /// <param name="Id">The model's ID.</param>
 /// <param name="Name">The model's human-readable display name.</param>
-public readonly record struct OpenRouterModel(string Id, string? Name);
+/// <param name="ContextWindowTokens">How much the model reads and writes in one conversation, in tokens.</param>
+public readonly record struct OpenRouterModel(string Id, string? Name, [property: JsonPropertyName("context_length")] int? ContextWindowTokens);

app/MindWork AI Studio/Provider/OpenRouter/ProviderOpenRouter.cs

@@ -2,6 +2,7 @@ using System.Net.Http.Headers;
 using System.Runtime.CompilerServices;
 
 using AIStudio.Chat;
+using AIStudio.Models.Live;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Settings;
 
@@ -125,9 +126,21 @@ public sealed class ProviderOpenRouter() : BaseProvider(LLMProviders.OPEN_ROUTER
                 request.Headers.Add("HTTP-Referer", PROJECT_WEBSITE);
                 request.Headers.Add("X-Title", PROJECT_NAME);
             },
+            listingFactory: modelResponse => modelResponse.Data.Select(n => ModelListing.For(n.Id, n.ContextWindowTokens)),
             token: token);
     }
 
+    /// <summary>
+    /// Loads the models OpenRouter offers for embedding, which live on a route of their own.
+    /// </summary>
+    /// <remarks>
+    /// Nothing is reported from here: this route answers with the embedding models alone, and what
+    /// is reported replaces everything an instance said before. The windows of the chat models
+    /// would go missing the moment somebody opens the embedding settings.
+    /// </remarks>
+    /// <param name="apiKeyProvisional">An API key which is not stored yet.</param>
+    /// <param name="token">The cancellation token to use.</param>
+    /// <returns>The embedding models.</returns>
     private Task<ModelLoadResult> LoadEmbeddingModels(string? apiKeyProvisional, CancellationToken token)
     {
         return this.LoadModelsResponse<OpenRouterModelsResponse>(

app/MindWork AI Studio/Provider/SelfHosted/ProviderSelfHosted.cs

@@ -3,7 +3,6 @@ using System.Runtime.CompilerServices;
 using System.Text.Json;
 
 using AIStudio.Chat;
-using AIStudio.Models;
 using AIStudio.Models.Live;
 using AIStudio.Provider.OpenAI;
 using AIStudio.Settings;
@@ -322,23 +321,9 @@ public sealed class ProviderSelfHosted(Host host, string hostname) : BaseProvide
     /// <summary>
     /// What an engine stated about the models it serves.
     /// </summary>
-    /// <remarks>
-    /// A window of zero or less is dropped rather than repaired. An engine answering that way is
-    /// telling us something we cannot interpret, and falling back to what the rules say about the
-    /// weights is the one answer nobody has to invent.
-    /// </remarks>
     /// <param name="models">The models exactly as the engine listed them.</param>
-    /// <returns>One listing per model the engine said something usable about.</returns>
-    private static IEnumerable<ModelListing> ListingsOf(IEnumerable<Model> models)
-    {
-        foreach (var model in models)
-        {
-            if (string.IsNullOrWhiteSpace(model.Id) || model.ContextWindowTokens is not > 0)
-                continue;
-
-            yield return new(model.Id, ContextWindow.Of(model.ContextWindowTokens.Value));
-        }
-    }
+    /// <returns>One listing per model, which says nothing for the models the engine was silent about.</returns>
+    private static IEnumerable<ModelListing> ListingsOf(IEnumerable<Model> models) => models.Select(model => ModelListing.For(model.Id, model.ContextWindowTokens));
 
     private static bool IsMatchingLlamaCppTextModel(Model model, string[] ignorePhrases, string[] filterPhrases)
     {