mirror of
https://github.com/MindWorkAI/AI-Studio.git
synced 2026-10-05 11:29:40 +00:00
Delete the old capability rules
This commit is contained in:
17 files changed
+308
-2862
No files matched your search
@@ -1,76 +1,11 @@
|
||||
using AIStudio.Models;
|
||||
using AIStudio.Models;
|
||||
using AIStudio.Models.Registry;
|
||||
using AIStudio.Provider;
|
||||
using AIStudio.Provider.HuggingFace;
|
||||
|
||||
namespace AIStudio.Settings;
|
||||
|
||||
public static partial class ProviderExtensions
|
||||
{
|
||||
/// <summary>
|
||||
/// The longest model ID we normalize without going to the heap.
|
||||
/// </summary>
|
||||
private const int MAX_STACK_ALLOCATED_MODEL_ID_LENGTH = 256;
|
||||
|
||||
/// <summary>
|
||||
/// Brings a model ID into the form the capability rules are written in.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Every provider names the same model differently, and the difference is rarely in the words:
|
||||
/// it is in what sits between them. Ollama separates the variant with a colon
|
||||
/// ("qwen3.8:27b-mlx"), Blablador answers with a whole sentence ("10 - Muse Glimmer 30b - the
|
||||
/// newest META model"), Fireworks puts a path in front
|
||||
/// ("accounts/fireworks/models/llama-v3p1-405b-instruct"), and the hubs use hyphens. Without
|
||||
/// this, every rule would have to spell out each of those writings, which is what the Llama
|
||||
/// block used to do with four variants of one check.
|
||||
///
|
||||
/// The dots stay. They carry the version boundary: llama3 and llama3.1 are different models,
|
||||
/// and only the latter calls functions. Dropping them would merge the two.
|
||||
///
|
||||
/// The patterns in the rules are written in this normalized form already, which is why they
|
||||
/// use lowercase and hyphens throughout.
|
||||
/// </remarks>
|
||||
/// <param name="modelId">The model ID as the provider reports it.</param>
|
||||
/// <returns>The model ID in lowercase, with every separator written as a single hyphen.</returns>
|
||||
private static string NormalizeModelId(string modelId)
|
||||
{
|
||||
if (string.IsNullOrWhiteSpace(modelId))
|
||||
return string.Empty;
|
||||
|
||||
//
|
||||
// Normalizing never makes a name longer, so the original length is always enough room.
|
||||
// Model IDs are short, which is why the buffer lives on the stack: the longest ones we
|
||||
// know of are the descriptive names Blablador answers with, at around 75 characters. A
|
||||
// provider reporting something longer still gets a correct answer, just from the heap.
|
||||
//
|
||||
Span<char> normalized = modelId.Length <= MAX_STACK_ALLOCATED_MODEL_ID_LENGTH
|
||||
? stackalloc char[modelId.Length]
|
||||
: new char[modelId.Length];
|
||||
|
||||
var length = 0;
|
||||
foreach (var character in modelId)
|
||||
{
|
||||
if (char.IsAsciiLetterOrDigit(character) || character is '.')
|
||||
{
|
||||
normalized[length++] = char.ToLowerInvariant(character);
|
||||
continue;
|
||||
}
|
||||
|
||||
// Anything else separates two parts of the name. A leading separator, and a repeated
|
||||
// one, say nothing and would only get in the way of the patterns:
|
||||
if (length is 0 || normalized[length - 1] is '-')
|
||||
continue;
|
||||
|
||||
normalized[length++] = '-';
|
||||
}
|
||||
|
||||
// A trailing separator carries no meaning either:
|
||||
if (length > 0 && normalized[length - 1] is '-')
|
||||
length--;
|
||||
|
||||
return new string(normalized[..length]);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Everything the app knows about the model this provider instance is configured with.
|
||||
/// </summary>
|
||||
@@ -118,17 +53,6 @@ public static partial class ProviderExtensions
|
||||
: stated;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Get the capabilities of the model used by the configured provider.
|
||||
/// </summary>
|
||||
/// <param name="provider">The configured provider.</param>
|
||||
/// <returns>The capabilities of the configured model.</returns>
|
||||
public static List<Capability> GetModelCapabilities(this Provider provider)
|
||||
{
|
||||
var automaticCapabilities = provider.UsedLLMProvider.GetModelCapabilities(provider.Model);
|
||||
return provider.CapabilityOverrides?.ApplyTo(automaticCapabilities) ?? automaticCapabilities;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Get whether the model used by the configured provider accepts images as input.
|
||||
/// </summary>
|
||||
@@ -141,56 +65,4 @@ public static partial class ProviderExtensions
|
||||
/// <param name="provider">The configured provider.</param>
|
||||
/// <returns><c>true</c> when the model accepts image input.</returns>
|
||||
public static bool SupportsImageInput(this Provider provider) => provider.GetModelProfile().HasAny(Capability.SINGLE_IMAGE_INPUT | Capability.MULTIPLE_IMAGE_INPUT);
|
||||
|
||||
/// <summary>
|
||||
/// Get the capabilities of a model for a specific provider.
|
||||
/// </summary>
|
||||
/// <param name="provider">The LLM provider.</param>
|
||||
/// <param name="model">The model to get the capabilities for.</param>
|
||||
/// <returns>>The capabilities of the model.</returns>
|
||||
public static List<Capability> GetModelCapabilities(this LLMProviders provider, Model model)
|
||||
{
|
||||
if (string.IsNullOrWhiteSpace(model.Id))
|
||||
return [];
|
||||
|
||||
return provider switch
|
||||
{
|
||||
LLMProviders.OPEN_AI => GetModelCapabilitiesOpenAI(model),
|
||||
LLMProviders.MISTRAL => GetModelCapabilitiesMistral(model),
|
||||
LLMProviders.ANTHROPIC => GetModelCapabilitiesAnthropic(model),
|
||||
LLMProviders.GOOGLE => GetModelCapabilitiesGoogle(model),
|
||||
LLMProviders.X => GetModelCapabilitiesOpenSource(model),
|
||||
LLMProviders.DEEP_SEEK => GetModelCapabilitiesDeepSeek(model),
|
||||
LLMProviders.ALIBABA_CLOUD => GetModelCapabilitiesAlibaba(model),
|
||||
LLMProviders.PERPLEXITY => GetModelCapabilitiesPerplexity(model),
|
||||
LLMProviders.OPEN_ROUTER => GetModelCapabilitiesGateway(model),
|
||||
LLMProviders.HETZNER or LLMProviders.IONOS => GetModelCapabilitiesOpenSource(model),
|
||||
|
||||
//
|
||||
// LiteLLM is a gateway just like OpenRouter, and it names its models the same way:
|
||||
// "vendor/model", e.g. "anthropic/claude-opus-5" or "azure/gpt-5.6". So we let the
|
||||
// gateway detection handle it, which resolves the vendor prefix and asks the
|
||||
// provider who really knows the model. Everything it cannot place is treated as
|
||||
// an open source model, which is the right fallback for a freely named alias:
|
||||
//
|
||||
LLMProviders.LITE_LLM => GetModelCapabilitiesGateway(model),
|
||||
|
||||
LLMProviders.GROQ or LLMProviders.FIREWORKS => GetModelCapabilitiesOpenSource(model),
|
||||
|
||||
//
|
||||
// Hugging Face names its models the way the hub does, "org/model", which is the same
|
||||
// shape the other gateways use. So we let the gateway detection resolve the organization
|
||||
// and ask the provider implementation which really knows the model. The routing suffix
|
||||
// has to go first: it says which inference provider answers, not what the model is.
|
||||
//
|
||||
LLMProviders.HUGGINGFACE => GetModelCapabilitiesGateway(model.WithoutRoutingSuffix()),
|
||||
|
||||
LLMProviders.HELMHOLTZ => GetModelCapabilitiesOpenSource(model),
|
||||
LLMProviders.GWDG => GetModelCapabilitiesOpenSource(model),
|
||||
|
||||
LLMProviders.SELF_HOSTED => GetModelCapabilitiesOpenSource(model),
|
||||
|
||||
_ => []
|
||||
};
|
||||
}
|
||||
}
|
||||
Reference in new issue
Block a user