Namespace JD.AI.Core.Providers

Classes

AmazonBedrockDetector

Detects AWS Bedrock availability via AWS credentials. Uses the official Microsoft.SemanticKernel.Connectors.Amazon package.

AnthropicDetector

Detects Anthropic API availability via API key. Dynamically discovers models from the Anthropic API, falling back to a curated catalog of well-known Claude models.

ApiKeyProviderDetectorBase

Shared template for API-key based provider detectors. Subclasses supply a static KnownModels catalog as a fallback and may override DiscoverModelsAsync(string, CancellationToken) to query their API for the current model list at detection time.

AzureOpenAIDetector

Detects Azure OpenAI availability via API key and endpoint. Requires apikey, endpoint, and optional comma-separated deployments list.

ClaudeCodeDetector

Detects a local Claude Code session and exposes its models. When running as a Windows service, scans user profiles for credentials. When the session token is expired, attempts a silent refresh via the Claude CLI.

CopilotDetector

Detects a local GitHub Copilot session and enumerates its models. When running as a Windows service, scans user profiles for credentials. When authentication fails, attempts a silent refresh via the gh CLI.

FoundryLocalDetector

Detects a locally-running Microsoft Foundry Local instance and enumerates its models via the OpenAI-compatible REST API. Port is discovered dynamically via the Foundry CLI or process scanning.

GoogleGeminiDetector

Detects Google Gemini API availability via API key. Dynamically discovers models from the Generative Language API, filtering to generateContent-capable models.

HuggingFaceDetector

Detects HuggingFace Inference API availability via API key. Dynamically discovers the most popular inference-ready chat models from the HuggingFace Hub, falling back to a curated catalog when offline.

HuggingFace deprecated the old api-inference.huggingface.co endpoint (returns 410 Gone). All requests now use the OpenAI-compatible router: https://router.huggingface.co/v1/

MistralDetector

Detects Mistral AI availability via API key. Dynamically discovers models from the Mistral API, falling back to a curated catalog when the endpoint is unreachable.

ModelCapabilitiesExtensions

Helpers for ModelCapabilities display.

ModelCapabilityEntry

Registry projection for model capability-aware routing.

ModelCapabilityRegistry

Default in-memory capability registry populated from provider model info.

OllamaDetector

Detects Ollama instances (local and optionally named remote endpoints) and enumerates models.

OpenAICodexDetector

Detects a local OpenAI Codex session and exposes its models. When running as a Windows service, scans user profiles for credentials. Supports automatic token refresh and device code login.

OpenAICompatibleDetector

Generic detector for OpenAI-compatible API endpoints. Supports multiple named instances (Groq, Together, DeepSeek, etc.). Each instance is stored as openai-compat:{alias} in the credential store.

OpenAIDetector

Detects OpenAI API availability via API key and discovers models.

OpenRouterDetector

First-class provider for OpenRouter — a unified API that exposes hundreds of AI models from multiple vendors through an OpenAI-compatible interface. Dynamically discovers models from the OpenRouter catalog, enriching each with context length, pricing, and capability metadata.

ProviderInfo

Describes a detected AI provider and its available models.

ProviderModelInfo

A model available through a provider.

ProviderRegistry

Aggregates all IProviderDetector instances and exposes a unified model catalog with kernel-building capability.

ProviderServiceCollectionExtensions

Service registration helpers for the default gateway/daemon provider detector set.

Interfaces

IModelCapabilityRegistry

Tracks model capabilities and supports capability-based lookups.

IProviderDetector

Detects a single AI backend and builds kernels for its models.

IProviderRegistry

Detects and registers AI providers, builds kernels on demand.

Enums

ModelCapabilities

Flags describing what a model supports. A model may support one or more capabilities simultaneously.

ModelCapability

Capability flags for model-level routing decisions.

ModelCostTier

Coarse pricing tier used for capability routing preferences.