LLM Providers
HttpLlmProviderBase (Orkeon.Infrastructure.LLMs.Base) provides the abstract base for all LLM providers. It integrates HTTP handling (IHttpClientFactory), Polly resilience policies (retry, circuit breaker, timeout), and JSON serialization.
Retry budget. Llm:MaxRetries (10 by default) covers the failures that come back in seconds — request errors, 5xx, 429. A buffered call that hits Llm:TimeoutSeconds is retried once (ResilienceDefaults.LlmTimeoutRetries: every attempt costs the whole timeout), then the provider answers with a failure naming the setting and the two ways out (a longer timeout, thinking off); a caller's cancellation is never retried. A failed call travels as LlmResponse.Error and, through the chat-client adapter, as an exception — it is never mistaken for an empty answer, so the agent loop fails the task at once with the provider's reason instead of retrying without tools (LLM-11).
Implemented providers:
| Provider | Class | Namespace |
|---|---|---|
| OpenAI | OpenAIProvider (via OpenAICompatibleProviderBase) |
Orkeon.Infrastructure.LLMs |
| Anthropic | AnthropicLlmProvider |
Orkeon.Infrastructure.LLMs |
| Azure OpenAI | AzureOpenAILlmProvider |
Orkeon.Infrastructure.LLMs |
| Ollama | OllamaLlmProvider |
Orkeon.Infrastructure.LLMs |
| Together AI | TogetherAiLlmProvider |
Orkeon.Infrastructure.LLMs |
| DeepSeek | DeepSeekLlmProvider |
Orkeon.Infrastructure.LLMs |
| HuggingFace | HuggingFaceLlmProvider |
Orkeon.Infrastructure.LLMs |
| Kimi | KimiLlmProvider |
Orkeon.Infrastructure.LLMs |
| Qwen | QwenLlmProvider |
Orkeon.Infrastructure.LLMs |
| Mistral AI | MistralLlmProvider |
Orkeon.Infrastructure.LLMs |
| Z.AI (Zhipu GLM) | ZaiLlmProvider |
Orkeon.Infrastructure.LLMs |
| Google Gemini | GeminiLlmProvider |
Orkeon.Infrastructure.LLMs |
| Grok (x.AI) | GrokLlmProvider |
Orkeon.Infrastructure.LLMs |
| MiniMax | MiniMaxLlmProvider |
Orkeon.Infrastructure.LLMs |
| OpenRouter (aggregator) | OpenRouterLlmProvider |
Orkeon.Infrastructure.LLMs |
| Mammouth AI (aggregator) | MammouthLlmProvider |
Orkeon.Infrastructure.LLMs |
Generic adapters (ChatClientToLlmProviderAdapter, LlmProviderToChatClientAdapter, ChatClientToBasicLlmProviderAdapter) are available in Orkeon.Infrastructure.LLMs.Adapters to integrate other providers compatible with the IChatClient interface.
LlmProviderFactory (Orkeon.Infrastructure.LLMs) automatically resolves the provider from the LlmConfig (detection by URL, model name, or API key).
Declared capabilities
Every provider declares a LlmProviderCapabilities value object (Domain, exposed on
ILlmProvider; the base defaults to LlmProviderCapabilities.Unknown): ResponseFormat
(None/JsonObject/JsonSchema), Thinking (None/EffortOnly/Toggle/Budget),
Vision, ExplicitPromptCaching, RequiresJsonKeywordInPrompt, ReplaysReasoningContent.
OpenAICompatibleProviderBase translates the declaration into the OpenAI dialect once
(vision payloads, response_format, thinking, the CapabilityMismatchHint diagnostics);
Anthropic, Ollama and Qwen override the hook for their own dialects. An option a provider
cannot honour produces a structured warning — never a silent drop. The mirror-image case —
a constraint only the server can state — has its own seam:
OpenAICompatibleProviderBase.TryAdaptRejectedPayload gives a provider one chance to adapt
a payload the API rejected with a 4xx and re-send it once (generate and chat paths;
streaming never retries). Kimi uses it for Moonshot's invalid temperature: only 1 is allowed for this model — the mandated value is read from the rejection itself (which
models mandate it is decided server-side, a hard-coded list would drift) and the
substitution is logged as a structured warning. Two more dialect seams serve the
aggregators (LLM-09): ReasoningFieldName names the vendor field the reasoning trace is
read from (reasoning_content by default, reasoning on OpenRouter — the Orkeon metadata
key stays reasoning_content), and usage.cost becomes the cost metadata wherever a
vendor bills in the response. A chunk carrying a root-level error after the HTTP 200
ends a stream the way a pre-stream refusal does — error metadata on the chat stream, an
HttpRequestException on the token stream — never as a clean completion. All 16 providers
are IStreamingLlmProviders, and RateLimitedLlmProvider decorates any of them. The
per-provider matrix lives in the provider comparison.
Validating a provider against its real API
Every unit test in this area speaks to a mocked HTTP handler, which proves Orkeon sends what
we believe it sends — not that the vendor accepts it. The second proof is produced by the
campaign kit in llmproviders-test/:
llmproviders-test/run-campaign.sh --provider ollama --model llama3.2 # zero-cost first run
llmproviders-test/run-campaign.sh --all --config providers.local.json --dry-run
It drives orkeon llm probe over modes M1–M10, M12 and M13 of the provider
test protocol (maintainers' internal matrix) and archives
one Markdown report per campaign. orkeon llm models -p <provider> --filter 'gpt-5.6-*'
lists what a provider currently serves, so a campaign never depends on a hand-maintained
model list.
See also: Memory system · Security · Back to index