Moonshot
The Moonshot provider uses provider key moonshot and the HPD-Agent.Providers.Moonshot package. It is backed by the shared OpenAI-compatible chat-completions client, so ModelName is a Moonshot or Kimi chat model id such as kimi-k2.5, kimi-k2.6, or moonshot-v1-128k.
Set an API key:
export MOONSHOT_API_KEY="..."KIMI_API_KEY is also accepted. The default endpoint is https://api.moonshot.ai/v1/, which sends chat requests to chat/completions. Pass endpoint to WithMoonshot(...), set Endpoint in ChatClientConfig, or use MOONSHOT_ENDPOINT, MOONSHOT_BASE_URL, KIMI_ENDPOINT, or KIMI_BASE_URL when you need a compatible proxy or custom base URL.
Use fluent chat setup first:
using HPD.Agent;
using HPD.Agent.Providers.Moonshot;
var agent = await new AgentBuilder()
.WithMoonshot(model: "kimi-k2.5")
.BuildAsync();
var result = await agent.RunAsync("Write one sentence about Moonshot setup.");
Console.WriteLine(result.Text);Provider construction stays small. Put model behavior on the agent's Clients.Chat default or a run's complete Clients.Chat replacement, and use Moonshot request options only for Kimi-specific request fields:
using HPD.Agent;
using HPD.Agent.Providers.Moonshot;
var agent = await new AgentBuilder()
.WithMoonshot(model: "kimi-k2.6")
.WithMoonshotChatRequestOptions(moonshot =>
{
moonshot.ThinkingKeep = MoonshotThinkingKeep.All;
})
.BuildAsync();The provider helper writes the same portable fields, typed providerConfig, and typed providerOptions into Clients.Chat. Use the runtime selection guide for the canonical JSON and YAML shape.
Runtime Chat Options
Use ChatClientConfig for shared model-call behavior such as temperature, top-p, max output tokens, stop sequences, response format, tools, tool mode, seed, and generic reasoning options. Put agent-level defaults in AgentConfig.Clients.Chat; use AgentRunConfig.Clients.Chat for complete per-run replacements.
Moonshot/Kimi thinking type is derived from ChatClientConfig.Reasoning when the selected model supports Kimi thinking.
Use MoonshotChatRequestOptions for Moonshot/Kimi-only request properties:
| Option | Moonshot key | Meaning |
|---|---|---|
ThinkingKeep | thinking_keep | Preserves reasoning content from historical assistant messages. Supported value: all. |
Moonshot Construction Options
Moonshot currently has no provider-specific provider configuration. Use ChatClientConfig.Endpoint, the fluent endpoint argument, or endpoint environment variables for custom base URLs.
Use ChatClientConfig.ResponseFormat for text, JSON object, or JSON schema response formatting where the selected Moonshot/Kimi model supports it.
Caveats
This package currently registers only the chat family. Moonshot files, embeddings, media upload, token counting, billing, batches, and model listing are deferred until those families have dedicated HPD provider support.
Moonshot chat uses HPD's shared OpenAI-compatible chat-completions client, not the generated Moonshot SDK. The generated SDK is useful as a reference for Kimi request fields such as thinking, but HPD sends chat through the OpenAI-compatible endpoint. It supports token streaming, function tools, response format and tool mode, seed, max output tokens, top-p, temperature, and stop sequences where the selected Moonshot/Kimi model supports them. ChatClientConfig.Reasoning and MoonshotChatRequestOptions.ThinkingKeep add a Kimi thinking object to the request for models that support those fields.
Runtime validation checks endpoint shape. Runtime chat options are applied through Clients.Chat and AgentRunConfig.Clients.Chat. Credentials are resolved lazily when a run first acquires this provider client; a missing credential fails that run, not BuildAsync().