메뉴
Assist with development in the Twilight AI Go SDK. Use when working in this repository, adding or updating providers, embeddings, tool calling, streaming, examples, or docs for Twilight AI.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/felinics/Memoh --skill twilight-ai명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Configure Memoh hooks in /data/.memoh/hooks.json, including events, matchers, command/tool actions, decisions, on_error behavior, append_context, and validation. Use when the user asks to inspect, create, update, debug, or explain workspace hooks.
Create or update Memoh workspace skills under /data/skills, including SKILL.md, scripts, references, assets, and validation. Use when the user asks to create, scaffold, revise, audit, or explain a workspace skill.
Memoh 统一错误处理架构（apperror + Problem Details + SSE envelope + 前端稳定 code 解析）。 新增任何面向用户的后端错误、或前端需要根据错误做业务分支时，必须先读本 skill。 核心约束：业务判断只依赖稳定 code，用户文案由前端按语言本地化，私有诊断永不进入响应。
SKILL.md 표시 중
| name | twilight-ai |
| description | Assist with development in the Twilight AI Go SDK. Use when working in this repository, adding or updating providers, embeddings, tool calling, streaming, examples, or docs for Twilight AI. |
Use this skill when the task involves twilight-ai, especially:
sdk/provider/GenerateText, GenerateTextResult, StreamText, Embed, or EmbedManyTwilight AI is a lightweight Go AI SDK with a provider-agnostic core API.
sdk.GenerateText, sdk.GenerateTextResult, sdk.StreamTextsdk.Embed, sdk.EmbedManysdk.Tool, sdk.NewTool[T], WithMaxSteps, approval flowsdk.CreateMCPClient, sdk.MCPClient, sdk.MCPClientConfigStreamPart events over Go channelsprovider/openai/completionsprovider/openai/responsesprovider/anthropic/messagesprovider/google/generativeaiprovider/openai/embeddingprovider/google/embeddingPrefer the high-level SDK API first, then drop to provider details only when needed.
sdk.Model binds a chat model to a sdk.Providersdk.EmbeddingModel binds an embedding model to an sdk.EmbeddingProvidersdk.Tool valuesChoose the narrowest API that matches the task:
sdk.GenerateTextsdk.GenerateTextResultsdk.StreamTextsdk.Embedsdk.EmbedManyIf the task introduces examples or docs, prefer simple end-to-end snippets that start with:
openai/completions for broad OpenAI-compatible support such as DeepSeek, Groq, Ollama, Azure-style compatible endpoints, and generic /chat/completions backends.openai/responses when the task needs OpenAI Responses API features such as first-class reasoning models, reasoning summaries, URL citation annotations, or flat input mapping.anthropic/messages for Claude and Anthropic extended thinking via WithThinking.google/generativeai for Gemini chat, tool calling, vision, streaming, and Gemini reasoning.openai/embedding or google/embedding for embeddings. Keep embedding-provider work separate from chat-provider work.If adding or changing a chat provider, preserve the sdk.Provider contract:
Name()ListModels(ctx)Test(ctx)TestModel(ctx, modelID)DoGenerate(ctx, params)DoStream(ctx, params)Keep provider responsibilities focused:
sdk.GenerateResultsdk.StreamPart valuesEmbedding providers are separate from chat providers. Use sdk.EmbeddingProvider and return an sdk.EmbeddingModel via EmbeddingModel(id).
When updating embeddings:
sdk.Embed for single-string conveniencesdk.EmbedMany for batched requestsUsage.TokensPrefer sdk.NewTool[T] for new tool examples and integrations. It gives typed input and inferred JSON Schema.
Use these defaults unless the task requires something else:
WithToolChoice("auto") for normal useWithMaxSteps(0) for inspection-only tool callsWithMaxSteps(N) for automatic execution loopsRequireApproval: true only for sensitive side effectsWhen streaming with tools, ensure the implementation can emit:
Use MCP when the task needs remote tools exposed by an MCP server rather than locally implemented Execute handlers.
Default guidance:
sdk.CreateMCPClient(ctx, &sdk.MCPClientConfig{...})sdk.MCPTransportHTTP for streamable HTTP MCP serverssdk.MCPTransportSSE only when the server exposes legacy SSE transportTransport: ...mcpClient.Tools(ctx) and pass the result into sdk.WithTools(...)defer mcpClient.Close() after successful creationImportant behavior:
sdk.Tool values from the caller's perspectiveInputSchema into *jsonschema.Schematools/call on the remote serverTwilight AI streaming is channel-first and type-safe. Prefer type switches over loosely typed event parsing.
Important expectations:
StreamText returns *sdk.StreamResultsr.Stream must be consumed before relying on sr.Steps or sr.MessagesText() and ToResult() are the convenience paths when callers do not want manual event handlingPreserve the SDK message model and avoid backend-specific shapes leaking into public usage.
stop, length, content-filter, and tool-callsUse this structure:
ChatModel or EmbeddingModelsdk functionCheck all affected layers:
Use the built-in providers as the template. A custom provider should feel identical to existing ones from the caller's perspective.
Minimum behavior:
ChatModelDoGenerateDoStream with correct lifecycle partsWhen writing Twilight AI docs or README content:
Use these terms consistently:
WithMaxStepsStreamTextBefore finishing work in this repo, verify:
sdk APIs unless lower-level behavior is the pointStreamPart handlingWithTools(...) usage when relevant