# Kiln

> Kiln is an inference API by Bricksum that serves OpenAI-style Chat Completions, Completions and Responses endpoints and an Anthropic-style Messages endpoint (`POST /chat/completions`, `POST /completions`, `POST /responses`, `POST /messages`, `POST /messages/count_tokens`, `GET /models`, `GET /key`). Base URL `https://api.bricksum.com/v1`, API keys start with `sk-bk-`, list models with `GET https://api.bricksum.com/v1/models`. Default model ID: `qwen3-32b`. Console: https://kiln.bricksum.com. Status: https://status.bricksum.com.

The docs are HTML pages in English, Korean and Japanese; each section below lists one locale.
Every page has a plain-Markdown twin at the (md) link on its line. Each locale also has a single file with every page of that locale, linked at the top of its section.

## Docs (English)

- [All pages of this locale in one Markdown file](https://kiln.bricksum.com/docs/en/llms-full.txt)

- [Documentation](https://kiln.bricksum.com/docs/en) ([md](https://kiln.bricksum.com/docs/en/md/overview.md)): The Kiln API: AI inference you call with the OpenAI SDK. Set base_url and an API key to call models.
- [Quickstart](https://kiln.bricksum.com/docs/en/quickstart) ([md](https://kiln.bricksum.com/docs/en/md/quickstart.md)): A step-by-step walkthrough from creating an API key to your first API request.
- [Models](https://kiln.bricksum.com/docs/en/models) ([md](https://kiln.bricksum.com/docs/en/md/models.md)): How to list the models you can call through the API, and the context length, pricing and capabilities of each one.
- [API Reference](https://kiln.bricksum.com/docs/en/api-reference) ([md](https://kiln.bricksum.com/docs/en/md/api-reference.md)): The hub of the API reference: base URL, authentication, the endpoints (Chat Completions, Completions, Responses, Messages, models, key), the parameters they share, the response envelope, streaming, errors and limits.
- [Chat completions](https://kiln.bricksum.com/docs/en/api-reference/chat-completions) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/chat-completions.md)): POST /chat/completions reference: every request parameter with its support status, response and stream chunk fields, errors, limits and examples.
- [Responses](https://kiln.bricksum.com/docs/en/api-reference/responses) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/responses.md)): POST /responses reference: parameter support status, stream events, where usage.cost appears, errors and examples. Responses are not stored; previous_response_id is rejected.
- [Completions](https://kiln.bricksum.com/docs/en/api-reference/completions) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/completions.md)): POST /completions reference: every request parameter with its support status, response and stream chunk fields, errors, limits and examples.
- [Streaming](https://kiln.bricksum.com/docs/en/streaming) ([md](https://kiln.bricksum.com/docs/en/md/streaming.md)): Server-sent events: chunk format, keepalive comments, usage and cost in the final chunk, what happens on disconnect.
- [Usage & billing](https://kiln.bricksum.com/docs/en/usage-billing) ([md](https://kiln.bricksum.com/docs/en/md/usage-billing.md)): The usage object and usage.cost, how cost is computed, the generation ID header, checking your key, and what is billed.
- [Authentication](https://kiln.bricksum.com/docs/en/authentication) ([md](https://kiln.bricksum.com/docs/en/md/authentication.md)): Bearer API keys, key management, and org/project scoping.
- [Organizations & budgets](https://kiln.bricksum.com/docs/en/organizations) ([md](https://kiln.bricksum.com/docs/en/md/organizations.md)): Roles, members and invites, projects and keys, key spend limits and expiry, the monthly budget, credits and their expiry, payments, status.
- [Rate limits](https://kiln.bricksum.com/docs/en/rate-limits) ([md](https://kiln.bricksum.com/docs/en/md/rate-limits.md)): Per-organization RPM/TPM limits, 429 behavior, and retry guidance.
- [Errors](https://kiln.bricksum.com/docs/en/errors) ([md](https://kiln.bricksum.com/docs/en/md/errors.md)): The error response shape and common HTTP status codes.
- [Integrations](https://kiln.bricksum.com/docs/en/integrations) ([md](https://kiln.bricksum.com/docs/en/md/integrations.md)): Index of the integrations section: a page each for the OpenAI SDK, the Anthropic SDK, the Vercel AI SDK, LangChain and PydanticAI, and the six coding tools — Claude Code, Codex CLI, Hermes Agent, OpenCode, OpenClaw and Cursor — with their configuration on this page.
- [FAQ](https://kiln.bricksum.com/docs/en/faq) ([md](https://kiln.bricksum.com/docs/en/md/faq.md)): Billing, trials, models, SDK compatibility, streaming, and data usage.
- [Messages](https://kiln.bricksum.com/docs/en/api-reference/messages) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/messages.md)): POST /messages reference (Anthropic Messages format): headers, every request parameter with its translation status, response fields, stream events, errors, limits and examples.
- [Count tokens](https://kiln.bricksum.com/docs/en/api-reference/count-tokens) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/count-tokens.md)): POST /messages/count_tokens reference: request fields, the input_tokens response, the 404 that lets a client estimate, and why the request is not billed.
- [Migrating from OpenAI](https://kiln.bricksum.com/docs/en/migrating-from-openai) ([md](https://kiln.bricksum.com/docs/en/md/migrating-from-openai.md)): Move an OpenAI client to Kiln: the base_url, key and model ID to change, the known limitations, and a checklist before production traffic.
- [List models](https://kiln.bricksum.com/docs/en/api-reference/models) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/models.md)): GET /models reference: the model object the gateway returns, what the list leaves out, the fallback when the catalog is empty, errors and examples.
- [Key info](https://kiln.bricksum.com/docs/en/api-reference/key) ([md](https://kiln.bricksum.com/docs/en/md/api-reference/key.md)): GET /key reference: read a key's organization, project, status, spend limit, expiry and spend with the key itself. Its error body is {message, code} and every rejected key gets 401.
- [OpenAI SDK](https://kiln.bricksum.com/docs/en/sdks/openai) ([md](https://kiln.bricksum.com/docs/en/md/sdks/openai.md)): Call Kiln with the official OpenAI clients in Python, TypeScript and Go: install, a verified request, the tool-calling round trip, and what differs — the cost field the OpenAI schema has no room for, and the reasoning budget inside max_tokens.
- [Anthropic SDK](https://kiln.bricksum.com/docs/en/sdks/anthropic) ([md](https://kiln.bricksum.com/docs/en/md/sdks/anthropic.md)): Call Kiln with the official Anthropic clients in Python and TypeScript: the base URL without /v1, why content[0] is a thinking block, and what the usage object does and does not carry.
- [Vercel AI SDK](https://kiln.bricksum.com/docs/en/sdks/vercel-ai) ([md](https://kiln.bricksum.com/docs/en/md/sdks/vercel-ai.md)): Stream from Kiln with the Vercel AI SDK: install ai and the openai-compatible provider, a verified streamText call, and why @ai-sdk/openai is the wrong package here.
- [LangChain](https://kiln.bricksum.com/docs/en/sdks/langchain) ([md](https://kiln.bricksum.com/docs/en/md/sdks/langchain.md)): Point LangChain's ChatOpenAI at Kiln: install, a verified invoke and stream call, and what the message content holds once the model has reasoned.
- [PydanticAI](https://kiln.bricksum.com/docs/en/sdks/pydantic-ai) ([md](https://kiln.bricksum.com/docs/en/md/sdks/pydantic-ai.md)): Run a PydanticAI agent on Kiln: install, a verified run_sync call, OpenAIChatModel in place of the renamed OpenAIModel, and where max_tokens belongs.
- [Cookbook](https://kiln.bricksum.com/docs/en/cookbook) ([md](https://kiln.bricksum.com/docs/en/md/cookbook.md)): Worked, verified recipes for building against Kiln: tutorials and integrations.
- [Streaming chat CLI quickstart](https://kiln.bricksum.com/docs/en/cookbook/streaming-chat-cli) ([md](https://kiln.bricksum.com/docs/en/md/cookbook/streaming-chat-cli.md)): Build a terminal chat client that prints Kiln's reply as it streams in, token by token.
- [Tool calling end to end](https://kiln.bricksum.com/docs/en/cookbook/tool-calling) ([md](https://kiln.bricksum.com/docs/en/md/cookbook/tool-calling.md)): One round trip with tool_choice "auto": ask for a tool, run it, send the result back.

## Docs (한국어)

- [All pages of this locale in one Markdown file](https://kiln.bricksum.com/docs/ko/llms-full.txt)

- [문서](https://kiln.bricksum.com/docs/ko) ([md](https://kiln.bricksum.com/docs/ko/md/overview.md)): Kiln API는 OpenAI SDK와 호환되는 AI 추론 API입니다. base_url만 바꾸면 바로 모델을 호출할 수 있습니다.
- [빠른 시작](https://kiln.bricksum.com/docs/ko/quickstart) ([md](https://kiln.bricksum.com/docs/ko/md/quickstart.md)): API 키 발급부터 첫 API 요청까지 단계별로 안내합니다.
- [모델](https://kiln.bricksum.com/docs/ko/models) ([md](https://kiln.bricksum.com/docs/ko/md/models.md)): 호출할 수 있는 모델을 API로 조회하는 방법과 모델별 컨텍스트 길이·단가·지원 기능을 설명합니다.
- [API 레퍼런스](https://kiln.bricksum.com/docs/ko/api-reference) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference.md)): API 레퍼런스의 허브입니다. Base URL, 인증, 엔드포인트 목록(Chat Completions, Completions, Responses, Messages, models, key), 공통 파라미터, 응답 형식, 스트리밍, 에러, 제한을 설명합니다.
- [Chat completions](https://kiln.bricksum.com/docs/ko/api-reference/chat-completions) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/chat-completions.md)): POST /chat/completions 레퍼런스입니다. 요청 파라미터별 지원 상태, 응답과 스트림 청크 필드, 에러, 제한, 예제를 설명합니다.
- [Responses](https://kiln.bricksum.com/docs/ko/api-reference/responses) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/responses.md)): POST /responses 레퍼런스입니다. 파라미터별 지원 상태, 스트림 이벤트, usage.cost가 포함되는 위치, 에러, 예제를 설명합니다. 응답은 저장되지 않으며 previous_response_id는 거부됩니다.
- [Completions](https://kiln.bricksum.com/docs/ko/api-reference/completions) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/completions.md)): POST /completions 레퍼런스입니다. 요청 파라미터별 지원 상태, 응답과 스트림 청크 필드, 에러, 제한, 예제를 설명합니다.
- [스트리밍](https://kiln.bricksum.com/docs/ko/streaming) ([md](https://kiln.bricksum.com/docs/ko/md/streaming.md)): Server-sent events의 청크 형식, keepalive 주석, 마지막 청크의 usage와 cost, 연결이 끊겼을 때의 동작을 설명합니다.
- [사용량과 과금](https://kiln.bricksum.com/docs/ko/usage-billing) ([md](https://kiln.bricksum.com/docs/ko/md/usage-billing.md)): usage 객체와 usage.cost, 비용 계산 방식, generation ID 헤더, 키 확인, 과금 대상을 설명합니다.
- [인증](https://kiln.bricksum.com/docs/ko/authentication) ([md](https://kiln.bricksum.com/docs/ko/md/authentication.md)): Bearer API 키 인증, 키 관리, 키가 속하는 조직과 프로젝트를 설명합니다.
- [조직과 예산](https://kiln.bricksum.com/docs/ko/organizations) ([md](https://kiln.bricksum.com/docs/ko/md/organizations.md)): 역할, 멤버와 초대, 프로젝트와 키, 키 지출 한도와 만료, 월 예산, 크레딧과 만료, 결제, 서비스 상태를 설명합니다.
- [레이트 리밋](https://kiln.bricksum.com/docs/ko/rate-limits) ([md](https://kiln.bricksum.com/docs/ko/md/rate-limits.md)): 조직 단위 RPM/TPM 한도, 429 동작, 재시도 가이드를 설명합니다.
- [에러](https://kiln.bricksum.com/docs/ko/errors) ([md](https://kiln.bricksum.com/docs/ko/md/errors.md)): 에러 응답 형식과 주요 HTTP 상태 코드를 설명합니다.
- [연동](https://kiln.bricksum.com/docs/ko/integrations) ([md](https://kiln.bricksum.com/docs/ko/md/integrations.md)): 연동 섹션의 목차입니다. OpenAI SDK, Anthropic SDK, Vercel AI SDK, LangChain, PydanticAI는 각각 별도 페이지로 있고, 코딩 도구 여섯 개(Claude Code, Codex CLI, Hermes Agent, OpenCode, OpenClaw, Cursor)의 설정은 이 페이지에 있습니다.
- [자주 묻는 질문](https://kiln.bricksum.com/docs/ko/faq) ([md](https://kiln.bricksum.com/docs/ko/md/faq.md)): 과금·크레딧, 체험 크레딧, 모델, SDK 호환성, 스트리밍, 데이터 사용에 관한 질문과 답입니다.
- [Messages](https://kiln.bricksum.com/docs/ko/api-reference/messages) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/messages.md)): POST /messages 레퍼런스(Anthropic Messages 형식)입니다. 헤더, 요청 파라미터별 변환 상태, 응답 필드, 스트림 이벤트, 에러, 제한, 예제를 설명합니다.
- [Count tokens](https://kiln.bricksum.com/docs/ko/api-reference/count-tokens) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/count-tokens.md)): POST /messages/count_tokens 레퍼런스입니다. 요청 필드, input_tokens 응답, 클라이언트가 추정값을 쓰게 하는 404, 과금되지 않는 이유를 설명합니다.
- [OpenAI에서 옮겨오기](https://kiln.bricksum.com/docs/ko/migrating-from-openai) ([md](https://kiln.bricksum.com/docs/ko/md/migrating-from-openai.md)): OpenAI 클라이언트를 Kiln으로 옮기는 방법입니다. 바꿔야 하는 base_url·키·모델 ID, 알려진 제한, 운영 전환 전 점검 목록을 정리합니다.
- [List models](https://kiln.bricksum.com/docs/ko/api-reference/models) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/models.md)): GET /models 레퍼런스입니다. 게이트웨이가 반환하는 모델 객체, 목록에 담기지 않는 정보, 카탈로그가 비었을 때의 폴백, 에러, 예제를 설명합니다.
- [Key info](https://kiln.bricksum.com/docs/ko/api-reference/key) ([md](https://kiln.bricksum.com/docs/ko/md/api-reference/key.md)): GET /key 레퍼런스입니다. 키 자체로 그 키의 조직, 프로젝트, 상태, 지출 한도, 만료, 사용액을 조회합니다. 에러 본문은 {message, code} 형식이고, 거부되는 키는 모두 401을 받습니다.
- [OpenAI SDK](https://kiln.bricksum.com/docs/ko/sdks/openai) ([md](https://kiln.bricksum.com/docs/ko/md/sdks/openai.md)): Python, TypeScript, Go 공식 OpenAI 클라이언트로 Kiln을 호출하는 방법입니다. 설치, 검증한 요청 예제, 툴 호출 왕복, 그리고 OpenAI 스키마에 없는 cost 필드와 max_tokens 안의 reasoning 예산 등 다른 점을 정리합니다.
- [Anthropic SDK](https://kiln.bricksum.com/docs/ko/sdks/anthropic) ([md](https://kiln.bricksum.com/docs/ko/md/sdks/anthropic.md)): Python, TypeScript 공식 Anthropic 클라이언트로 Kiln을 호출하는 방법입니다. /v1을 뺀 base URL, content[0]이 thinking 블록인 이유, usage 객체에 담기는 값과 담기지 않는 값을 설명합니다.
- [Vercel AI SDK](https://kiln.bricksum.com/docs/ko/sdks/vercel-ai) ([md](https://kiln.bricksum.com/docs/ko/md/sdks/vercel-ai.md)): Vercel AI SDK로 Kiln에서 스트리밍하는 방법입니다. ai와 openai-compatible 프로바이더 설치, 검증한 streamText 호출, @ai-sdk/openai가 맞지 않는 이유를 설명합니다.
- [LangChain](https://kiln.bricksum.com/docs/ko/sdks/langchain) ([md](https://kiln.bricksum.com/docs/ko/md/sdks/langchain.md)): LangChain의 ChatOpenAI를 Kiln으로 연결하는 방법입니다. 설치, 검증한 invoke와 stream 호출, 모델이 reasoning을 마친 뒤 메시지 content에 남는 내용을 설명합니다.
- [PydanticAI](https://kiln.bricksum.com/docs/ko/sdks/pydantic-ai) ([md](https://kiln.bricksum.com/docs/ko/md/sdks/pydantic-ai.md)): PydanticAI 에이전트를 Kiln에서 실행하는 방법입니다. 설치, 검증한 run_sync 호출, 이름이 바뀐 OpenAIModel 대신 쓰는 OpenAIChatModel, max_tokens를 넣는 위치를 설명합니다.
- [쿡북](https://kiln.bricksum.com/docs/ko/cookbook) ([md](https://kiln.bricksum.com/docs/ko/md/cookbook.md)): Kiln으로 만드는 검증된 레시피 모음입니다. 튜토리얼과 통합 가이드를 담고 있습니다.
- [스트리밍 채팅 CLI 빠른 시작](https://kiln.bricksum.com/docs/ko/cookbook/streaming-chat-cli) ([md](https://kiln.bricksum.com/docs/ko/md/cookbook/streaming-chat-cli.md)): Kiln의 응답을 토큰 단위로 실시간 출력하는 터미널 채팅 클라이언트를 만듭니다.
- [툴 호출, 처음부터 끝까지](https://kiln.bricksum.com/docs/ko/cookbook/tool-calling) ([md](https://kiln.bricksum.com/docs/ko/md/cookbook/tool-calling.md)): tool_choice "auto"로 둔 왕복 한 번입니다. 툴을 요청받아 실행하고 결과를 돌려보냅니다.

## Docs (日本語)

- [All pages of this locale in one Markdown file](https://kiln.bricksum.com/docs/ja/llms-full.txt)

- [ドキュメント](https://kiln.bricksum.com/docs/ja) ([md](https://kiln.bricksum.com/docs/ja/md/overview.md)): Kiln API は OpenAI SDK 互換の AI 推論 API です。base_url を変えるだけでモデルを呼び出せます。
- [クイックスタート](https://kiln.bricksum.com/docs/ja/quickstart) ([md](https://kiln.bricksum.com/docs/ja/md/quickstart.md)): API キーの発行から最初の API リクエストまでを手順どおりに案内します。
- [モデル](https://kiln.bricksum.com/docs/ja/models) ([md](https://kiln.bricksum.com/docs/ja/md/models.md)): 呼び出せるモデルを API で一覧取得する方法と、モデルごとのコンテキスト長・単価・対応機能。
- [API リファレンス](https://kiln.bricksum.com/docs/ja/api-reference) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference.md)): API リファレンスのハブです。Base URL、認証、エンドポイントの一覧（Chat Completions、Completions、Responses、Messages、models、key）、共通パラメータ、レスポンス、ストリーミング、エラー、制限。
- [Chat completions](https://kiln.bricksum.com/docs/ja/api-reference/chat-completions) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/chat-completions.md)): POST /chat/completions のリファレンスです。リクエストパラメータごとの対応状況、レスポンスとストリームチャンクのフィールド、エラー、制限、サンプルを説明します。
- [Responses](https://kiln.bricksum.com/docs/ja/api-reference/responses) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/responses.md)): POST /responses のリファレンスです。パラメータごとの対応状況、ストリームイベント、usage.cost が含まれる位置、エラー、サンプルを説明します。レスポンスは保存されず、previous_response_id は拒否されます。
- [Completions](https://kiln.bricksum.com/docs/ja/api-reference/completions) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/completions.md)): POST /completions のリファレンスです。リクエストパラメータごとの対応状況、レスポンスとストリームチャンクのフィールド、エラー、制限、サンプルを説明します。
- [ストリーミング](https://kiln.bricksum.com/docs/ja/streaming) ([md](https://kiln.bricksum.com/docs/ja/md/streaming.md)): Server-sent events（SSE）のチャンク形式、keepalive コメント、最終チャンクの usage と cost、切断時の動作。
- [使用量と課金](https://kiln.bricksum.com/docs/ja/usage-billing) ([md](https://kiln.bricksum.com/docs/ja/md/usage-billing.md)): usage オブジェクトと usage.cost、コストの計算方法、リクエスト ID ヘッダー、キーの確認、課金対象。
- [認証](https://kiln.bricksum.com/docs/ja/authentication) ([md](https://kiln.bricksum.com/docs/ja/md/authentication.md)): Bearer API キー、キー管理、組織・プロジェクトのスコープ。
- [組織と予算](https://kiln.bricksum.com/docs/ja/organizations) ([md](https://kiln.bricksum.com/docs/ja/md/organizations.md)): ロール、メンバーと招待、プロジェクトとキー、キーの支出上限と有効期限、月間予算、クレジットと有効期限、支払い、ステータス。
- [レート制限](https://kiln.bricksum.com/docs/ja/rate-limits) ([md](https://kiln.bricksum.com/docs/ja/md/rate-limits.md)): 組織単位の RPM/TPM 上限、429 の挙動、リトライガイド。
- [エラー](https://kiln.bricksum.com/docs/ja/errors) ([md](https://kiln.bricksum.com/docs/ja/md/errors.md)): エラーレスポンスの形式と主な HTTP ステータスコード。
- [連携](https://kiln.bricksum.com/docs/ja/integrations) ([md](https://kiln.bricksum.com/docs/ja/md/integrations.md)): 連携セクションの目次です。OpenAI SDK、Anthropic SDK、Vercel AI SDK、LangChain、PydanticAI はそれぞれ個別のページにあり、コーディングツール 6 種（Claude Code、Codex CLI、Hermes Agent、OpenCode、OpenClaw、Cursor）の設定はこのページにあります。
- [よくある質問](https://kiln.bricksum.com/docs/ja/faq) ([md](https://kiln.bricksum.com/docs/ja/md/faq.md)): 課金・クレジット、トライアル、モデル、SDK 互換性、ストリーミング、データ利用。
- [Messages](https://kiln.bricksum.com/docs/ja/api-reference/messages) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/messages.md)): POST /messages のリファレンス（Anthropic Messages 形式）です。ヘッダー、リクエストパラメータごとの変換状況、レスポンスフィールド、ストリームイベント、エラー、制限、サンプルを説明します。
- [Count tokens](https://kiln.bricksum.com/docs/ja/api-reference/count-tokens) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/count-tokens.md)): POST /messages/count_tokens のリファレンスです。リクエストフィールド、input_tokens レスポンス、クライアントが推定値を使うための 404、課金されない理由を説明します。
- [OpenAI からの移行](https://kiln.bricksum.com/docs/ja/migrating-from-openai) ([md](https://kiln.bricksum.com/docs/ja/md/migrating-from-openai.md)): OpenAI クライアントを Kiln に移行する方法です。変更する base_url・キー・モデル ID、既知の制限、本番移行前のチェックリストをまとめます。
- [List models](https://kiln.bricksum.com/docs/ja/api-reference/models) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/models.md)): GET /models のリファレンスです。ゲートウェイが返すモデルオブジェクト、一覧に含まれない情報、カタログが空のときのフォールバック、エラー、サンプルを説明します。
- [Key info](https://kiln.bricksum.com/docs/ja/api-reference/key) ([md](https://kiln.bricksum.com/docs/ja/md/api-reference/key.md)): GET /key のリファレンスです。キー自体で、そのキーの組織、プロジェクト、ステータス、支出上限、有効期限、支出額を照会します。エラーボディは {message, code} の形式で、拒否されるキーはすべて 401 を受け取ります。
- [OpenAI SDK](https://kiln.bricksum.com/docs/ja/sdks/openai) ([md](https://kiln.bricksum.com/docs/ja/md/sdks/openai.md)): Python、TypeScript、Go の公式 OpenAI クライアントから Kiln を呼び出す方法です。インストール、検証済みのリクエスト例、ツール呼び出しの往復に加え、OpenAI のスキーマにない cost フィールドと、max_tokens に含まれる reasoning の上限といった違いをまとめます。
- [Anthropic SDK](https://kiln.bricksum.com/docs/ja/sdks/anthropic) ([md](https://kiln.bricksum.com/docs/ja/md/sdks/anthropic.md)): Python と TypeScript の公式 Anthropic クライアントから Kiln を呼び出す方法です。/v1 を含まない base URL、content[0] が thinking ブロックである理由、usage オブジェクトに入る値と入らない値を説明します。
- [Vercel AI SDK](https://kiln.bricksum.com/docs/ja/sdks/vercel-ai) ([md](https://kiln.bricksum.com/docs/ja/md/sdks/vercel-ai.md)): Vercel AI SDK で Kiln からストリーミングする方法です。ai と openai-compatible のプロバイダーのインストール、検証済みの streamText 呼び出し、@ai-sdk/openai が適さない理由を説明します。
- [LangChain](https://kiln.bricksum.com/docs/ja/sdks/langchain) ([md](https://kiln.bricksum.com/docs/ja/md/sdks/langchain.md)): LangChain の ChatOpenAI を Kiln に向ける方法です。インストール、検証済みの invoke と stream の呼び出し、モデルが reasoning を終えたあとメッセージの content に残る内容を説明します。
- [PydanticAI](https://kiln.bricksum.com/docs/ja/sdks/pydantic-ai) ([md](https://kiln.bricksum.com/docs/ja/md/sdks/pydantic-ai.md)): PydanticAI のエージェントを Kiln で動かす方法です。インストール、検証済みの run_sync 呼び出し、名前が変わった OpenAIModel の代わりに使う OpenAIChatModel、max_tokens を書く場所を説明します。
- [クックブック](https://kiln.bricksum.com/docs/ja/cookbook) ([md](https://kiln.bricksum.com/docs/ja/md/cookbook.md)): Kiln で構築するための検証済みレシピ集です。チュートリアルとインテグレーションを収録しています。
- [ストリーミングチャット CLI クイックスタート](https://kiln.bricksum.com/docs/ja/cookbook/streaming-chat-cli) ([md](https://kiln.bricksum.com/docs/ja/md/cookbook/streaming-chat-cli.md)): Kiln の返信をトークン単位でリアルタイムに表示するターミナルチャットクライアントを作ります。
- [ツール呼び出し、最初から最後まで](https://kiln.bricksum.com/docs/ja/cookbook/tool-calling) ([md](https://kiln.bricksum.com/docs/ja/md/cookbook/tool-calling.md)): tool_choice を "auto" にした往復 1 回です。ツールを要求されたら実行し、結果を返します。