helicone
🧊 Open source LLM observability platform. One line of code to monitor, evaluate, and experiment. YC W23 🍓
What is helicone?
🧊 Open source LLM observability platform. One line of code to monitor, evaluate, and experiment. YC W23 🍓
Overview of helicone
What it is
Helicone is an open source LLM observability platform and AI Gateway. It allows developers to monitor, evaluate, and experiment with large language models using a single line of code. The project is from Y Combinator W23 and is licensed under Apache-2.0.
Key features
- AI Gateway: Access 100+ AI models with one API key through the OpenAI API, with intelligent routing and automatic fallbacks.
- Quick integration: One line of code to log all requests from OpenAI, Anthropic, LangChain, Gemini, Vercel AI SDK, and more.
- Observe: Inspect and debug traces and sessions for agents, chatbots, document processing pipelines, and more.
- Analyze: Track metrics like cost, latency, quality, and export to PostHog for custom dashboards.
- Playground: Rapidly test and iterate on prompts, sessions, and traces in the UI.
- Prompt Management: Version prompts using production data and deploy through the AI Gateway without code changes.
- Fine-tune: Fine-tune with partners like OpenPipe or Autonomi.
- Enterprise Ready: SOC 2 and GDPR compliant.
Use cases
- Monitoring and debugging LLM-powered applications such as chatbots, agents, and document processing pipelines.
- Tracking cost and latency across multiple AI providers.
- Experimenting with prompts and model versions in a controlled environment.
- Building custom dashboards for LLM performance metrics.
Getting started
- Sign up at helicone.ai to get an API key and add credits.
- Update the
baseURLin your code tohttps://ai-gateway.helicone.aiand set your API key. - Use the OpenAI client as shown:
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://ai-gateway.helicone.ai",
apiKey: process.env.HELICONE_API_KEY,
});
const response = await client.chat.completions.create({
model: "gpt-4o-mini",
messages: [{ role: "user", content: "Hello!" }]
});
For self-hosting, use Docker:
git clone https://github.com/Helicone/helicone.git
cd docker
cp .env.example .env
./helicone-compose.sh helicone up
License and pricing
Licensed under Apache-2.0. The project is open source with a generous monthly free tier of 10,000 requests per month (no credit card required). Pricing for higher usage is not stated in the provided documentation.
這是什麼
Helicone 是一個開源的 LLM 可觀測性平台與 AI 閘道器。開發者只需一行程式碼,即可監控、評估和實驗大型語言模型。該專案來自 Y Combinator W23,採用 Apache-2.0 授權。
主要功能
- AI 閘道器:透過一個 API 金鑰存取超過 100 種 AI 模型,支援智慧路由與自動備援。
- 快速整合:一行程式碼即可記錄來自 OpenAI、Anthropic、LangChain、Gemini、Vercel AI SDK 等服務的所有請求。
- 觀察:檢查與除錯代理程式、聊天機器人、文件處理管線等的追蹤與工作階段。
- 分析:追蹤成本、延遲、品質等指標,並可匯出至 PostHog 建立自訂儀表板。
- 遊樂場:在 UI 中快速測試與迭代提示詞、工作階段與追蹤。
- 提示詞管理:使用生產資料對提示詞進行版本控制,並透過 AI 閘道器部署,無需修改程式碼。
- 微調:可與 OpenPipe 或 Autonomi 等合作夥伴進行微調。
- 企業就緒:符合 SOC 2 與 GDPR 規範。
使用情境
- 監控與除錯基於 LLM 的應用程式,例如聊天機器人、代理程式與文件處理管線。
- 跨多個 AI 提供者追蹤成本與延遲。
- 在受控環境中實驗提示詞與模型版本。
- 為 LLM 效能指標建立自訂儀表板。
快速開始
- 在 helicone.ai 註冊以取得 API 金鑰並加值。
- 將程式碼中的
baseURL改為https://ai-gateway.helicone.ai,並設定您的 API 金鑰。 - 使用 OpenAI 客戶端,如下所示:
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://ai-gateway.helicone.ai",
apiKey: process.env.HELICONE_API_KEY,
});
const response = await client.chat.completions.create({
model: "gpt-4o-mini",
messages: [{ role: "user", content: "Hello!" }]
});
若要自行託管,請使用 Docker:
git clone https://github.com/Helicone/helicone.git
cd docker
cp .env.example .env
./helicone-compose.sh helicone up
授權與價格
採用 Apache-2.0 授權。此專案為開源軟體,提供慷慨的每月免費額度(每月 10,000 次請求,無需信用卡)。更高使用量的價格在提供的文件中未說明。
Based on the project's public README and repository metadata.