# Integration Options for AI Metering {% hint style="info" %} **Python SDK:** `pip install revenium-python-sdk` — available on [PyPI](https://pypi.org/project/revenium-python-sdk/) with extras for each provider.\ **Node.js SDKs:** available on [npm](https://www.npmjs.com/org/revenium).\ **Go SDKs:** available on [GitHub](https://github.com/orgs/revenium/repositories?q=middleware+go). {% endhint %} ## Overview Revenium SDKs wrap your AI provider’s client library to automatically capture token usage, costs, latencies, and metadata — no changes to your API logic required. Data flows to the AI Analytics and Alerts dashboards in real time. *** ### Python SDK Install: `pip install revenium-python-sdk` The unified Python SDK supports all major AI providers through optional extras. Install only the providers you need: | Provider | Install Command | Chat Completions | Embeddings | | ------------------------------------- | ------------------------------------------------------ | :--------------: | :----------------------------------------------: | | OpenAI | `pip install "revenium-python-sdk[openai]"` | ✅ | ✅ | | Azure OpenAI | `pip install "revenium-python-sdk[openai]"` | ✅ | ✅ | | Anthropic | `pip install "revenium-python-sdk[anthropic]"` | ✅ | Anthropic has no embedding models | | Anthropic via AWS Bedrock | `pip install "revenium-python-sdk[anthropic]"` | ✅ | Anthropic has no embedding models | | Google Gemini (AI SDK) | `pip install "revenium-python-sdk[google-genai]"` | ✅ | ✅ | | Google Vertex AI (Enterprise) | `pip install "revenium-python-sdk[google-vertex]"` | ✅ | ✅ | | LiteLLM | `pip install "revenium-python-sdk[litellm]"` | ✅ | ✅ | | LiteLLM Proxy | `pip install "revenium-python-sdk[litellm-proxy]"` | ✅ | ✅ | | Ollama | `pip install "revenium-python-sdk[ollama]"` | ✅ | ✅ | | Perplexity (OpenAI-compatible) | `pip install "revenium-python-sdk[perplexity-openai]"` | ✅ | Perplexity does not currently support embeddings | | Perplexity (Native) | `pip install "revenium-python-sdk[perplexity-native]"` | ✅ | Perplexity does not currently support embeddings | | Fal.ai (image, video, audio metering) | `pip install "revenium-python-sdk[fal]"` | – | – | | **LangChain** | `pip install "revenium-python-sdk[langchain]"` | ✅ | ✅ | You can install multiple providers at once: `pip install "revenium-python-sdk[openai,anthropic,langchain]"` {% hint style="info" %} The Python SDK is available on [PyPI](https://pypi.org/project/revenium-python-sdk/). Source code and examples are on [GitHub](https://github.com/revenium/revenium-python-sdk). {% endhint %} *** ### Node.js SDKs | Provider | Package | Chat Completions | Embeddings | | ----------------------------- | --------------------------------------------------------------------------------------------- | :--------------: | :----------------------------------------------: | | OpenAI | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/openai`) | ✅ | ✅ | | Azure OpenAI | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/openai`) | ✅ | ✅ | | Anthropic | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/anthropic`) | ✅ | Anthropic has no embedding models | | Google Vertex AI (Enterprise) | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/google/vertex`) | ✅ | ✅ | | Google AI SDK | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/google/genai`) | ✅ | ✅ | | Perplexity | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/perplexity`) | ✅ | Perplexity does not currently support embeddings | | LiteLLM | [@revenium/middleware](https://www.npmjs.com/package/@revenium/middleware) (`/litellm`) | ✅ | ✅ | {% hint style="info" %} All Node.js SDKs are available on [npm](https://www.npmjs.com/org/revenium). {% endhint %} *** ### Go SDKs | Provider | Chat Completions | Embeddings | Image | Video | Audio | | ------------ | :--------------: | :----------------------------------------------: | :---: | :---: | :---: | | OpenAI | ✅ | ✅ | – | – | – | | Azure OpenAI | ✅ | ✅ | – | – | – | | Anthropic | ✅ | Anthropic has no embedding models | – | – | – | | Google | ✅ | ✅ | – | – | – | | Perplexity | ✅ | Perplexity does not currently support embeddings | – | – | – | | Fal.ai | – | – | ✅ | ✅ | ✅ | | Runway | – | – | – | ✅ | – | {% hint style="info" %} Go SDKs are available on [GitHub](https://github.com/orgs/revenium/repositories?q=middleware+go). {% endhint %} *** ### LangChain ```bash pip install "revenium-python-sdk[langchain]" ``` ```python from revenium_middleware.openai.langchain import wrap # bundled with the OpenAI module from langchain_openai import ChatOpenAI llm = wrap(ChatOpenAI(model="gpt-4o")) response = llm.invoke("Hello from Revenium!") ``` The `wrap()` function attaches a callback handler that reports usage data to Revenium's metering API. See the [SDK README](https://github.com/revenium/revenium-python-sdk) for additional examples. For LangChain apps already using OpenTelemetry, see [OpenTelemetry Integration](https://docs.revenium.io/opentelemetry-integration) as an alternative path. *** ### Other Framework Integrations | Framework / Platform | Language | Install | | --------------------- | -------- | ----------------------------------------------------------------------- | | n8n – OpenAI Agent | Node.js | [GitHub](https://github.com/revenium/revenium-middleware-openai-n8n) | | n8n – Anthropic Agent | Node.js | [GitHub](https://github.com/revenium/revenium-middleware-anthropic-n8n) | | Griptape | Python | Via griptape driver | | OpenTelemetry | Any | [See OTEL docs](https://docs.revenium.io/opentelemetry-integration) | *** ### AI Coding Assistants Revenium ingests telemetry from AI coding assistants via OpenTelemetry, tracking adoption and usage across your engineering team. | Tool | Status | | ----------- | ----------- | | Claude Code | ✅ Supported | | Gemini CLI | ✅ Supported | Setup: configure the tool's OTLP exporter to point at Revenium — see [OpenTelemetry Integration](https://docs.revenium.io/opentelemetry-integration). Usage data appears in the [AI Coding Dashboard](https://docs.revenium.io/ai-coding-dashboard). For a full list of collected data points, see the [AI Coding Data Reference](https://docs.revenium.io/ai-coding-data-reference). *** ### Multimodal Cost Tracking Revenium tracks costs across all AI modalities in a single view: [Completions](https://revenium.readme.io/reference/meter_ai_completion), [Images](https://revenium.readme.io/reference/meter_ai_images), [Video](https://revenium.readme.io/reference/meter_ai_video), and [Audio](https://revenium.readme.io/reference/meter_ai_audio). *** > **Using a different SDK or framework?** Revenium supports [direct API calls](#direct-api-integration) and [OpenTelemetry](https://docs.revenium.io/opentelemetry-integration) for any application. Contact for integration assistance. > > **Using an AI agent?** Revenium docs are available in Claude Code, Cursor, Windsurf, and other MCP-compatible agents via Context7. See [For AI Agents](https://docs.revenium.io/for-ai-agents). ## Usage Metadata Reference Each SDK accepts an optional `usage_metadata` object for billing, cost attribution, and alerting. The more fields you provide, the more granular your reporting. Key fields: `traceId`, `taskType`, `organizationId`, `subscriptionId`, `productId`, `agent`, `responseQualityScore`, `subscriber` (with `id`, `email`, `credential`). For field definitions, naming conventions, and examples, see each SDK's documentation on [npm](https://www.npmjs.com/org/revenium) or [PyPI](https://pypi.org/org/revenium/). For OTEL-based attribution, see the `revenium.*` attributes in the [OpenTelemetry Integration](https://docs.revenium.io/opentelemetry-integration#deep-attribution-with-revenium-attributes) docs. *** ### Direct API Integration For complete customization, use the metering API directly: * [Completions API](https://revenium.readme.io/reference/meter_ai_completion) | [Images API](https://revenium.readme.io/reference/meter_ai_images) | [Video API](https://revenium.readme.io/reference/meter_ai_video) | [Audio API](https://revenium.readme.io/reference/meter_ai_audio) * [Agent-friendly API documentation](https://docs.revenium.io/for-ai-agents) {% hint style="danger" %} **A 201 response does NOT confirm your payload was processed correctly.** The metering API returns `201 Created` even when your payload contains unrecognized fields. Misspelled or incorrect field names are silently ignored, which can result in **$0 cost calculations**. Always verify the calculated cost in the AI Analytics dashboard after your first integration. Common field-name mistakes: * `imageCount` (ignored) vs `actualImageCount` (correct) * `audioLength` (ignored) vs `audioDurationSeconds` (correct) * `stop_reason` (ignored) vs `stopReason` (correct) {% endhint %} {% hint style="info" %} **Required field: `stopReason`** — For completion metering, the `stopReason` field is **required**. Valid values: `END`, `END_SEQUENCE`, `TOKEN_LIMIT`, `COST_LIMIT`, `COMPLETION_LIMIT`, `ERROR`, `TIMEOUT`, `CANCELLED`. Omitting this field or sending an invalid value (e.g., `MAX_TOKENS`, `CONTENT_FILTER`, `STOP`) will result in a `400 Bad Request`. {% endhint %} {% hint style="warning" %} **Authentication errors return 403, not 401.** If your API key is missing or invalid, the metering API returns `403 Forbidden`. You will not receive a `401 Unauthorized` response. {% endhint %}