@raindrop-ai/bedrock

Raindrop integration for AWS Bedrock. Automatically captures Converse and InvokeModel calls by wrapping the BedrockRuntimeClient.

Installation

npm install @raindrop-ai/bedrock @aws-sdk/client-bedrock-runtime

Usage

import { createRaindropBedrock } from "@raindrop-ai/bedrock";
import { BedrockRuntimeClient, ConverseCommand } from "@aws-sdk/client-bedrock-runtime";

const raindrop = createRaindropBedrock({
  writeKey: "your-write-key",
  userId: "user-123",
});

const client = new BedrockRuntimeClient({ region: "us-east-1" });
const wrapped = raindrop.wrap(client);

const response = await wrapped.send(
  new ConverseCommand({
    modelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
    messages: [
      { role: "user", content: [{ text: "Hello!" }] },
    ],
  }),
);

await raindrop.flush();

What gets captured

• Converse API: input messages, output text, model ID, token usage (inputTokens/outputTokens)
• InvokeModel API: raw request/response bodies, model ID, token usage (Claude, Titan, and Llama formats)
• Errors: captured with OTLP error status, re-thrown to the caller

Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | writeKey | string | - | Raindrop API write key (omit to disable telemetry) | | endpoint | string | https://api.raindrop.ai/v1/ | API endpoint | | userId | string | - | Associate all events with a user | | convoId | string | - | Group events into a conversation | | projectId | string | - | Route events to a specific project (slug); omit for the default Production project | | debug | boolean | false | Enable verbose logging | | maxTextFieldChars | number | 1000000 | Per-field character cap for captured input/output text |

Projects

If your org has multiple projects, route events to a specific one by passing its slug as projectId:

const raindrop = createRaindropBedrock({
  writeKey: "your-write-key",
  projectId: "support-prod",
});

This sets the X-Raindrop-Project-Id header on every event. Omit it (or pass "default") to use your org's default Production project — the existing behavior. Single-project orgs need nothing new.

Payload size limits

Captured input/output text is capped at 1,000,000 characters per field by default and truncated with a ...[truncated by raindrop] marker (the result, marker included, never exceeds the limit). The cap is enforced before the text enters the event/span pipeline, so multi-MB prompts or responses cost the cap — not the payload — on your calling thread, and oversized events land truncated instead of being dropped at the ingest size limit. Tune it via maxTextFieldChars.

Testing

pnpm test

Tests use MSW to intercept HTTP requests — no real AWS calls are made.