Raindrop

Raindrop is an AI observability platform for monitoring and evaluating LLM applications. With Raindrop, you can track conversations, analyze model performance, and debug AI workflows.

Step 1: Get your Raindrop write key

In Raindrop, navigate to your project settings:

1. Log in to your Raindrop account
2. Go to Settings and find your project’s Write Key
3. Copy the write key

Step 2: Enable Broadcast in OpenRouter

Go to Settings > Observability and toggle Enable Broadcast.

Step 3: Configure Raindrop

Click the edit icon next to Raindrop and enter:

• Write Key: Your Raindrop project write key
• Base URL (optional): Default is https://api.raindrop.ai. Change only if using a custom endpoint

Step 4: Test and save

Click Test Connection to verify the setup. The configuration only saves if the test passes.

Step 5: Send a test trace

Make an API request through OpenRouter and view the event in your Raindrop dashboard under Events.

Each event includes:

• User Input: The latest user message from the conversation
• Assistant Output: The model’s completion text
• Properties: Token counts, cost, latency, model, provider, and finish reason

Custom Metadata

Raindrop receives events with custom metadata included as event properties. Use the trace field to attach additional context to your events.

Supported Metadata Keys

Example

1
{
2
  "model": "anthropic/claude-sonnet-4",
3
  "messages": [{ "role": "user", "content": "What time is it?" }],
4
  "user": "user_35",
5
  "session_id": "session_abc",
6
  "trace": {
7
    "trace_name": "Time Check",
8
    "feature": "assistant"
9
  }
10
}

Every key inside trace that is not in the table above is forwarded as-is (e.g. feature becomes the property feature).

Additional Context

• The user field maps to Raindrop’s user_id for user-level analytics
• The session_id field maps to convo_id for grouping conversation turns
• Events include system properties like model, provider, total_cost, prompt_tokens, completion_tokens, duration_ms, and finish_reason

Privacy Mode

When Privacy Mode is enabled for this destination, the input and output fields are excluded from events. All other event data — token usage, costs, timing, model information, and custom metadata — is still sent normally. See Privacy Mode for details.