Observe

Risicare's observability layer provides deep visibility into your AI agent's behavior.

Overview

Data Model

Session (user interaction)
  └── Trace (single request)
       └── Span (individual operation)
            ├── LLM Call
            ├── Tool Execution
            ├── Agent Decision
            └── Child Spans...

Key Metrics

Dashboard Views

Trace List

View all traces with filtering:

• Time range
• Status (success/error)
• Agent name
• Session ID
• Latency thresholds
• Cost thresholds

Trace Detail

Deep dive into a single trace:

• Span waterfall visualization
• LLM prompts and completions
• Tool inputs and outputs
• Error details and stack traces
• Cost breakdown by model

Analytics

Aggregate views:

• Trace volume over time
• Error rate trends
• Latency distributions
• Cost by model/agent
• Top error codes

Real-Time Streaming

View traces as they happen:

# Traces appear in dashboard within ~100ms
response = client.chat.completions.create(...)

Filtering and Search

Search

The search box takes plain text (minimum 2 characters) and matches on exactly two things:

• the root span name, as a case-insensitive substring
• the trace ID, as a prefix

It does not search agent names, models, content, or error codes.

Filters

Filtering is done with separate controls, not search terms:

To narrow by anything else — model, cost, latency, agent, error code — query the REST API and filter client-side, or use GET /api/v1/diagnoses?error_code=... for error codes specifically.

Data Retention

Retention is a fixed ClickHouse TTL that varies by data type, uniform across every project during beta:

Captured prompt and completion text lives in the span's attributes, so it is retained for the span's full 90 days — not the 30 days stated by earlier versions of these docs.

The retention_days project setting is reserved but not yet honored, and per-plan retention tiers will ship after beta. See Data Management for the full table.

Next Steps