Observability

Complete Visibility Into Every AI Request

Track, analyze, and optimize your AI interactions with real-time tracing, intelligent session grouping, and comprehensive analytics dashboards.

Real-time

Live trace updates with 5-second polling

30min

Auto session timeout for grouping

90 days

Data retention for compliance

CSV Export

One-click data download

Three Pillars of Observability

Full visibility into your AI stack with tracing, sessions, and analytics working together.

Request Tracing

Every LLM request captured with full context, timing, tokens, costs, and security metadata.

  • Live streaming traces
  • Request/response payloads
  • Security event annotations
  • Sortable & filterable columns

Session Management

Group related requests into sessions automatically or with explicit session IDs.

  • Auto-sessions by user ID
  • Explicit X-Session-ID header
  • 30-minute inactivity timeout
  • Conversation timeline view

Analytics Dashboard

Visualize trends with charts for volume, latency, cost, tokens, and security.

  • 7/14/30/90 day views
  • Model cost breakdowns
  • Security threat trends
  • CSV export

Request Tracing

Every request through your proxy is captured as a trace with comprehensive metadata for debugging and analysis.

What's Captured

Trace ID
Timestamp
Model used
Latency (ms)
Input tokens
Output tokens
Cost (USD)
Status code
Session ID
User ID
Security events
Full payloads

Example Trace

{
  "trace_id": "tr_7kL9mN2pQr",
  "timestamp": "2025-01-13T10:23:45Z",
  "model": "gpt-4o",
  "latency_ms": 847,
  "tokens": {
    "input": 1234,
    "output": 567
  },
  "cost_usd": 0.0023,
  "status": "success",
  "session_id": "sess_abc123",
  "security": {
    "score": 0.12,
    "threats": []
  }
}

Smart Filtering

Filter by status, model, time range, cost, or security events

Full-Text Search

Search through request and response content

Payload Viewer

Inspect full request and response payloads

Session Management

Sessions group related requests together, making it easy to track multi-turn conversations and analyze user journeys.

Automatic Sessions

When you provide a user parameter, Bastio automatically groups requests from the same user into sessions based on a 30-minute inactivity timeout.

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    user="user_12345"  # Auto-session!
)
Explicit Sessions

For full control, use the X-Session-ID header to explicitly group requests into named sessions.

curl -X POST .../chat/completions \
  -H "X-API-Key: YOUR_KEY" \
  -H "X-Session-ID: checkout-flow-123" \
  -d '{"model":"gpt-4o",...}'

Session Timeline View

Click into any session to see a chronological timeline of all requests, making it easy to debug multi-turn conversations and understand user behavior.

Conversation flow
Time between messages
Security events per turn

Analytics Dashboard

Five interactive charts give you complete visibility into your AI usage, costs, and security posture.

Volume Chart

Track request volume over time with daily, weekly, and monthly breakdowns. Spot usage spikes and trends.

Latency Chart

Monitor P50, P95, and P99 latencies. Identify slow providers and optimize response times.

Token Chart

Analyze input vs output token distribution. Track context window usage and optimize prompts.

Cost Chart

Break down costs by model, user, and time period. Set budgets and track spend against goals.

Security Chart

Visualize threat detection trends, blocked requests, and security events over time.

CSV Export

Export any chart data to CSV for custom analysis, reporting, or compliance documentation.

Evaluation & Scoring

Rate and evaluate AI responses with custom rubrics. Build annotation queues and track quality metrics over time.

Quality Scoring

Score responses on a 1-10 scale directly from the trace view. Track average scores by model, user, or time period.

  • One-click scoring from trace detail
  • Custom evaluation criteria
  • Team collaboration

Annotation Queues

Add traces to annotation queues for systematic review. Perfect for QA workflows and building fine-tuning datasets.

  • Create multiple queues
  • Assign to team members
  • Export for fine-tuning

Included in Every Plan

Observability features are included with all Bastio plans. No additional charges for traces, sessions, or analytics.

FeatureFreeStarterProEnterprise
Request Tracing
Session Management
Analytics Dashboard
Data Retention7 days30 days90 daysCustom
CSV Export
Evaluation & Scoring

Start Observing Your AI Requests

Get full visibility into every AI interaction. Traces, sessions, and analytics are included with all Bastio plans at no extra cost.

Questions about observability? Contact us for a free consultation.