Observability

Observability

Both runtimes provide observability features for production deployments.

Rust: Structured Logging

ai-lib-rust uses the tracing ecosystem:

use tracing_subscriber;

// Enable logging
tracing_subscriber::init();

// All AI-Lib operations emit structured log events
let client = AiClient::from_model("openai/gpt-4o").await?;

Log levels:

• INFO — Request/response summaries
• DEBUG — Protocol loading, pipeline stages
• TRACE — Individual frames, JSONPath matches

Rust: Call Statistics

Every request returns usage statistics:

let (response, stats) = client.chat()
    .user("Hello")
    .execute_with_stats()
    .await?;

println!("Model: {}", stats.model);
println!("Provider: {}", stats.provider);
println!("Prompt tokens: {}", stats.prompt_tokens);
println!("Completion tokens: {}", stats.completion_tokens);
println!("Total tokens: {}", stats.total_tokens);
println!("Latency: {}ms", stats.latency_ms);

Python: Metrics (Prometheus)

from ai_lib_python.telemetry import MetricsCollector

metrics = MetricsCollector()

client = await AiClient.builder() \
    .model("openai/gpt-4o") \
    .metrics(metrics) \
    .build()

# After some requests...
prometheus_text = metrics.export_prometheus()

Tracked metrics:

• ai_lib_requests_total — Request count by model/provider
• ai_lib_request_duration_seconds — Latency histogram
• ai_lib_tokens_total — Token usage by type
• ai_lib_errors_total — Error count by type

Python: Distributed Tracing (OpenTelemetry)

from ai_lib_python.telemetry import Tracer

tracer = Tracer(
    service_name="my-app",
    endpoint="http://jaeger:4317",
)

client = await AiClient.builder() \
    .model("openai/gpt-4o") \
    .tracer(tracer) \
    .build()

Traces include spans for:

• Protocol loading
• Request compilation
• HTTP transport
• Pipeline processing
• Event mapping

Python: Health Monitoring

from ai_lib_python.telemetry import HealthChecker

health = HealthChecker()
status = await health.check()

print(f"Healthy: {status.is_healthy}")
print(f"Details: {status.details}")

Python: User Feedback

Collect feedback on AI responses:

from ai_lib_python.telemetry import FeedbackCollector

feedback = FeedbackCollector()

# After getting a response
feedback.record(
    request_id=stats.request_id,
    rating=5,
    comment="Helpful response",
)

Resilience Observability

Monitor circuit breaker and rate limiter state:

// Rust
let state = client.circuit_state(); // Closed, Open, HalfOpen
let inflight = client.current_inflight();

# Python
signals = client.signals_snapshot()
print(f"Circuit: {signals.circuit_state}")
print(f"Inflight: {signals.current_inflight}")