llm-observatory-collector

OpenTelemetry collector with LLM-specific processors for LLM Observatory.

Overview

High-performance telemetry collector designed specifically for LLM applications:

• OTLP Receiver: gRPC and HTTP endpoints for OpenTelemetry data
• LLM-Aware Processing: Token counting, cost calculation, PII redaction
• Intelligent Sampling: Head and tail sampling strategies
• Context Propagation: Distributed tracing support
• Metric Enrichment: Automatic LLM-specific metadata

Features

• 100k+ spans/sec: High-throughput processing
• PII Redaction: Optional privacy-preserving transformations
• Cost Calculation: Real-time LLM cost tracking
• Multiple Backends: TimescaleDB, Tempo, Loki support

Usage

# Run collector
llm-observatory-collector --config config.yaml

Configuration:

receivers:
  otlp:
    grpc:
      endpoint: 0.0.0.0:4317
    http:
      endpoint: 0.0.0.0:4318

processors:
  llm_enrichment:
    enable_cost: true
    enable_pii_redaction: false

exporters:
  timescaledb:
    connection_string: postgres://...

Documentation

See the collector documentation for detailed configuration.

License

Apache-2.0