sideseat 0.1.0

OpenTelemetry SDK for AI/LLM observability with SideSeat

SideSeat Python SDK

OpenTelemetry SDK for AI/LLM observability with SideSeat.

Installation

pip install sideseat

Quick Start

from sideseat import SideSeatTelemetry

# Initialize telemetry
telemetry = SideSeatTelemetry()

# Configure exporters (chainable)
telemetry.setup_otlp_exporter(endpoint="http://localhost:5388/otel/default/v1/traces")
telemetry.setup_console_exporter()

# Your AI/LLM code here...

# Graceful shutdown (also registered with atexit for file exporters)
telemetry.shutdown()

Features

• Easy Setup: One-liner OpenTelemetry configuration
• Binary Encoding: Automatic base64 encoding for images/audio
• Multiple Exporters: OTLP, Console, and JSONL file output
• Framework Support: Works with Strands, LangChain, CrewAI, AutoGen
• Method Chaining: Fluent API for configuration
• Thread-Safe: Safe for concurrent use

Configuration

Custom Service Name

telemetry = SideSeatTelemetry(
    service_name="my-ai-app",
    service_version="1.0.0",
)

By default, the SDK auto-detects common AI frameworks (Strands, LangChain, CrewAI, AutoGen).

Disable Binary Encoding

telemetry = SideSeatTelemetry(encode_binary_as_base64=False)

Exporters

OTLP (recommended)

telemetry.setup_otlp_exporter(
    endpoint="http://localhost:5388/otel/default/v1/traces"
)

Console

telemetry.setup_console_exporter()

File (JSONL)

# Append mode (default) - safe for multiple runs
telemetry.setup_file_exporter(path="traces.jsonl")

# Overwrite mode
telemetry.setup_file_exporter(path="traces.jsonl", mode="w")

Metrics

telemetry.setup_meter(
    enable_console_exporter=True,
    enable_otlp_exporter=True,
)

Utilities

encode_value

Encode Python values for JSON, with base64 encoding for binary data:

from sideseat import encode_value

encode_value(b"binary")           # "YmluYXJ5" (base64)
encode_value({"key": b"value"})   # {"key": "dmFsdWU="}
encode_value({1, 2, 3})           # [1, 2, 3] (sorted list)

License

MIT