Glossary and References
On this page
This appendix separates terms used by OpenTelemetry, application architecture, evaluation systems, and observability backends. Similar words do not always represent the same data model.
Glossary
| Term | Meaning in this book |
|---|---|
| Agent | Software that uses a model to select or sequence actions toward a task. |
| Attribute | Typed key-value metadata attached to a span, resource, log, or metric point. |
| Baggage | Application key-value context propagated across boundaries; not a telemetry record. |
| Conversation | A set of related user and agent interactions, usually correlated by gen_ai.conversation.id. |
| Evaluation | A versioned judgment against an explicit criterion, produced by code, a human, or a model. |
| Event | A timestamped occurrence; in tracing, a span event belongs to one span. |
| Exemplar | A link from an aggregated metric observation to a representative trace. |
| Guardrail | A control that detects, transforms, blocks, or escalates an input, output, or action. |
| Instrumentation scope | Identity and version of the component that emitted telemetry. |
| Link | A causal or correlational relationship between a span and another span context that is not its parent. |
| Log | A timestamped record that can carry severity, body, attributes, trace ID, and span ID. |
| Metric | An aggregated measurement over time and a bounded set of dimensions. |
| Observation | Langfuse object for an operation inside a trace; it can represent several observation types. |
| OTLP | OpenTelemetry Protocol for exporting traces, metrics, and logs. |
| Resource | Attributes describing the entity that produced telemetry, such as a service. |
| Root span | A span without a parent in its trace. |
| Sampling | Selection of telemetry for recording or export; not a privacy mechanism. |
| Session | Backend or application grouping for related interactions. It is not an OpenTelemetry parent of a trace. |
| Span | A timed operation with context, attributes, events, links, status, and parent relationship. |
| SLI | A quantitative indicator of service behavior. |
| SLO | A target for an SLI over a defined population and time window. |
| Task | A business unit of work that can span requests, traces, workers, or conversation turns. |
| Trace | A graph of causally related spans that share a trace ID. |
| Trace context | Identifiers and flags propagated so spans can join a trace. |
OpenTelemetry
- OpenTelemetry documentation
- OpenTelemetry specification
- OpenTelemetry specification overview
- OpenTelemetry signals
- OpenTelemetry traces
- OpenTelemetry baggage
- OpenTelemetry context propagation
- OpenTelemetry sampling
- OpenTelemetry span links
- OpenTelemetry semantic conventions
- OpenTelemetry GenAI semantic conventions
- OpenTelemetry GenAI semantic conventions repository
- OpenTelemetry GenAI schema registry
- GenAI spans
- GenAI agent spans
- GenAI metrics
- GenAI OpenAI conventions
- OpenTelemetry GenAI content capture
- OpenTelemetry attribute requirement levels
- OpenTelemetry semantic convention version selection
- OpenTelemetry database semantic conventions
- OpenTelemetry messaging span conventions
- OpenTelemetry session semantic conventions
- OpenTelemetry Python
- OpenTelemetry Python propagation
- OpenTelemetry Python exporters
- OpenTelemetry SDK configuration
- OpenTelemetry Python testing utilities
- OpenTelemetry Collector
- OpenTelemetry Collector configuration
- OpenTelemetry Collector processors
- OpenTelemetry Collector tail sampling processor
- OpenTelemetry Collector resiliency
- OpenTelemetry Collector internal telemetry
- OpenTelemetry Metrics SDK cardinality limits and exemplars
- OpenTelemetry Trace SDK sampling decisions
- OpenTelemetry handling sensitive data
- Open proposal for GenAI task conventions
- W3C Trace Context
The GenAI conventions are under active development. Check the linked repository and schema version before implementing or upgrading attributes.
OpenAI
- OpenAI API documentation
- Responses API migration guide
- OpenAI Responses API reference
- Function calling
- Streaming responses
- Conversation state
- Safety best practices
- Production best practices
- OpenAI data controls
- OpenAI evaluation best practices
- OpenAI guardrails and human review
- OpenAI prompt caching
- OpenAI cost optimization
- Pricing
- OpenAI Python SDK
- OpenAI Python SDK retries and timeouts
- OpenAI Agents SDK guardrails
LangGraph and Langfuse
- LangGraph documentation
- LangGraph graph API
- LangGraph subgraphs
- LangGraph durable execution
- LangGraph streaming
- LangGraph event streaming
- LangChain multi-agent patterns
- LangGraph personal assistant with subagents
- Langfuse documentation
- Langfuse OpenTelemetry endpoint
- Langfuse native OpenTelemetry integration
- Langfuse self-hosting
- Langfuse Docker Compose deployment
- Langfuse data model
- Langfuse data masking
- Langfuse masking
- Langfuse data retention
- Langfuse sampling
- Langfuse sessions
- Langfuse users
- Langfuse metadata
- Langfuse Prompt Management
- Langfuse Prompt Management get started
- Langfuse prompt version control
- Langfuse prompt A/B testing
- Langfuse link prompts to traces
- Langfuse Prompt Management for OpenAI functions
- Langfuse scores
- Langfuse scores data model
- Langfuse scores via SDK
- Langfuse user feedback
- Langfuse Score Analytics
- Langfuse Score Configs
Security, privacy, and risk
- OWASP Top 10 for LLM Applications
- OWASP LLM01: Prompt Injection
- OWASP LLM06: Excessive Agency
- OWASP Logging Cheat Sheet
- OWASP Logging Cheat Sheet verification
- NIST AI Risk Management Framework
- NIST AI RMF Core
- NIST Generative AI Profile
- NIST Privacy Framework
- Microsoft Presidio
Evaluation and operations
- Langfuse evaluation overview
- Langfuse evaluation core concepts
- Langfuse LLM-as-a-Judge
- Langfuse code evaluators
- Langfuse annotation queues
- Langfuse datasets and experiments
- Langfuse datasets
- Langfuse experiments via SDK
- Langfuse experiments data model
- Langfuse experiments in CI/CD
- Langfuse versioned dataset experiments
- Who Validates the Validators?
- Google SRE: Monitoring Distributed Systems
- Google SRE: Service Level Objectives