Span Attributes
groundcover adheres to the OpenTelemetry GenAI Semantic Conventions. The more attributes your spans carry, the richer the experience — cost analytics, prompt debugging, execution chain visibility.
For how groundcover handles older SDK attribute conventions (Era 1 gen_ai.prompt/gen_ai.completion and Era 2 log events), see OTel GenAI Conventions.
Core Attributes
gen_ai.provider.name
string
GenAI provider
gen_ai.operation.name
string
Operation type
gen_ai.request.model
string
Model requested
gen_ai.response.model
string
Model that responded
gen_ai.response.id
string
Completion ID
gen_ai.usage.input_tokens
int
Tokens consumed by the prompt
gen_ai.usage.output_tokens
int
Tokens generated in the response
gen_ai.response.finish_reasons
string[]
Why the model stopped
gen_ai.conversation.id
string
Session/thread identifier
Additional Attributes
Token Usage
gen_ai.usage.cache_read.input_tokens
int
Input tokens served from provider cache
gen_ai.usage.cache_creation.input_tokens
int
Input tokens written to provider cache
Content
gen_ai.input.messages
JSON
Structured input messages (role + parts)
gen_ai.output.messages
JSON
Structured output messages (role + parts)
gen_ai.system_instructions
JSON
System prompt, separate from chat history
gen_ai.tool.definitions
JSON
Tool schemas passed to the model
eBPF captures content automatically for supported providers. For SDK instrumentation, content capture is controlled by your SDK configuration — see Privacy Controls for details.
Agent
gen_ai.agent.name
string
Human-readable agent name
gen_ai.agent.id
string
Unique agent identifier
gen_ai.agent.description
string
Agent description
gen_ai.agent.version
string
Agent version
Tool Execution
gen_ai.tool.name
string
Name of the tool
gen_ai.tool.type
string
function, extension, datastore
gen_ai.tool.call.id
string
Tool call identifier
gen_ai.tool.call.arguments
JSON
Parameters passed to the tool
gen_ai.tool.call.result
JSON
Result returned by the tool
For request parameters, response metadata, and the complete specification, see the OpenTelemetry GenAI Semantic Conventions.
groundcover Enrichment
groundcover adds the following attributes at ingestion time. These are not part of the OTel specification — they're computed by groundcover for cost analytics and filtering.
Cost
gc.llm_cost.input
float
Cost of input tokens, computed from the model's pricing table
gc.llm_cost.output
float
Cost of output tokens, computed from the model's pricing table
gc.llm_cost.status
string
computed when pricing is available, unpriced when the model isn't in the pricing table
Cost is calculated per-span using gen_ai.request.model and gen_ai.provider.name to look up the correct pricing. When cache tokens are present (gen_ai.usage.cache_read.input_tokens), cached and non-cached portions are priced separately.
If gc.llm_cost.status shows unpriced, the model isn't in groundcover's pricing table yet. This typically resolves automatically as the table is updated. Contact us if it persists.
Filtering
groundcover assigns span_type:gen_ai automatically — eBPF spans from supported providers always receive it; SDK spans receive it when gen_ai.operation.name is present (best-effort). You do not need to set it yourself.
See Example Queries for patterns using these attributes.
Last updated