Skip to content

Agent Attributes Reference

Complete reference for all agenttel.agentic.* attributes used to instrument AI agent runtimes. These attributes capture agent lifecycle, reasoning traces, orchestration patterns, cost aggregation, quality signals, and safety guardrails. They are distinct from the core agenttel.* attributes (see attribute-dictionary.md) which cover the backend services that agents interact with.

All attributes live in the agenttel.agentic.* namespace and are defined as AttributeKey constants in io.agenttel.api.attributes.AgenticAttributes.

Quick navigation: Agent Identity | Invocation | Task Tracking | Step / Reasoning | Orchestration | Handoff / Delegation | Cost Aggregation | Quality Signals | Guardrail / Safety | Memory Access | Human-in-the-Loop | Code Execution | Evaluation / Scoring | Error Classification | Agent Capabilities | Conversation / Message Tracking | Retrieval (RAG) | Reranker | Span Name Reference | Java Constant Reference


Alphabetical Index

All agenttel.agentic.* attributes sorted alphabetically. Click any key to jump to its category.

Attribute Key Type Category
agenttel.agentic.agent.framework String Agent Identity
agenttel.agentic.agent.name String Agent Identity
agenttel.agentic.agent.type String Agent Identity
agenttel.agentic.agent.version String Agent Identity
agenttel.agentic.capability.system_prompt_hash String Agent Capabilities
agenttel.agentic.capability.tool_count Long Agent Capabilities
agenttel.agentic.capability.tools String[] Agent Capabilities
agenttel.agentic.code.exit_code Long Code Execution
agenttel.agentic.code.language String Code Execution
agenttel.agentic.code.sandboxed Boolean Code Execution
agenttel.agentic.code.status String Code Execution
agenttel.agentic.conversation.id String Conversation / Message Tracking
agenttel.agentic.conversation.message_count Long Conversation / Message Tracking
agenttel.agentic.conversation.speaker_role String Conversation / Message Tracking
agenttel.agentic.conversation.turn Long Conversation / Message Tracking
agenttel.agentic.cost.cached_read_tokens Long Cost Aggregation
agenttel.agentic.cost.cached_write_tokens Long Cost Aggregation
agenttel.agentic.cost.input_tokens Long Cost Aggregation
agenttel.agentic.cost.llm_calls Long Cost Aggregation
agenttel.agentic.cost.output_tokens Long Cost Aggregation
agenttel.agentic.cost.reasoning_tokens Long Cost Aggregation
agenttel.agentic.cost.total_usd Double Cost Aggregation
agenttel.agentic.error.category String Error Classification
agenttel.agentic.error.retryable Boolean Error Classification
agenttel.agentic.error.source String Error Classification
agenttel.agentic.eval.criteria String Evaluation / Scoring
agenttel.agentic.eval.feedback String Evaluation / Scoring
agenttel.agentic.eval.score Double Evaluation / Scoring
agenttel.agentic.eval.scorer_name String Evaluation / Scoring
agenttel.agentic.eval.type String Evaluation / Scoring
agenttel.agentic.guardrail.action String Guardrail / Safety
agenttel.agentic.guardrail.name String Guardrail / Safety
agenttel.agentic.guardrail.reason String Guardrail / Safety
agenttel.agentic.guardrail.triggered Boolean Guardrail / Safety
agenttel.agentic.handoff.chain_depth Long Handoff / Delegation
agenttel.agentic.handoff.from_agent String Handoff / Delegation
agenttel.agentic.handoff.reason String Handoff / Delegation
agenttel.agentic.handoff.to_agent String Handoff / Delegation
agenttel.agentic.human.checkpoint_type String Human-in-the-Loop
agenttel.agentic.human.decision String Human-in-the-Loop
agenttel.agentic.human.wait_ms Long Human-in-the-Loop
agenttel.agentic.invocation.goal String Invocation
agenttel.agentic.invocation.id String Invocation
agenttel.agentic.invocation.max_steps Long Invocation
agenttel.agentic.invocation.status String Invocation
agenttel.agentic.invocation.steps Long Invocation
agenttel.agentic.memory.items Long Memory Access
agenttel.agentic.memory.operation String Memory Access
agenttel.agentic.memory.store_type String Memory Access
agenttel.agentic.orchestration.aggregation String Orchestration
agenttel.agentic.orchestration.coordinator_id String Orchestration
agenttel.agentic.orchestration.parallel_branches Long Orchestration
agenttel.agentic.orchestration.pattern String Orchestration
agenttel.agentic.orchestration.stage Long Orchestration
agenttel.agentic.orchestration.total_stages Long Orchestration
agenttel.agentic.quality.eval_score Double Quality Signals
agenttel.agentic.quality.goal_achieved Boolean Quality Signals
agenttel.agentic.quality.human_interventions Long Quality Signals
agenttel.agentic.quality.loop_detected Boolean Quality Signals
agenttel.agentic.quality.loop_iterations Long Quality Signals
agenttel.agentic.reranker.input_documents Long Reranker
agenttel.agentic.reranker.model String Reranker
agenttel.agentic.reranker.output_documents Long Reranker
agenttel.agentic.reranker.top_score Double Reranker
agenttel.agentic.retrieval.document_count Long Retrieval (RAG)
agenttel.agentic.retrieval.query String Retrieval (RAG)
agenttel.agentic.retrieval.relevance_score_avg Double Retrieval (RAG)
agenttel.agentic.retrieval.relevance_score_min Double Retrieval (RAG)
agenttel.agentic.retrieval.store_type String Retrieval (RAG)
agenttel.agentic.retrieval.top_k Long Retrieval (RAG)
agenttel.agentic.step.iteration Long Step / Reasoning
agenttel.agentic.step.number Long Step / Reasoning
agenttel.agentic.step.tool_name String Step / Reasoning
agenttel.agentic.step.tool_status String Step / Reasoning
agenttel.agentic.step.type String Step / Reasoning
agenttel.agentic.task.depth Long Task Tracking
agenttel.agentic.task.id String Task Tracking
agenttel.agentic.task.name String Task Tracking
agenttel.agentic.task.parent_id String Task Tracking
agenttel.agentic.task.status String Task Tracking

Agent Identity

Attributes that identify the agent instance and its framework context. Set on the root invocation span and inherited by child spans.

Key Type Values / Description
agenttel.agentic.agent.name String Agent identifier (e.g., "research-assistant", "code-reviewer")
agenttel.agentic.agent.type String Agent archetype. See Agent Type values
agenttel.agentic.agent.framework String Framework name (e.g., "langchain4j", "spring-ai", "autogen")
agenttel.agentic.agent.version String Agent version string (e.g., "1.2.0", "2024-03-beta")

Agent Type Values

Value Description
single Standalone agent operating independently
orchestrator Agent that coordinates other agents
worker Agent that executes tasks assigned by an orchestrator
evaluator Agent that evaluates outputs from other agents
critic Agent that critiques and suggests improvements
router Agent that routes requests to appropriate agents

Invocation

Attributes that describe a single agent invocation -- from goal to completion. Set on the root span created by AgentTracer.invoke().

Key Type Values / Description
agenttel.agentic.invocation.id String Unique invocation identifier (UUID or application-defined)
agenttel.agentic.invocation.goal String Goal description given to the agent (e.g., "Summarize quarterly earnings")
agenttel.agentic.invocation.status String Terminal status. See Invocation Status values
agenttel.agentic.invocation.steps Long Total number of steps executed during this invocation
agenttel.agentic.invocation.max_steps Long Maximum steps guardrail -- agent stops if this limit is reached

Invocation Status Values

Value Description
success Agent completed the goal successfully
failure Agent failed to complete the goal
timeout Agent exceeded its time or step limit
escalated Agent escalated to a human or higher-level agent
human_intervened A human intervened during execution

Task Tracking

Attributes for tracking task decomposition within an invocation. Tasks can nest arbitrarily deep, forming a tree structure.

Key Type Values / Description
agenttel.agentic.task.id String Task identifier (unique within the invocation)
agenttel.agentic.task.name String Human-readable task name (e.g., "extract-entities", "validate-schema")
agenttel.agentic.task.status String Task status. See Task Status values
agenttel.agentic.task.parent_id String Parent task ID for nested task hierarchies
agenttel.agentic.task.depth Long Depth level in the task decomposition tree (root = 0)

Task Status Values

Value Description
in_progress Task is currently being executed
completed Task finished successfully
failed Task failed
delegated Task was delegated to another agent

Step / Reasoning

Attributes that capture individual reasoning steps within an agent's execution loop (e.g., ReAct thought-action-observation cycles).

Key Type Values / Description
agenttel.agentic.step.number Long Sequential step number within the invocation (1-indexed)
agenttel.agentic.step.type String Step type in the reasoning loop. See Step Type values
agenttel.agentic.step.iteration Long Iteration number when the agent loops (e.g., retry or refinement cycle)
agenttel.agentic.step.tool_name String Name of the tool called during this step (e.g., "web_search", "calculator")
agenttel.agentic.step.tool_status String Outcome of the tool call. See Tool Status values

Step Type Values

Value Description
thought Agent reasoning / chain-of-thought
action Agent taking an action (tool call, API request)
observation Agent processing the result of an action
evaluation Agent evaluating progress toward the goal
revision Agent revising its approach based on evaluation

Tool Status Values

Value Description
success Tool executed successfully
error Tool returned an error
timeout Tool call timed out

Orchestration

Attributes describing multi-agent orchestration patterns. Set on session-level spans created by AgentTracer.orchestrate().

Key Type Values / Description
agenttel.agentic.orchestration.pattern String Orchestration pattern. See Orchestration Pattern values
agenttel.agentic.orchestration.coordinator_id String Identifier of the coordinating agent
agenttel.agentic.orchestration.stage Long Current stage number (for sequential patterns)
agenttel.agentic.orchestration.total_stages Long Total number of stages (for sequential patterns)
agenttel.agentic.orchestration.parallel_branches Long Number of parallel branches executing concurrently
agenttel.agentic.orchestration.aggregation String Strategy for aggregating results from parallel branches (e.g., "merge", "vote", "first_success")

Orchestration Pattern Values

Value Description
react ReAct (Reasoning + Acting) single-agent loop
sequential Agents execute in a predefined sequence (pipeline)
parallel Agents execute concurrently on the same input
handoff Agent-to-agent handoff (one at a time)
orchestrator_workers Central orchestrator dispatches tasks to worker agents
evaluator_optimizer Iterative loop between a generator and evaluator agent
group_chat Multiple agents converse in a shared context
swarm Decentralized multi-agent collaboration
hierarchical Tree-structured agent delegation

Handoff / Delegation

Attributes that track agent-to-agent handoffs and delegation chains.

Key Type Values / Description
agenttel.agentic.handoff.from_agent String Name of the source agent initiating the handoff
agenttel.agentic.handoff.to_agent String Name of the target agent receiving the handoff
agenttel.agentic.handoff.reason String Reason for the handoff (e.g., "requires domain expertise", "capability mismatch")
agenttel.agentic.handoff.chain_depth Long Depth in the handoff chain (0 = original agent, 1 = first handoff, etc.)

Cost Aggregation

Attributes for tracking cumulative LLM costs across an agent invocation. Typically set on the root invocation span after all LLM calls complete.

Key Type Values / Description
agenttel.agentic.cost.total_usd Double Total cost in US dollars for the invocation
agenttel.agentic.cost.input_tokens Long Total input (prompt) tokens consumed across all LLM calls
agenttel.agentic.cost.output_tokens Long Total output (completion) tokens generated across all LLM calls
agenttel.agentic.cost.llm_calls Long Number of LLM API calls made during the invocation
agenttel.agentic.cost.reasoning_tokens Long Extended thinking / chain-of-thought tokens (e.g., Claude extended thinking, OpenAI o1 reasoning)
agenttel.agentic.cost.cached_read_tokens Long Tokens served from prompt cache reads
agenttel.agentic.cost.cached_write_tokens Long Tokens written to prompt cache

Quality Signals

Attributes that capture quality and reliability indicators for the agent's execution.

Key Type Values / Description
agenttel.agentic.quality.goal_achieved Boolean Whether the agent achieved its stated goal
agenttel.agentic.quality.human_interventions Long Number of times a human intervened during execution
agenttel.agentic.quality.loop_detected Boolean Whether the agent was detected to be looping
agenttel.agentic.quality.loop_iterations Long Number of loop iterations detected before breaking
agenttel.agentic.quality.eval_score Double Evaluation score for the invocation outcome (0.0 = worst, 1.0 = best)

Guardrail / Safety

Attributes that record guardrail activations and safety interventions.

Key Type Values / Description
agenttel.agentic.guardrail.triggered Boolean Whether a guardrail was activated
agenttel.agentic.guardrail.name String Name of the guardrail that triggered (e.g., "pii-filter", "cost-limit")
agenttel.agentic.guardrail.action String Action taken. See Guardrail Action values
agenttel.agentic.guardrail.reason String Explanation of why the guardrail triggered

Guardrail Action Values

Value Description
block Request or action was blocked entirely
warn Warning was emitted but execution continued
log Event was logged for audit without interrupting execution
escalate Issue was escalated to a human or higher-level agent

Memory Access

Attributes for tracking agent interactions with external memory stores (vector databases, key-value stores, etc.).

Key Type Values / Description
agenttel.agentic.memory.operation String Memory operation type. See Memory Operation values
agenttel.agentic.memory.store_type String Type of memory store (e.g., "vector", "relational", "key_value", "graph")
agenttel.agentic.memory.items Long Number of items read, written, or deleted

Memory Operation Values

Value Description
read Reading from memory
write Writing to memory
delete Deleting from memory
search Searching / querying memory

Human-in-the-Loop

Attributes for tracking human checkpoints and interventions within agent workflows.

Key Type Values / Description
agenttel.agentic.human.checkpoint_type String Type of human checkpoint. See Checkpoint Type values
agenttel.agentic.human.wait_ms Long Time the agent waited for human input, in milliseconds
agenttel.agentic.human.decision String The decision or input provided by the human

Checkpoint Type Values

Value Description
approval Human must approve before the agent proceeds
feedback Human provides feedback on agent output
correction Human corrects agent output or behavior
decision Human makes a decision the agent cannot make autonomously

Code Execution

Attributes for tracking code execution steps (e.g., Python sandboxes, shell commands, REPL interactions).

Key Type Values / Description
agenttel.agentic.code.language String Programming language of the executed code (e.g., "python", "javascript", "bash")
agenttel.agentic.code.status String Execution outcome. See Code Status values
agenttel.agentic.code.exit_code Long Process exit code (0 = success)
agenttel.agentic.code.sandboxed Boolean Whether the code ran in a sandboxed / isolated environment

Code Status Values

Value Description
success Code executed successfully
error Code execution produced an error
timeout Code execution timed out

Evaluation / Scoring

Attributes for evaluation and scoring steps, where an evaluator agent or heuristic grades an output.

Key Type Values / Description
agenttel.agentic.eval.scorer_name String Identifier of the evaluation scorer (e.g., "faithfulness", "relevance", "toxicity")
agenttel.agentic.eval.criteria String Description of the evaluation criteria applied
agenttel.agentic.eval.score Double Evaluation score (0.0 = worst, 1.0 = best)
agenttel.agentic.eval.feedback String Free-text feedback from the evaluator
agenttel.agentic.eval.type String Evaluation method. See Evaluation Type values

Evaluation Type Values

Value Description
llm_judge LLM used as a judge to evaluate output quality
heuristic Rule-based or algorithmic evaluation
human Human evaluator
custom Custom evaluation logic

Error Classification

Attributes that classify errors by their source and recoverability, enabling targeted retry and escalation policies.

Key Type Values / Description
agenttel.agentic.error.source String Origin of the error. See Error Source values
agenttel.agentic.error.category String Application-defined error category (e.g., "rate_limit", "invalid_input", "permission_denied")
agenttel.agentic.error.retryable Boolean Whether the error is eligible for automatic retry

Error Source Values

Value Description
llm Error from the LLM provider (e.g., rate limit, context length exceeded)
tool Error from a tool execution
agent Error in agent logic itself
guardrail Error triggered by a guardrail
timeout Timeout error
network Network connectivity error

Agent Capabilities

Attributes describing what the agent is configured to do -- its available tools and system prompt version. Useful for correlating behavior changes with configuration changes.

Key Type Values / Description
agenttel.agentic.capability.tools String[] List of tool names available to the agent (e.g., ["web_search", "calculator", "code_interpreter"])
agenttel.agentic.capability.tool_count Long Number of tools available to the agent
agenttel.agentic.capability.system_prompt_hash String Hash of the system prompt, for tracking prompt version changes (e.g., SHA-256 prefix)

Conversation / Message Tracking

Attributes for tracking multi-turn conversations and message flow within agent interactions.

Key Type Values / Description
agenttel.agentic.conversation.id String Conversation identifier (groups multiple turns)
agenttel.agentic.conversation.turn Long Turn number within the conversation (1-indexed)
agenttel.agentic.conversation.message_count Long Number of messages exchanged in this turn
agenttel.agentic.conversation.speaker_role String Role of the speaker. See Speaker Role values

Speaker Role Values

Value Description
agent Message from the agent
user Message from the user
system System message (e.g., system prompt, context injection)

Retrieval (RAG)

Attributes for Retrieval-Augmented Generation (RAG) operations, tracking document retrieval quality and parameters.

Key Type Values / Description
agenttel.agentic.retrieval.query String The query string used for retrieval
agenttel.agentic.retrieval.store_type String Type of vector or document store (e.g., "pinecone", "weaviate", "elasticsearch", "pgvector")
agenttel.agentic.retrieval.document_count Long Number of documents retrieved
agenttel.agentic.retrieval.top_k Long The K parameter used for top-K retrieval
agenttel.agentic.retrieval.relevance_score_avg Double Average relevance score across retrieved documents (0.0 - 1.0)
agenttel.agentic.retrieval.relevance_score_min Double Minimum relevance score among retrieved documents (0.0 - 1.0)

Reranker

Attributes for reranking operations that refine retrieval results using a cross-encoder or other reranking model.

Key Type Values / Description
agenttel.agentic.reranker.model String Reranker model name (e.g., "cohere-rerank-v3", "bge-reranker-large")
agenttel.agentic.reranker.input_documents Long Number of documents passed to the reranker
agenttel.agentic.reranker.output_documents Long Number of documents returned after reranking
agenttel.agentic.reranker.top_score Double Highest relevance score from the reranker (0.0 - 1.0)

Span Name Reference

Standard span names created by the AgentTel agentic tracing API, along with their entry points and the attribute categories typically set on each span.

Span Name API Entry Point Primary Attribute Categories
invoke_agent AgentTracer.invoke() Agent Identity, Invocation, Quality Signals, Cost Aggregation
agenttel.agentic.step AgentInvocation.step() / AgentInvocation.beginStep() Step / Reasoning
agenttel.agentic.tool_call AgentInvocation.toolCall() Step (step.tool_name, step.tool_status)
agenttel.agentic.task AgentInvocation.task() Task Tracking
agenttel.agentic.handoff AgentInvocation.handoff() Handoff / Delegation
agenttel.agentic.human_input AgentInvocation.humanCheckpoint() Human-in-the-Loop
agenttel.agentic.code_execution AgentInvocation.codeExecution() Code Execution
agenttel.agentic.evaluate AgentInvocation.evaluate() Evaluation / Scoring
agenttel.agentic.retriever AgentInvocation.retrieve() Retrieval (RAG)
agenttel.agentic.reranker AgentInvocation.rerank() Reranker
agenttel.agentic.guardrail AgentInvocation.guardrail() Guardrail / Safety
agenttel.agentic.memory AgentTracer.memory() Memory Access
agenttel.agentic.session AgentTracer.orchestrate() Orchestration

Java Constant Reference

All attributes are defined as AttributeKey constants in io.agenttel.api.attributes.AgenticAttributes. The naming convention is AgenticAttributes.CATEGORY_FIELD.

Java Constant Attribute Key
AgenticAttributes.AGENT_NAME agenttel.agentic.agent.name
AgenticAttributes.AGENT_TYPE agenttel.agentic.agent.type
AgenticAttributes.AGENT_FRAMEWORK agenttel.agentic.agent.framework
AgenticAttributes.AGENT_VERSION agenttel.agentic.agent.version
AgenticAttributes.INVOCATION_ID agenttel.agentic.invocation.id
AgenticAttributes.INVOCATION_GOAL agenttel.agentic.invocation.goal
AgenticAttributes.INVOCATION_STATUS agenttel.agentic.invocation.status
AgenticAttributes.INVOCATION_STEPS agenttel.agentic.invocation.steps
AgenticAttributes.INVOCATION_MAX_STEPS agenttel.agentic.invocation.max_steps
AgenticAttributes.TASK_ID agenttel.agentic.task.id
AgenticAttributes.TASK_NAME agenttel.agentic.task.name
AgenticAttributes.TASK_STATUS agenttel.agentic.task.status
AgenticAttributes.TASK_PARENT_ID agenttel.agentic.task.parent_id
AgenticAttributes.TASK_DEPTH agenttel.agentic.task.depth
AgenticAttributes.STEP_NUMBER agenttel.agentic.step.number
AgenticAttributes.STEP_TYPE agenttel.agentic.step.type
AgenticAttributes.STEP_ITERATION agenttel.agentic.step.iteration
AgenticAttributes.STEP_TOOL_NAME agenttel.agentic.step.tool_name
AgenticAttributes.STEP_TOOL_STATUS agenttel.agentic.step.tool_status
AgenticAttributes.ORCHESTRATION_PATTERN agenttel.agentic.orchestration.pattern
AgenticAttributes.ORCHESTRATION_COORDINATOR_ID agenttel.agentic.orchestration.coordinator_id
AgenticAttributes.ORCHESTRATION_STAGE agenttel.agentic.orchestration.stage
AgenticAttributes.ORCHESTRATION_TOTAL_STAGES agenttel.agentic.orchestration.total_stages
AgenticAttributes.ORCHESTRATION_PARALLEL_BRANCHES agenttel.agentic.orchestration.parallel_branches
AgenticAttributes.ORCHESTRATION_AGGREGATION agenttel.agentic.orchestration.aggregation
AgenticAttributes.HANDOFF_FROM_AGENT agenttel.agentic.handoff.from_agent
AgenticAttributes.HANDOFF_TO_AGENT agenttel.agentic.handoff.to_agent
AgenticAttributes.HANDOFF_REASON agenttel.agentic.handoff.reason
AgenticAttributes.HANDOFF_CHAIN_DEPTH agenttel.agentic.handoff.chain_depth
AgenticAttributes.COST_TOTAL_USD agenttel.agentic.cost.total_usd
AgenticAttributes.COST_INPUT_TOKENS agenttel.agentic.cost.input_tokens
AgenticAttributes.COST_OUTPUT_TOKENS agenttel.agentic.cost.output_tokens
AgenticAttributes.COST_LLM_CALLS agenttel.agentic.cost.llm_calls
AgenticAttributes.COST_REASONING_TOKENS agenttel.agentic.cost.reasoning_tokens
AgenticAttributes.COST_CACHED_READ_TOKENS agenttel.agentic.cost.cached_read_tokens
AgenticAttributes.COST_CACHED_WRITE_TOKENS agenttel.agentic.cost.cached_write_tokens
AgenticAttributes.QUALITY_GOAL_ACHIEVED agenttel.agentic.quality.goal_achieved
AgenticAttributes.QUALITY_HUMAN_INTERVENTIONS agenttel.agentic.quality.human_interventions
AgenticAttributes.QUALITY_LOOP_DETECTED agenttel.agentic.quality.loop_detected
AgenticAttributes.QUALITY_LOOP_ITERATIONS agenttel.agentic.quality.loop_iterations
AgenticAttributes.QUALITY_EVAL_SCORE agenttel.agentic.quality.eval_score
AgenticAttributes.GUARDRAIL_TRIGGERED agenttel.agentic.guardrail.triggered
AgenticAttributes.GUARDRAIL_NAME agenttel.agentic.guardrail.name
AgenticAttributes.GUARDRAIL_ACTION agenttel.agentic.guardrail.action
AgenticAttributes.GUARDRAIL_REASON agenttel.agentic.guardrail.reason
AgenticAttributes.MEMORY_OPERATION agenttel.agentic.memory.operation
AgenticAttributes.MEMORY_STORE_TYPE agenttel.agentic.memory.store_type
AgenticAttributes.MEMORY_ITEMS agenttel.agentic.memory.items
AgenticAttributes.HUMAN_CHECKPOINT_TYPE agenttel.agentic.human.checkpoint_type
AgenticAttributes.HUMAN_WAIT_MS agenttel.agentic.human.wait_ms
AgenticAttributes.HUMAN_DECISION agenttel.agentic.human.decision
AgenticAttributes.CODE_LANGUAGE agenttel.agentic.code.language
AgenticAttributes.CODE_STATUS agenttel.agentic.code.status
AgenticAttributes.CODE_EXIT_CODE agenttel.agentic.code.exit_code
AgenticAttributes.CODE_SANDBOXED agenttel.agentic.code.sandboxed
AgenticAttributes.EVAL_SCORER_NAME agenttel.agentic.eval.scorer_name
AgenticAttributes.EVAL_CRITERIA agenttel.agentic.eval.criteria
AgenticAttributes.EVAL_SCORE agenttel.agentic.eval.score
AgenticAttributes.EVAL_FEEDBACK agenttel.agentic.eval.feedback
AgenticAttributes.EVAL_TYPE agenttel.agentic.eval.type
AgenticAttributes.ERROR_SOURCE agenttel.agentic.error.source
AgenticAttributes.ERROR_CATEGORY agenttel.agentic.error.category
AgenticAttributes.ERROR_RETRYABLE agenttel.agentic.error.retryable
AgenticAttributes.CAPABILITY_TOOLS agenttel.agentic.capability.tools
AgenticAttributes.CAPABILITY_TOOL_COUNT agenttel.agentic.capability.tool_count
AgenticAttributes.CAPABILITY_SYSTEM_PROMPT_HASH agenttel.agentic.capability.system_prompt_hash
AgenticAttributes.CONVERSATION_ID agenttel.agentic.conversation.id
AgenticAttributes.CONVERSATION_TURN agenttel.agentic.conversation.turn
AgenticAttributes.CONVERSATION_MESSAGE_COUNT agenttel.agentic.conversation.message_count
AgenticAttributes.CONVERSATION_SPEAKER_ROLE agenttel.agentic.conversation.speaker_role
AgenticAttributes.RETRIEVAL_QUERY agenttel.agentic.retrieval.query
AgenticAttributes.RETRIEVAL_STORE_TYPE agenttel.agentic.retrieval.store_type
AgenticAttributes.RETRIEVAL_DOCUMENT_COUNT agenttel.agentic.retrieval.document_count
AgenticAttributes.RETRIEVAL_TOP_K agenttel.agentic.retrieval.top_k
AgenticAttributes.RETRIEVAL_RELEVANCE_SCORE_AVG agenttel.agentic.retrieval.relevance_score_avg
AgenticAttributes.RETRIEVAL_RELEVANCE_SCORE_MIN agenttel.agentic.retrieval.relevance_score_min
AgenticAttributes.RERANKER_MODEL agenttel.agentic.reranker.model
AgenticAttributes.RERANKER_INPUT_DOCUMENTS agenttel.agentic.reranker.input_documents
AgenticAttributes.RERANKER_OUTPUT_DOCUMENTS agenttel.agentic.reranker.output_documents
AgenticAttributes.RERANKER_TOP_SCORE agenttel.agentic.reranker.top_score