Agent Observability Python Framework Module: LlamaIndex

agento11y-llamaindex provides callback handlers that map LlamaIndex workflow/agent events into agento11y generation recorder lifecycles.

Installation

pip install agento11y agento11y-llamaindex
pip install llama-index

Quickstart

from agento11y import Client
from agento11y_llamaindex import with_agento11y_llamaindex_callbacks

client = Client()
config = with_agento11y_llamaindex_callbacks(None, client=client, provider_resolver="auto")
# query_engine = index.as_query_engine(callback_manager=config["callback_manager"])

Native callback manager wiring

from agento11y import Client
from agento11y_llamaindex import with_agento11y_llamaindex_callbacks

client = Client()
config = with_agento11y_llamaindex_callbacks(None, client=client, provider_resolver="auto")
# query_engine = index.as_query_engine(callback_manager=config["callback_manager"])

Conversation mapping

Conversation ID precedence:

1. conversation_id / session_id / group_id
2. thread_id
3. fallback agento11y:framework:llamaindex:<run_id>

Metadata and lineage

• Required: agento11y.framework.run_type
• Optional lineage: agento11y.framework.run_id, agento11y.framework.thread_id, agento11y.framework.parent_run_id, agento11y.framework.component_name, agento11y.framework.retry_attempt, agento11y.framework.event_id

Provider resolver

Model prefix inference:

• gpt-/o1/o3/o4 -> openai
• claude- -> anthropic
• gemini- -> gemini
• fallback custom

Troubleshooting

• Reuse stable workflow/session IDs to keep conversation grouping stable.
• Keep capture_inputs/capture_outputs enabled while validating mappings.
• Call client.flush() at checkpoints and client.shutdown() on exit.