langgraph_agent_toolkit.core.observability.phoenix module
Send isolated agent traces and prompts to Phoenix.
- class langgraph_agent_toolkit.core.observability.phoenix.PhoenixObservability(remote_first=False, *, endpoint=None, project_name=None, api_key=None, timeout=None, client=None, http_client=None, tracer_provider=None, trace_config=None)[source][source]
Bases:
BaseObservabilityPlatformUse Phoenix without global tracing or LangChain patches.
Each instance owns its default HTTP client and tracer provider. The caller owns any injected client or provider.
- Parameters:
remote_first (bool)
endpoint (str | None)
project_name (str | None)
api_key (str | None)
timeout (float | None)
client (Any)
http_client (Client | None)
tracer_provider (Any)
trace_config (TraceConfig | None)
- __init__(remote_first=False, *, endpoint=None, project_name=None, api_key=None, timeout=None, client=None, http_client=None, tracer_provider=None, trace_config=None)[source][source]
- Parameters:
remote_first (bool)
endpoint (str | None)
project_name (str | None)
api_key (str | None)
timeout (float | None)
client (Any)
http_client (Client | None)
tracer_provider (Any)
trace_config (TraceConfig | None)
- Return type:
None
- get_callback_handler(**kwargs)[source][source]
Create one native OpenInference callback for each run.
- Parameters:
kwargs (Any)
- Return type:
Any
- trace_context(run_id, **kwargs)[source]
Start a root trace with the request UUID as its trace ID.
- Parameters:
run_id (str)
kwargs (Any)
- update_trace(trace, **attributes)[source][source]
Record root input and output through OpenInference masking.
- Parameters:
trace (Any)
attributes (Any)
- Return type:
None
- record_feedback(run_id, key, score, **kwargs)[source][source]
Save human feedback as a native trace annotation.
- Parameters:
run_id (str)
key (str)
score (float)
kwargs (Any)
- Return type:
None
- push_prompt(name, prompt_template, metadata=None, force_create_new_version=True)[source][source]
Create a native chat prompt version and optional version tags.
- Parameters:
name (str)
prompt_template (str | List[ChatMessageDict])
metadata (dict[str, Any] | None)
force_create_new_version (bool)
- Return type:
None
- pull_prompt(name, template_format='f-string', **kwargs)[source][source]
Get the latest, tagged, or explicitly identified prompt version.
- Parameters:
name (str)
template_format (Literal['f-string', 'mustache', 'jinja2'])
kwargs (Any)
- Return type:
ChatPromptTemplate | str | dict | None