loongsuite-python-agent 0.6.0

loongsuite-python-agent 0.6.0

Installation

pip install loongsuite-distro==0.6.0
loongsuite-bootstrap -a install --version 0.6.0

Package Versions

• loongsuite-* packages: 0.6.0
• opentelemetry-* packages: 0.60b1

PyPI packages

The following distributions are built and uploaded to PyPI for this release:

• loongsuite-otel-util-genai
• loongsuite-distro
• loongsuite-site-bootstrap
• loongsuite-instrumentation-agentscope
• loongsuite-instrumentation-algotune
• loongsuite-instrumentation-bfclv4
• loongsuite-instrumentation-claude-agent-sdk
• loongsuite-instrumentation-claw-eval
• loongsuite-instrumentation-crewai
• loongsuite-instrumentation-dashscope
• loongsuite-instrumentation-google-adk
• loongsuite-instrumentation-hermes-agent
• loongsuite-instrumentation-langchain
• loongsuite-instrumentation-langgraph
• loongsuite-instrumentation-litellm
• loongsuite-instrumentation-mem0
• loongsuite-instrumentation-minisweagent
• loongsuite-instrumentation-openhands
• loongsuite-instrumentation-qwen-agent
• loongsuite-instrumentation-qwenpaw
• loongsuite-instrumentation-slop-code
• loongsuite-instrumentation-terminus2
• loongsuite-instrumentation-vita
• loongsuite-instrumentation-webarena
• loongsuite-instrumentation-widesearch
• loongsuite-instrumentation-wildtool

---

loongsuite

Changed

• Publish LoongSuite GenAI utilities as loongsuite-otel-util-genai for new
  releases. New LoongSuite distro and instrumentation packages depend on
  loongsuite-otel-util-genai; the previous loongsuite-util-genai
  distribution remains available for existing installations but is no longer
  the package receiving new LoongSuite GenAI utility updates. The Python import
  namespace remains opentelemetry.util.genai.

---

loongsuite-instrumentation-agno

Removed

• Drop Agno 1.x support and require Agno 2.x public Agent.run/Agent.arun
  APIs. Users that still depend on Agno 1.x should pin
  loongsuite-instrumentation-agno < 0.6.

Changed

• Align message content capture with opentelemetry-util-genai controls such as
  OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY.
• Migrate Agno instrumentation to Agno 2.x public Agent.run/Agent.arun
  APIs and opentelemetry-util-genai ExtendedTelemetryHandler.
• Emit standardized AGENT, LLM, and TOOL GenAI spans for agent runs, model
  calls, streaming calls, async streaming calls, and function executions.

Added

• Add a DashScope smoke example that exercises non-streaming, streaming, and
  concurrent Agno calls.
• Add local test requirements for the Agno LoongSuite tox environment.

---

loongsuite-instrumentation-bfclv4

Added

• Initial release of loongsuite-instrumentation-bfclv4.
• ENTRY span around bfcl_eval._llm_response_generation.generate_results.
• AGENT span around bfcl_eval.model_handler.base_handler.BaseHandler.inference
  with cross-thread OTel context propagation via a narrow patch of
  bfcl_eval._llm_response_generation.ThreadPoolExecutor.
• STEP spans created by reflectively wrapping each handler's
  _query_FC / _query_prompting (discovered via
  bfcl_eval.constants.model_config.MODEL_CONFIG_MAPPING).
• Per-call TOOL spans emitted by wrapping
  bfcl_eval.eval_checker.multi_turn_eval.multi_turn_utils.execute_multi_turn_func_call.
• Provider override mapping for OSS handlers (vLLM / SGLang).
• Multi-turn bfcl.turn_idx and ReAct gen_ai.react.round tracking via
  contextvars.

---

loongsuite-instrumentation-crewai

Breaking

• Align CrewAI GenAI span names with opentelemetry-util-genai
  extended semantic conventions. gen_ai.operation.name now reports
  enter, invoke_agent, or execute_tool; the CrewAI framework operation
  is reported in gen_ai.crewai.operation instead.
• Replace the legacy gen_ai.system=crewai attribute with
  gen_ai.provider.name=crewai.
• Use the current content-capture environment values:
  OTEL_SEMCONV_STABILITY_OPT_IN=gen_ai_latest_experimental and
  OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY.

Changed

• Migrate CrewAI entry, task, agent, and tool spans to
  opentelemetry-util-genai ExtendedTelemetryHandler.
• Keep instrumentation post-processing failures from changing successful
  CrewAI calls into user-visible errors, avoid duplicate nested agent spans,
  and gate content-like CrewAI task and agent attributes behind util-genai
  content capture controls.

Added

• Add a real CrewAI smoke example covering sync, streaming, and concurrent
  calls for local otel-gui and Robin/ARMS verification.

---

loongsuite-instrumentation-dashscope

Fixed

• Fix extraction of gen_ai.output.messages for message-format text responses
  that omit optional tool_calls.

---

loongsuite-instrumentation-google-adk

Changed

• Route Google ADK AGENT, LLM, and TOOL spans through
  opentelemetry-util-genai, emitting current GenAI attributes such as
  gen_ai.input.messages, gen_ai.output.messages,
  gen_ai.tool.call.arguments, gen_ai.tool.call.result,
  gen_ai.span.kind, and gen_ai.provider.name=google_adk. (#199)

Fixed

• Keep Google ADK streaming model spans open until the final response and
  protect same-session concurrent invocations from cross-finishing spans.
• Ensure Google ADK spans include LoongSuite gen_ai.span.kind values such as
  AGENT.

---

loongsuite-instrumentation-litellm

Changed

• Improved LiteLLM GenAI util invocation mapping for positional arguments,
  streaming time-to-first-token, multi-choice outputs, tool-call deltas, and
  a real smoke example (#191).