Issues

• [bug] litellm: streaming span ends with status UNSET when the stream fails mid-way

  bugtriage
  Arize-ai/openinference#4013 · Filter by author armanunix opened on Oct 10, 2026

• [bug] litellm: a stream the app stops reading early records no output (llm.output_messages / output.value missing)

  bugtriage
  Arize-ai/openinference#4011 · Filter by author armanunix opened on Oct 10, 2026

• [bug] bedrock: invoke_model drops the system prompt and tools; streaming writes system blocks under a malformed key

  bugtriage
  Arize-ai/openinference#4010 · Filter by author armanunix opened on Oct 10, 2026

• [bug] openinference-vercel: Eve 0.76 agent.step / agent.approval spans classified as LLM again (regression of #3876)

  Arize-ai/openinference#4008 · Filter by author jimbobbennett opened on Oct 9, 2026

• [refactor] [ai-sdk] Consider moving all span conversion to enrichSpan primitive

  enhancementlanguage: js
  Arize-ai/openinference#4006 · Filter by author cephalization opened on Oct 9, 2026
  

• [bug] [agno] Callable tools factory is called without awaiting it: async tool factories crash the run ("'coroutine' object is not iterable")

  bugtriage
  Arize-ai/openinference#3997 · Filter by author SemGa92 opened on Oct 9, 2026

• [bug] oogle-genai: contents passed as a tuple are not recorded as llm.input_messages Describe the bug

  bugtriage
  Arize-ai/openinference#3990 · Filter by author armanunix opened on Oct 8, 2026
  1

• auto instrumentor for elevenlabs python sdk

  enhancementinstrumentationlanguage: pythonnew instrumentationsize:L
  Arize-ai/openinference#3956 · Filter by author caroger opened on Oct 5, 2026
  

• [claude-agent-sdk] Token counts and cost are cumulative per session after resume (claude-agent-sdk >= 0.2.157)

  Arize-ai/openinference#3945 · Filter by author hassiebp opened on Oct 5, 2026

• [bug] bedrock: Converse llm.invocation_parameters omits additionalModelRequestFields (e.g. top_k) and includes the response stop_reason

  buggood first issueinstrumentation: bedrockinstrumentationlanguage: pythonsize:S
  Arize-ai/openinference#3944 · Filter by author armanunix opened on Oct 5, 2026
  2

  

• [bug] bedrock: converse_stream and invoke_model_with_response_stream spans are never ended when the stream is left early (break, close(), abandoned)

  buggood first issueinstrumentation: bedrockinstrumentationlanguage: pythonsize:M
  Arize-ai/openinference#3943 · Filter by author armanunix opened on Oct 5, 2026
  2

  

• [bug] bedrock: system prompt recorded with OPENINFERENCE_HIDE_INPUTS=true (in llm.invocation_parameters for invoke_model, and under an unprefixed message.contents key for streamed system blocks)

  Arize-ai/openinference#3933 · Filter by author armanunix opened on Oct 5, 2026
  1

  

• [bug] bedrock: invoke_model body passed as a file-like object breaks invoke_model_with_response_stream (TypeError after the request is sent, span never ends) and is not traced by invoke_model

  buggood first issueinstrumentation: bedrockinstrumentationlanguage: pythonsize:S
  Arize-ai/openinference#3932 · Filter by author armanunix opened on Oct 5, 2026
  2

  

• [bug] anthropic: stream.close() and abandoned streams never end the LLM span (sync and async)

  bug
  Arize-ai/openinference#3930 · Filter by author armanunix opened on Oct 4, 2026
  2

  

• [bug] JS openai-agents: tool spans are never made active, so spans created inside a tool start a new trace

  buggood first issueinstrumentation: openai agents sdkinstrumentationlanguage: jssize:M
  Bug

  Arize-ai/openinference#3925 · Filter by author selimaytac opened on Oct 2, 2026
  1

  

• [bug] claude-agent-sdk (JS): AGENT span prompt/total token counts exclude prompt-cache tokens

  bug
  Arize-ai/openinference#3923 · Filter by author jimbobbennett opened on Oct 2, 2026
  

• [feature request] Add instrumentation for pi-durable

  enhancementlanguage: js
  Arize-ai/openinference#3921 · Filter by author cephalization opened on Oct 2, 2026
  

• [bug] anthropic: early exit from a stream records the SDK's placeholder tool arguments ({}) as the final tool call

  bugtriage
  Arize-ai/openinference#3904 · Filter by author armanunix opened on Oct 1, 2026
  1

• auto instrumentor for livekit python sdk

  instrumentation
  Feature

  Arize-ai/openinference#3903 · Filter by author caroger opened on Oct 1, 2026
  

• [bug] pipecat: prompt token count not folded, cache_write and audio details never set

  Arize-ai/openinference#3893 · Filter by author chrikrah opened on Sep 30, 2026
  

• Together PR #3843 drops error spans when content generators raise

  Arize-ai/openinference#3884 · Filter by author caroger opened on Sep 29, 2026
  

• [feature request] Reduce CPU cost of input/output serialization in _update_span for LangChain autoinstrumentor

  enhancementtriage
  Arize-ai/openinference#3882 · Filter by author sofia099 opened on Sep 29, 2026
  

• [bug] anthropic (python and js): document blocks are silently dropped from recorded input messages

  c/semcovenhancementinstrumentation: anthropicinstrumentationlanguage: jslanguage: pythonsize:Ltriage
  Arize-ai/openinference#3879 · Filter by author Harsh23Kashyap opened on Sep 29, 2026
  1

  

• get_llm_input_message_attributes stores list-valued content in a string span attribute

  bugc/coregood first issuelanguage: pythonsize:S
  Arize-ai/openinference#3859 · Filter by author feiiiiii5 opened on Sep 28, 2026
  1

  

• [bug] openai-agents: cached and reasoning token counts dropped for Chat Completions generation spans

  Arize-ai/openinference#3848 · Filter by author devansh-sodhi opened on Sep 27, 2026
  1