Hacktoberfest 2026: as issues que os mantenedores marcaram para outubro, abertas e boas para iniciantes. Ver issues do Hacktoberfest

feat(sdk): support multiple OTLP span exporters (dual-write alongside Langfuse)

Aberta
  • Algo errado?
#1,716 2 comentários 1 reação 1 responsável Ver no GitHub

Mantenedores costumam responder em até 1 dia

@hassiebp já está trabalhando nisso.

Desde 18/6/2026.

Avaliação

Esta issue ainda não foi avaliada.

Descrição

feat-otel integration-otel integration-python-decorator sdk-python
Summary

We would like Langfuse SDKs to support exporting spans to more than one destination—specifically, keep the default Langfuse OTLP export and optionally forward the same spans to an additional OTLP endpoint (or custom SpanExporter).

This would follow the standard OpenTelemetry pattern of attaching multiple SpanProcessors to a single TracerProvider, without requiring users to replace Langfuse's built-in exporter.

Problem

Today, when instrumenting with the Langfuse Python SDK (@observe(), LangChain callbacks, etc.), spans are exported through Langfuse's internal OTLP pipeline to the configured Langfuse host.

Some deployments need parallel export to another OTLP-compatible backend—for example:

  • An internal observability or logging pipeline (same VPC / on-prem)
  • A secondary sink during migration or validation
  • Environment-specific routing (e.g. dev/staging copies to a local collector)

PR https://github.com/langfuse/langfuse-python/pull/1618 added optional span_exporter, which is helpful for customization, but it replaces the default Langfuse exporter rather than adding a second one. Setting generic OTEL_EXPORTER_OTLP_* env vars also does not appear to dual-write when using Langfuse's own TracerProvider initialization.

Proposed behavior

Support one or more additional exporters while preserving the default Langfuse export path.

Option A — constructor / client API

from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter

Langfuse(
    additional_span_exporters=[
        OTLPSpanExporter(endpoint="https://internal-collector.example.com/v1/traces"),
    ],
)

Option B — environment variables (no code changes)

# Existing Langfuse config unchanged
LANGFUSE_PUBLIC_KEY=...
LANGFUSE_SECRET_KEY=...
LANGFUSE_HOST=...

# Additional OTLP sink
LANGFUSE_ADDITIONAL_OTLP_ENDPOINT=https://internal-collector.example.com
LANGFUSE_ADDITIONAL_OTLP_HEADERS=Authorization=Bearer ...

Requirements we care about:

  • Default behavior unchanged when additional exporters are not configured
  • Additional exporters receive the same spans as Langfuse (no silent filtering unless explicitly configured later)
  • get_client() / singleton initialization should preserve additional exporters
  • Clear documentation on auth headers, timeouts, and failure isolation (one exporter failing should not break the other)
Alternatives considered
  1. Replace exporter via span_exporter (#1618) — loses default Langfuse export unless manually re-wired
  2. User-owned OTEL Collector sidecar — works but adds operational overhead; SDK-level dual export is simpler for many teams
  3. Application-level double instrumentation — fragile across frameworks and easy to break on upgrades
Questions for maintainers
  1. Is dual / multi-exporter export something the Langfuse SDK team would consider?
  2. Would you prefer an explicit additional_span_exporters API, env-based configuration, or both?
  3. Should this live in langfuse-python first, with parity in langfuse-js?
  4. Any concerns about performance, backpressure, or support burden we should address in a PR?

Happy to contribute a PR if there is alignment on the API shape. Thanks!

Linguagem predominante
Python
Estrelas
468
Forks
349
Merge médio
16h 1min
PRs com merge (30d)
20

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de langfuse/langfuse-python

Todas as issues de langfuse/langfuse-python

Issues semelhantes

Mais issues de Python

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.