Add public force_flush() to OTel and Langfuse observers

src/openarmature/observability/langfuse/adapter.py

@@ -328,6 +328,27 @@ def generation(
         )
         return _SpanHandle(obs)
 
+    def force_flush(self, timeout_ms: int = 30_000) -> bool:
+        """Best-effort flush of the underlying Langfuse client.
+
+        ``timeout_ms`` is accepted for Protocol compatibility but
+        **ignored**: the v4 Langfuse SDK's ``flush()`` method takes
+        no timeout parameter and discards the underlying
+        ``TracerProvider.force_flush()`` return value. The call is
+        nonetheless synchronous — internally ``flush()`` waits on
+        OTel's ``force_flush`` (default 30 s) and then ``.join()`` on
+        the SDK's score and media ingestion queues — so by the time
+        we return the OTel batch processor and ingestion queues have
+        either drained or hit the SDK's internal default deadlines.
+
+        Returns ``True`` once the SDK call completes without raising;
+        a tight-deadline caller should pair this with its own
+        wall-clock guard rather than relying on the return value.
+        """
+        del timeout_ms
+        self._client.flush()
+        return True
+
     def _start_observation(
         self,
         *,

src/openarmature/observability/langfuse/client.py

@@ -217,6 +217,28 @@ def generation(
         prompt: Any = None,
     ) -> LangfuseGenerationHandle: ...
 
+    def force_flush(self, timeout_ms: int = 30_000) -> bool:
+        """Flush any pending outbound buffer in the underlying sink.
+
+        Returns ``True`` when the flush completes within the deadline,
+        ``False`` otherwise. The semantics mirror OTel's
+        ``TracerProvider.force_flush``: cover the export-buffer half
+        of fast-teardown races. The bundled
+        :class:`InMemoryLangfuseClient` has no buffer and returns
+        ``True`` immediately; SDK adapters delegate to the underlying
+        client's flush.
+
+        **Deadline honor is best-effort.** Adapters wrapping SDKs
+        that don't expose a timeout-propagation surface (the v4
+        Langfuse SDK is one such case — its ``flush()`` blocks on the
+        SDK's own internal defaults) may ignore ``timeout_ms`` and
+        return ``True`` once the underlying call returns. Callers
+        with a hard deadline should layer their own wall-clock guard
+        around this method rather than relying solely on the return
+        value.
+        """
+        ...
+
 
 # Concrete in-memory implementation ---------------------------------------
 # Used by tests and the conformance harness. Stores everything the
@@ -407,6 +429,13 @@ def generation(
         trace.observations.append(observation)
         return _InMemoryGenerationHandle(observation=observation)
 
+    def force_flush(self, timeout_ms: int = 30_000) -> bool:
+        # In-memory recorder has no outbound buffer; every observation
+        # is captured synchronously on its create call. The ``timeout_ms``
+        # parameter is accepted for Protocol compatibility but unused.
+        del timeout_ms
+        return True
+
     def _get_trace(self, trace_id: str) -> LangfuseTrace:
         trace = self.traces.get(trace_id)
         if trace is None:

src/openarmature/observability/langfuse/observer.py

@@ -741,6 +741,24 @@ def close_invocation(self, invocation_id: str) -> None:
         ):
             self._close_subgraph_observation(inv_state, prefix)
 
+    def force_flush(self, timeout_ms: int = 30_000) -> bool:
+        """Flush pending observations through the underlying client.
+
+        Returns ``True`` when the client's flush completes within the
+        deadline, ``False`` otherwise. Mirrors the OTel observer's
+        ``force_flush`` surface — distinct from
+        :meth:`~openarmature.graph.compiled.CompiledGraph.drain` (which
+        covers the engine's observer-event queue): this method covers
+        the outbound buffer of the Langfuse client (the SDK's OTel
+        BatchSpanProcessor when wrapped via :class:`LangfuseSDKAdapter`).
+
+        Useful in fast-teardown harnesses (CLI one-shots, serverless
+        functions, ASGI lifespan shutdown) where the SDK's
+        BatchSpanProcessor export thread would otherwise be cut off
+        before its buffer drains.
+        """
+        return self.client.force_flush(timeout_ms=timeout_ms)
+
     def shutdown(self) -> None:
         """Drain every in-flight invocation. Use for long-lived
         observers shared across requests; CLI / one-shot processes

src/openarmature/observability/otel/observer.py

@@ -1445,6 +1445,32 @@ def _close_invocation_span(self, invocation_id: str) -> None:
         self._run_enrichers(open_span.span, None)
         open_span.span.end()
 
+    def force_flush(self, timeout_ms: int = 30_000) -> bool:
+        """Flush any pending spans through every registered span processor.
+
+        Returns ``True`` when all processors finish flushing within the
+        deadline, ``False`` otherwise. Wraps the underlying OTel
+        :class:`TracerProvider`'s ``force_flush`` so callers don't have
+        to reach into the private ``_provider`` attribute.
+
+        **When to call.** Distinct from :meth:`drain` on
+        :class:`~openarmature.graph.compiled.CompiledGraph` (which covers
+        the engine's observer-event queue): this method covers the
+        outbound span-export buffer of each registered
+        :class:`SpanProcessor`. Under fast or unusual teardown
+        orderings (FastAPI ``TestClient`` teardown, CLI one-shots,
+        serverless functions) the :class:`BatchSpanProcessor`'s export
+        thread can be cut off before its buffer drains; calling
+        ``force_flush()`` from a ``finally`` block right before process
+        exit is the canonical hardening.
+
+        The default 30 s ``timeout_ms`` matches the OTel SDK's own
+        default. Pass a smaller value when running under a hard
+        deadline (a serverless function's max execution time, an
+        ASGI lifespan timeout).
+        """
+        return self._provider.force_flush(timeout_millis=timeout_ms)
+
     def shutdown(self) -> None:
         """Close any still-open spans across all in-flight invocations
         and shut down the underlying provider. Each per-invocation
@@ -1459,8 +1485,7 @@ def shutdown(self) -> None:
         thread finishes), the flush may not complete in time and spans
         can appear dropped. Workarounds:
 
-        - Call ``observer._provider.force_flush(timeout_millis=…)``
-          explicitly before this method.
+        - Call :meth:`force_flush` explicitly before this method.
         - Use :class:`SimpleSpanProcessor` instead of
           :class:`BatchSpanProcessor` in tests; it exports synchronously
           and is unaffected by teardown timing.

tests/unit/test_observability_langfuse.py

@@ -108,6 +108,25 @@ def test_in_memory_recorder_observation_id_is_unique_per_recorder() -> None:
     assert a.id != b.id
 
 
+def test_observer_force_flush_delegates_to_client() -> None:
+    # LangfuseObserver.force_flush() calls into the client; the
+    # InMemoryLangfuseClient's force_flush is a no-op that returns
+    # True, so this just verifies the delegation wires correctly.
+    client = InMemoryLangfuseClient()
+    observer = LangfuseObserver(client=client)
+    assert observer.force_flush() is True
+    assert observer.force_flush(timeout_ms=1000) is True
+
+
+def test_in_memory_recorder_force_flush_is_no_op() -> None:
+    # In-memory recorder has no outbound buffer; force_flush returns
+    # True immediately. The timeout_ms parameter is accepted for
+    # Protocol compatibility but unused.
+    client = InMemoryLangfuseClient()
+    assert client.force_flush() is True
+    assert client.force_flush(timeout_ms=5000) is True
+
+
 def test_in_memory_recorder_children_of_walks_parent_links() -> None:
     client = InMemoryLangfuseClient()
     client.trace(id="t1")

tests/unit/test_observability_langfuse_adapter.py

@@ -114,6 +114,26 @@ def test_adapter_update_trace_merges_into_cache() -> None:
     assert cached["metadata"] == {"key1": "v1", "key2": "v2"}
 
 
+def test_adapter_force_flush_delegates_to_client() -> None:
+    # force_flush() must invoke the wrapped SDK's flush() so callers
+    # in fast-teardown harnesses get the SDK's internal drain
+    # (OTel TracerProvider.force_flush + score/media queue joins).
+    # Mock(wraps=...) lets us assert delegation without simulating
+    # the SDK's full surface.
+    from unittest.mock import Mock
+
+    real_client = _dummy_client()
+    wrapped = Mock(wraps=real_client)
+    adapter = LangfuseSDKAdapter(wrapped)
+
+    assert adapter.force_flush() is True
+    wrapped.flush.assert_called_once_with()
+
+    # timeout_ms is accepted but unused per the documented contract.
+    assert adapter.force_flush(timeout_ms=1_000) is True
+    assert wrapped.flush.call_count == 2
+
+
 # ---------------------------------------------------------------------------
 # Integration test against real Langfuse Cloud (opt-in)
 # ---------------------------------------------------------------------------
@@ -173,11 +193,12 @@ async def test_adapter_against_real_langfuse_cloud() -> None:
     await graph.invoke(_S())
     await graph.drain()
     observer.shutdown()
-    # ``client.shutdown()`` is the synchronous drain — flush() returns
-    # immediately while the OTel BatchSpanProcessor exports in
-    # background, and the test process exits before that finishes.
-    # shutdown() blocks until all spans are exported (or the
-    # exporter's shutdown timeout elapses).
+    # Use ``client.shutdown()`` rather than ``client.flush()`` here:
+    # both block on the SDK's internal drain (OTel's force_flush plus
+    # the score/media queue joins), but shutdown() also tears down
+    # the resource manager so the test process exits cleanly. flush()
+    # is the appropriate call from a long-lived process that wants
+    # to drain without releasing SDK resources.
     client.shutdown()
     # Manual check: open the trace in the dashboard and confirm
     # "step_a" + "step_b" appear as Span observations under one Trace.

tests/unit/test_observability_otel.py

@@ -1382,3 +1382,16 @@ async def ask_llm(_s: _S) -> dict[str, str]:
     assert attrs.get("openarmature.prompt.label") == "production"
     assert attrs.get("openarmature.prompt.template_hash") == "sha256:tpl"
     assert attrs.get("openarmature.prompt.rendered_hash") == "sha256:rendered"
+
+
+def test_force_flush_delegates_to_provider() -> None:
+    # Public force_flush wraps TracerProvider.force_flush so downstream
+    # users don't reach into observer._provider to drain the
+    # BatchSpanProcessor buffer in fast-teardown harnesses.
+    exporter = InMemorySpanExporter()
+    observer = OTelObserver(span_processor=SimpleSpanProcessor(exporter))
+    try:
+        assert observer.force_flush() is True
+        assert observer.force_flush(timeout_ms=1000) is True
+    finally:
+        observer.shutdown()