From 8bc2c189a9b5a3a1defb892c1a10a078da269eb3 Mon Sep 17 00:00:00 2001
From: kamilbenkirane <benkirane.kamil@gmail.com>
Date: Wed, 6 May 2026 11:10:19 +0200
Subject: [PATCH] feat(telemetry): emit gen_ai.response.model from
 Output.metadata["model"]
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

`output_attributes` now emits `gen_ai.response.model` set to
`output.metadata["model"]` — the model id celeste already populates on
every Output. Both `_predict` (via `ModalityClient._build_metadata`) and
`_stream` (via the `stream_metadata` constructor arg) write `model` to
metadata, so the field is universally present; no guards needed.

For providers that echo a different served snapshot in their response
(Anthropic latest-tag → dated, OpenAI alias → snapshot), provider mixins
can opt in to overwrite `metadata["model"]` with the served value. None
do that today — `metadata["model"]` is the request id everywhere — and
the emitted attribute equals `gen_ai.request.model` in practice. That's
semconv-clean for the common case where served = requested.

Test fixture `TelemetryStream` now defaults `stream_metadata={"model":
"test-model"}` to mirror production wiring; one updated test passes the
metadata explicitly. New positive test asserts `gen_ai.response.model`
reads from `metadata["model"]`.

Real-call validated against running chat-backend (Gemini text + image
streaming): every GenAI span now carries `gen_ai.response.model`.
---
 src/celeste/telemetry.py                     |  1 +
 tests/unit_tests/_telemetry_helpers.py       | 15 ++++++++++++++-
 tests/unit_tests/test_telemetry_streaming.py | 17 ++++++++++++++++-
 3 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/src/celeste/telemetry.py b/src/celeste/telemetry.py
index c6ce838..a4fb07e 100644
--- a/src/celeste/telemetry.py
+++ b/src/celeste/telemetry.py
@@ -224,6 +224,7 @@ def output_attributes(output: Output[Any]) -> dict[str, Any]:
         attrs[semconv_key or f"celeste.usage.{field_name}"] = value
     if output.finish_reason is not None and output.finish_reason.reason is not None:
         attrs["gen_ai.response.finish_reasons"] = (output.finish_reason.reason,)
+    attrs["gen_ai.response.model"] = output.metadata["model"]
     return attrs
 
 
diff --git a/tests/unit_tests/_telemetry_helpers.py b/tests/unit_tests/_telemetry_helpers.py
index 56f1c6f..451c88d 100644
--- a/tests/unit_tests/_telemetry_helpers.py
+++ b/tests/unit_tests/_telemetry_helpers.py
@@ -1,7 +1,7 @@
 """Shared fixtures and helpers for telemetry-related unit tests."""
 
 from collections.abc import AsyncIterator
-from typing import Any, ClassVar
+from typing import Any, ClassVar, Unpack
 
 from celeste.io import Chunk, Output, Usage
 from celeste.parameters import Parameters
@@ -30,6 +30,19 @@ class TelemetryStream(Stream[TelemetryOutput, Parameters, Chunk]):
     _output_class: ClassVar[type[Output]] = TelemetryOutput
     _empty_content: ClassVar[str] = ""
 
+    def __init__(
+        self,
+        sse_iterator: AsyncIterator[dict[str, Any]],
+        stream_metadata: dict[str, Any] | None = None,
+        **parameters: Unpack[Parameters],
+    ) -> None:
+        """Default `stream_metadata={"model": "test-model"}` to mirror production wiring."""
+        super().__init__(
+            sse_iterator,
+            stream_metadata=stream_metadata or {"model": "test-model"},
+            **parameters,
+        )
+
     def _aggregate_content(self, chunks: list[Chunk]) -> str:
         """Concatenate chunk content."""
         return "".join(chunk.content for chunk in chunks)
diff --git a/tests/unit_tests/test_telemetry_streaming.py b/tests/unit_tests/test_telemetry_streaming.py
index fa91081..ee95a01 100644
--- a/tests/unit_tests/test_telemetry_streaming.py
+++ b/tests/unit_tests/test_telemetry_streaming.py
@@ -16,6 +16,7 @@
 from tests.unit_tests._telemetry_helpers import (
     TelemetryOutput,
     TelemetryStream,
+    TelemetryUsage,
     async_iter,
 )
 from tests.unit_tests.conftest import start_test_span
@@ -213,10 +214,24 @@ async def test_total_reasoning_cached_tokens_emitted(
     def test_off_spec_usage_emitted_under_celeste_namespace(self) -> None:
         """Modality-specific Usage fields fall through to `celeste.usage.<field>`."""
         usage = _OffSpecUsage(images_generated=4, audio_seconds=1.5)
-        output: Output[Any] = TelemetryOutput(content="", usage=usage)
+        output: Output[Any] = TelemetryOutput(
+            content="", usage=usage, metadata={"model": "test-model"}
+        )
 
         attrs = telemetry.output_attributes(output)
 
         assert attrs["celeste.usage.images_generated"] == 4
         assert attrs["celeste.usage.audio_seconds"] == 1.5
         assert "gen_ai.usage.images_generated" not in attrs
+
+    def test_response_model_emitted_from_metadata(self) -> None:
+        """`gen_ai.response.model` reads from `output.metadata["model"]`."""
+        output: Output[Any] = TelemetryOutput(
+            content="",
+            usage=TelemetryUsage(),
+            metadata={"model": "claude-opus-4-1-20250805"},
+        )
+
+        attrs = telemetry.output_attributes(output)
+
+        assert attrs["gen_ai.response.model"] == "claude-opus-4-1-20250805"