telemetry-dev-openai 0.1.0__tar.gz

telemetry_dev_openai-0.1.0/PKG-INFO

@@ -0,0 +1,114 @@
+Metadata-Version: 2.4
+Name: telemetry-dev-openai
+Version: 0.1.0
+Summary: OpenAI integration for telemetry.dev Python SDK
+Keywords: telemetry,opentelemetry,openai,llm,genai,tracing
+Author: telemetry.dev
+License-Expression: MIT
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Typing :: Typed
+Requires-Dist: telemetry-dev>=0.1.0
+Requires-Dist: openai>=2,<3
+Requires-Python: >=3.10
+Project-URL: Homepage, https://telemetry.dev
+Project-URL: Repository, https://github.com/telemetry-dev/telemetry.dev
+Description-Content-Type: text/markdown
+
+# telemetry-dev-openai
+
+OpenAI SDK instrumentation for telemetry.dev. It wraps the official `openai` Python SDK and emits telemetry.dev generation and embedding spans through `telemetry-dev`.
+
+## Install
+
+```sh
+pip install telemetry-dev-openai
+```
+
+Initialize the core SDK first:
+
+```py
+import telemetry_dev
+
+telemetry_dev.init(
+    api_key="td_live_...",
+    base_url="http://localhost:4318",
+    service_name="my-service",
+)
+```
+
+## Per-client wrapping
+
+```py
+from openai import OpenAI
+from telemetry_dev_openai import wrap_openai
+
+client = wrap_openai(OpenAI())
+
+client.chat.completions.create(
+    model="gpt-4o-mini",
+    messages=[{"role": "user", "content": "Tell me a joke about OpenTelemetry"}],
+)
+```
+
+Use this when you want explicit control over which sync or async clients are instrumented.
+
+## Global instrumentation
+
+```py
+from openai import OpenAI
+from telemetry_dev_openai import instrument_openai, uninstrument_openai
+
+instrument_openai()
+client = OpenAI()
+
+try:
+    client.responses.create(model="gpt-4o-mini", input="Tell me a joke about OpenTelemetry")
+finally:
+    uninstrument_openai()
+```
+
+Use this as the app-wide one-liner at startup when all OpenAI clients should be instrumented.
+
+## Instrumented surfaces
+
+Sync and async variants are covered:
+
+- `client.chat.completions.create(...)`
+- `client.chat.completions.parse(...)`
+- `client.chat.completions.stream(...)`
+- `client.responses.create(...)`
+- `client.responses.parse(...)`
+- `client.responses.stream(...)` when starting a new response
+- `client.embeddings.create(...)`
+
+The integration maps native OpenAI request/response shapes directly into telemetry.dev fields. It does not normalize messages into another schema.
+
+## Streaming
+
+Chat completion streams are traced. Requests are sent unchanged by default, so token usage is only captured when the caller sets `stream_options={"include_usage": True}` themselves. Pass `inject_stream_usage=True` to `wrap_openai` or `instrument_openai` to inject it automatically; the synthetic usage-only chunk is then hidden from the caller. Injection is opt-in because some providers reject `stream_options` — for example Azure OpenAI "on your data" (`data_sources`) returns 400 for it while plain `stream=True` works.
+
+```py
+client = wrap_openai(OpenAI(), inject_stream_usage=True)
+```
+
+Responses API streams are traced through `responses.create(stream=True)`; terminal `response.completed`, `response.failed`, and `response.incomplete` events close the span.
+
+## Embeddings
+
+Embedding calls emit `gen_ai.operation.name = "embeddings"`, request model/input, response model, and token usage. Embedding vectors are intentionally not captured as output.
+
+## Azure OpenAI
+
+`wrap_openai(AzureOpenAI(...))` and `wrap_openai(AsyncAzureOpenAI(...))` record provider `azure.ai.openai`. Global class instrumentation detects Azure from the resource client when available.
+
+## Limitations
+
+- `with_raw_response` snapshots bound methods on first access in the OpenAI Python SDK. Call `wrap_openai()` or `instrument_openai()` before accessing `with_raw_response` if those methods need instrumentation.
+- `responses.stream(response_id=...)` resumes an existing response through `retrieve()`, which is not instrumented in this version.
+- Unconsumed streams end their spans only when the stream is exhausted, errors, or is closed.

telemetry_dev_openai-0.1.0/README.md

@@ -0,0 +1,92 @@
+# telemetry-dev-openai
+
+OpenAI SDK instrumentation for telemetry.dev. It wraps the official `openai` Python SDK and emits telemetry.dev generation and embedding spans through `telemetry-dev`.
+
+## Install
+
+```sh
+pip install telemetry-dev-openai
+```
+
+Initialize the core SDK first:
+
+```py
+import telemetry_dev
+
+telemetry_dev.init(
+    api_key="td_live_...",
+    base_url="http://localhost:4318",
+    service_name="my-service",
+)
+```
+
+## Per-client wrapping
+
+```py
+from openai import OpenAI
+from telemetry_dev_openai import wrap_openai
+
+client = wrap_openai(OpenAI())
+
+client.chat.completions.create(
+    model="gpt-4o-mini",
+    messages=[{"role": "user", "content": "Tell me a joke about OpenTelemetry"}],
+)
+```
+
+Use this when you want explicit control over which sync or async clients are instrumented.
+
+## Global instrumentation
+
+```py
+from openai import OpenAI
+from telemetry_dev_openai import instrument_openai, uninstrument_openai
+
+instrument_openai()
+client = OpenAI()
+
+try:
+    client.responses.create(model="gpt-4o-mini", input="Tell me a joke about OpenTelemetry")
+finally:
+    uninstrument_openai()
+```
+
+Use this as the app-wide one-liner at startup when all OpenAI clients should be instrumented.
+
+## Instrumented surfaces
+
+Sync and async variants are covered:
+
+- `client.chat.completions.create(...)`
+- `client.chat.completions.parse(...)`
+- `client.chat.completions.stream(...)`
+- `client.responses.create(...)`
+- `client.responses.parse(...)`
+- `client.responses.stream(...)` when starting a new response
+- `client.embeddings.create(...)`
+
+The integration maps native OpenAI request/response shapes directly into telemetry.dev fields. It does not normalize messages into another schema.
+
+## Streaming
+
+Chat completion streams are traced. Requests are sent unchanged by default, so token usage is only captured when the caller sets `stream_options={"include_usage": True}` themselves. Pass `inject_stream_usage=True` to `wrap_openai` or `instrument_openai` to inject it automatically; the synthetic usage-only chunk is then hidden from the caller. Injection is opt-in because some providers reject `stream_options` — for example Azure OpenAI "on your data" (`data_sources`) returns 400 for it while plain `stream=True` works.
+
+```py
+client = wrap_openai(OpenAI(), inject_stream_usage=True)
+```
+
+Responses API streams are traced through `responses.create(stream=True)`; terminal `response.completed`, `response.failed`, and `response.incomplete` events close the span.
+
+## Embeddings
+
+Embedding calls emit `gen_ai.operation.name = "embeddings"`, request model/input, response model, and token usage. Embedding vectors are intentionally not captured as output.
+
+## Azure OpenAI
+
+`wrap_openai(AzureOpenAI(...))` and `wrap_openai(AsyncAzureOpenAI(...))` record provider `azure.ai.openai`. Global class instrumentation detects Azure from the resource client when available.
+
+## Limitations
+
+- `with_raw_response` snapshots bound methods on first access in the OpenAI Python SDK. Call `wrap_openai()` or `instrument_openai()` before accessing `with_raw_response` if those methods need instrumentation.
+- `responses.stream(response_id=...)` resumes an existing response through `retrieve()`, which is not instrumented in this version.
+- Unconsumed streams end their spans only when the stream is exhausted, errors, or is closed.

telemetry_dev_openai-0.1.0/pyproject.toml

@@ -0,0 +1,58 @@
+[project]
+name = "telemetry-dev-openai"
+version = "0.1.0"
+description = "OpenAI integration for telemetry.dev Python SDK"
+readme = "README.md"
+license = "MIT"
+requires-python = ">=3.10"
+authors = [{ name = "telemetry.dev" }]
+keywords = ["telemetry", "opentelemetry", "openai", "llm", "genai", "tracing"]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "Intended Audience :: Developers",
+  "Programming Language :: Python :: 3",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
+  "Typing :: Typed",
+]
+dependencies = [
+  "telemetry-dev>=0.1.0",
+  "openai>=2,<3",
+]
+
+[project.urls]
+Homepage = "https://telemetry.dev"
+Repository = "https://github.com/telemetry-dev/telemetry.dev"
+
+[tool.uv.sources]
+telemetry-dev = { path = "../python", editable = true }
+
+[dependency-groups]
+dev = [
+  "pytest>=8.3",
+  "pytest-asyncio>=0.25",
+  "ruff>=0.9",
+  "pyright>=1.1.390",
+]
+
+[build-system]
+requires = ["uv_build>=0.9.0,<0.10.0"]
+build-backend = "uv_build"
+
+[tool.pytest.ini_options]
+asyncio_mode = "auto"
+testpaths = ["tests"]
+
+[tool.ruff]
+line-length = 100
+target-version = "py310"
+
+[tool.ruff.lint]
+select = ["E", "F", "I", "UP", "B", "RUF"]
+
+[tool.pyright]
+include = ["src", "tests"]
+typeCheckingMode = "strict"
+pythonVersion = "3.10"