Releases: elastic/elastic-otel-python-instrumentations
Releases · elastic/elastic-otel-python-instrumentations
openai-v1.2.0
openai-v1.1.1
openai-v1.1.0
What's Changed
- Fix missing or double spans when completion stream is used with context manager (#80)
- Follow semantic conventions 1.31.0: use GEN_AI_OUTPUT_TYPE, GEN_AI_REQUEST_SEED and GEN_AI_REQUEST_CHOICE_COUNT attributes (#77, #76, #75)
New Contributors
Full Changelog: openai-v1.0.0...openai-v1.1.0
openai-v1.0.0
What's Changed
- openai: use proper genai suggested buckets for metrics by @xrmx in #72
- openai: fix instrumentation of with_raw_response by @xrmx in #73
Full Changelog: openai-v0.7.0...openai-v1.0.0
openai-v0.7.0
What's Changed
- Map non-string, non-dict schema as json_schema (#68, Adrian Cole)
- Don't set attributes if they are NotGiven (#67)
- Apply instrumentation to beta.chat.Completions.parse (#65, Anuraag (Rag) Agrawal)
- Fix README venv installation command and dotenv (#63, Anuraag (Rag) Agrawal)
New Contributors
Full Changelog: openai-v0.6.1...openai-v0.7.0
openai-v0.6.1
What's Changed
- Handle message with developer role (#58)
- Decouple metrics from span recording, fixes running with
OTEL_SDK_DISABLED=true(#56) - Add examples and simplify test matrix (#46, Adrian Cole)
Full Changelog: openai-v0.6.0...openai-v0.6.1
openai-v0.6.0
What's Changed
- Trace GEN_AI_OPENAI_REQUEST_RESPONSE_FORMAT, GEN_AI_OPENAI_REQUEST_SEED, GEN_AI_OPENAI_REQUEST_SERVICE_TIER and GEN_AI_OPENAI_RESPONSE_SERVICE_TIER (#41)
- make OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT work like upstream (#42)
Full Changelog: openai-v0.5.0...openai-v0.6.0
openai-v0.5.0
What's Changed
- Sync embeddings calls tracing with semantic conventions 1.29.0 (#36)
- Drop span events support for events, ELASTIC_OTEL_GENAI_EVENTS environment variable is gone (#38)
- Relax opentelemetry-api dependency so it supports 1.29.0+ (#40)
Full Changelog: openai-v0.4.0...openai-v0.5.0
openai-v0.4.0
What's Changed
- Add support for tracing embeddings calls (#20)
- Rewrite tests in pytest style (#21)
- Don't crash on calls without a model (#22)
- Implement log events support following 1.28.0 GenAI semantic conventions (#23)
- Add tests for asserting exported schema version (#24)
- Don't update non-recording spans (#25)
- Add integration tests and test with both latest and baseline openai client (#28)
- Normalize capture content env variable to upstream (#29, Adrian Cole)
- Format recorded responses as upstream (#30, Adrian Cole)
- Use 1.28.0 semantic conventions log events as default (#31, Adrian Cole)
- Bump required api and sdk to 1.28.2/0.49b2 (#33)
Full Changelog: openai-v0.3.0...openai-v0.4.0
openai-v0.3.0
What's Changed
- Delay loading of the openai module to close race condition with httpx instrumentation (#16)
Full Changelog: openai-v0.2.0...openai-v0.3.0