feat(llmobs): track prompt caching for openai chat completions

[lievan]

Tracks number of tokens read from the prompt cache for openai chat completions

openai does prompt caching by default and returns a cached_tokens field in prompt_tokens_details
https://platform.openai.com/docs/api-reference/chat/create

We rely on two keys in metrics for prompt caching:

• cache_read_input_tokens
• cache_write_input_tokens

We have both of these fields since bedrock/anthropic return info on cache read/writes

cached_tokens maps to cache_read_input_tokens

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

releasenotes/notes/oai-p-cache-78c511f97709a357.yaml                    @DataDog/apm-python
tests/contrib/openai/cassettes/v1/chat_completion_prompt_caching_cache_read.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/chat_completion_prompt_caching_cache_write.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/chat_completion_stream_prompt_caching_cache_read.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/chat_completion_stream_prompt_caching_cache_write.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/responses_prompt_caching_cache_read.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/responses_prompt_caching_cache_write.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/responses_stream_prompt_caching_cache_read.yaml  @DataDog/ml-observability
tests/contrib/openai/cassettes/v1/responses_stream_prompt_caching_cache_write.yaml  @DataDog/ml-observability
ddtrace/llmobs/_integrations/openai.py                                  @DataDog/ml-observability
tests/contrib/openai/test_openai_llmobs.py                              @DataDog/ml-observability

[github-actions]

Bootstrap import analysis

Comparison of import times between this PR and base.

Summary

The average import time from this PR is: 275 ± 2 ms.

The average import time from base is: 277 ± 2 ms.

The import time difference between this PR and base is: -1.95 ± 0.08 ms.

Import time breakdown

The following import paths have shrunk:

ddtrace.auto 1.974 ms (0.72%)

ddtrace.bootstrap.sitecustomize 1.299 ms (0.47%)

ddtrace.bootstrap.preload 1.299 ms (0.47%)

ddtrace.internal.remoteconfig.client 0.656 ms (0.24%)

ddtrace 0.675 ms (0.25%)

ddtrace.internal._unpatched 0.032 ms (0.01%)

json 0.032 ms (0.01%)

json.decoder 0.032 ms (0.01%)

re 0.032 ms (0.01%)

enum 0.032 ms (0.01%)

types 0.032 ms (0.01%)

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-07-09 15:29:00

Comparing candidate commit 749f06a in PR branch evan.li/openai-prompt-caching with baseline commit 573a530 in branch main.

Found 0 performance improvements and 1 performance regressions! Performance is the same for 523 metrics, 2 unstable metrics.

scenario:iastaspectsospath-ospathnormcase_aspect

• 🟥 execution_time [+398.337ns; +474.077ns] or [+11.400%; +13.567%]

[Yun-Kim]

Nicely done! Small nits but otherwise lgtm