Capture context_management In Anthropic Messages Tracer

[Atharva Mhaske (atharvamhaske)]

Summary

The Anthropic Messages tracer does not capture context_management. This field controls Claude's server-side context editing. It clears oldtool-use results or thinking blocks as a conversation grows.

Changes

• trace/contrib/anthropic/messages.go: add context_management to the request metadata fields. Add the response-side applied_edits field, nested under the same context_management key.
• trace/contrib/anthropic/traceanthropic_test.go: add a test for both the request config and the response applied_edits.
• examples/internal/anthropic/main.go: add a context management example.

Closes #222

Screenshot

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a7186e2f5c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "Codex (@codex) address that feedback".

[Atharva Mhaske (atharvamhaske)]

note: i haven't tried on braintrust dashboard as i dont have anthropic api key to test, you can test it !

[Atharva Mhaske (atharvamhaske)]

fixed above suggestion by bot, ClaudeStreamAccumulator now has a contextManagement field. Its Add method saves **context_management** from the message_start event into this field. a new ContextManagement() method returns this saved value. In parseStreamingResponse, the code reads applied_edits from this value and adds it to the span metadata, under the same context_management key that holds the request-side config.

this matches the pattern already used for non-streaming responses, so both paths now report applied edits the same way.

cc: Abhijeet Prasad (@AbhiPrasad) re run ci jobs, once you get time :)

[Abhijeet Prasad (AbhiPrasad)]

Let's delete the mocks/fakes tests we have in favor of adding a vcr test. Once we have the vcr test, I can record the cassette with my api key

[Atharva Mhaske (atharvamhaske)]

Let's delete the mocks/fakes tests we have in favor of adding a vcr test. Once we have the vcr test, I can record the cassette with my api key

sure on it

[Atharva Mhaske (atharvamhaske)]

reverted all fake tests in latest commit, ci is green
cc: Abhijeet Prasad (@AbhiPrasad)

[Atharva Mhaske (atharvamhaske)]

Abhijeet Prasad (@AbhiPrasad) PTAL

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-14T16:58:23.693368Z	c67cbe0	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review" or "Codex (@codex) security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[Abhijeet Prasad (AbhiPrasad)]

push up the cassette and made some fixes + cleanup

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c67cbe0459

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "Codex (@codex) address that feedback".