Uptake capi headers wave 2

[Sun Haoran (haoranpb)]

Follow up #756

Instead of handling the headers in BC-Bench, moved the logic upstream into bc-eval, and uptake, bc-eval now have those headers out of box. (see https://microsoft.ghe.com/bic/BC-LanguageModelTools/pull/69)

The implementation of those headers belongs in bc-eval, so the downstream (e.g. BC-Bench) does not need to handle them.

[Copilot]

Pull request overview

Moves CAPI header handling into bc-eval 0.3.13 and removes BC-Bench’s local shims.

Changes:

• Upgrades CAPI evaluation workflows to bc-eval 0.3.13.
• Removes local taxonomy-header injection and related tests/configuration.
• Simplifies the CAPI bridge to call the client directly.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
.github/workflows/bcal-evaluation.yml	Upgrades bc-eval and removes taxonomy environment variables.
.github/workflows/summarize-results.yml	Upgrades the evaluation tool dependency.
evaluator/_capi_taxonomy.py	Removes the local evaluator header shim.
evaluator/scores.py	Removes shim initialization.
src/bcbench/agent/bcal/bc_eval_capi_bridge.py	Removes header injection and retry code.
tests/test_bcal_capi_bridge.py	Removes tests for deleted bridge behavior.
tests/test_capi_taxonomy_evaluator.py	Removes tests for the deleted evaluator shim.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated no new comments.

Suppressed comments (1)

src/bcbench/agent/bcal/bc_eval_capi_bridge.py:137

• This bypasses the bridge's existing retry policy, which was independent of the taxonomy shim. A transient CAPI HttpResponseError such as DependencyFailure/server_error will now abort the entire agent turn on the first failure instead of retrying up to three times. Please retain _create_with_retry (and its tests) and remove only the taxonomy-specific patching.

    response = client.chat.completions.create(**kwargs)