feat: inject anthropic prompt cache control

[shudonglin]

Summary

Adds gateway-side Anthropic automatic prompt-cache control for native Anthropic requests when the client has not already supplied cache_control.

Changes

• Adds ANTHROPIC_PROMPT_CACHE_TTL support with off, 5m, 1h, and auto policies
• Applies top-level cache_control on typed and raw Anthropic Messages paths, streaming paths, and Anthropic Responses requests
• Preserves any client-authored cache_control marker anywhere in the request body
• Strips gateway-only prompt-cache policy headers before Anthropic passthrough forwarding
• Adds focused unit tests for injection, preservation, opt-out, auto workload handling, and raw body mutation

Type of change

• Bug fix
• Feature
• Refactor
• Documentation
• Chore/CI

Affected areas

• Core (Go)
• Transports (HTTP)
• Providers/Integrations
• Plugins
• UI (React)
• Docs

How to test

go test ./core/providers/anthropic -run 'TestApplyAnthropicPromptCacheControl'

Expected outcome: the Anthropic provider tests pass and cover top-level injection, client override preservation, header opt-out, auto workload TTL selection, and raw-body injection.

New configuration:

ANTHROPIC_PROMPT_CACHE_TTL=off|5m|1h|auto
x-anthropic-prompt-cache-ttl: off|5m|1h|auto
x-anthropic-prompt-cache-workload: eval|evaluation|benchmark|bench|batch|pipeline|long|long-running

Screenshots/Recordings

Not applicable; backend provider behavior only.

Breaking changes

• Yes
• No

Related issues

None

Security considerations

The request headers used to control gateway policy are stripped before Anthropic passthrough forwarding. Client-provided cache_control remains authoritative and is not overwritten.

Checklist

• I read docs/contributing/README.md and followed the guidelines
• I added/updated tests where appropriate
• I updated documentation where needed
• I verified builds succeed (Go and UI)
• I verified the CI pipeline passes locally if applicable

[shudonglin]

E2E cache validation

Ran on 2026-06-28 from branch feat/anthropic-cache-control-injection:

go test -count=1 ./core/providers/anthropic -run 'TestAnthropicProviderChatCompletionE2EPromptCacheControl|TestApplyAnthropicPromptCacheControl'

Result: passed (ok github.com/maximhq/bifrost/core/providers/anthropic 1.001s).

The E2E test uses a mock Anthropic upstream, so it validates gateway request/response behavior without spending real Anthropic tokens. It verifies top-level cache_control: {"type":"ephemeral","ttl":"1h"} is sent for long/eval workloads, gateway-only cache policy headers are stripped before upstream, client-supplied cache control is preserved, and Anthropic cache usage fields flow back through response handling.