feat(models): a caller can keep the cache breakpoint off per-request rows

[rezaho]

Why

A prompt-cache entry is only worth writing if the next request can read it, which requires its hashed prefix to consist of bytes that request still contains.

mark_conversation_tail_for_cache marks the last content block of the last message, unconditionally. When a caller appends per-request content after the durable conversation — a clock line, a budget figure, anything derived from "now" — the marker lands on exactly the row the next request cannot reproduce. Every request then writes a fresh entry and reads none, silently, forever.

Measured on Bedrock / anthropic.claude-opus-5, single-step turns, tools present:

marker position	turn 1	turn 2
on the volatile row (behaviour before this PR)	write=8211	write=8226 read=0
on the last durable row (this PR)	write=8425	write=15 read=8425

Note the tools array is load-bearing in any probe of this: without one, a messages-tier marker writes entries that never become readable, which produces false negatives.

What

CACHE_EXEMPT_KEY — a neutral per-item key riding the caller's own message dict, following the defer_loading precedent (the framework's nearest analogous feature also took no new request parameter). A caller marks its per-request rows; the marker moves back to the last durable row, which the next request contains by construction since durable rows only ever grow by append.

Semantics:

• Counted from the END, stopping at the first unmarked row — the exemption is about position, so a marked row with durable rows after it is not a tail and does not move the marker.
• An all-exempt list writes no marker at all (nothing durable to anchor to; a marker there would cost an entry per request and read none).
• The key never reaches the wire — both payload builders already rebuild rows with only the wire-legal fields.

Compatibility

volatile_tail defaults to 0, so with no exempt row the payload is byte-identical to before. A test pins that explicitly. No existing caller changes behaviour.

Twinned on AnthropicOAuthAdapter, which is not a subclass of AnthropicAdapter and has its own format_request_payload. Bedrock inherits unchanged.

Tests

10 new tests in tests/models/test_prompt_cache_breakpoint.py covering placement, the position semantics, the all-exempt case, wire-cleanliness, byte-identity of the no-key path, both legs, and idempotence. Full file: 39 passed. Suite: 4 failures, all pre-existing on main and none in adapters (3 reproduce at the merge base; the 4th is a live smoke test that skips without a key).

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.