fix(post): send LlmVerifier requests to a chat completions endpoint

[feiiiiii5]

Problem

LlmVerifier / llm_semantic_verification never returns a verified answer. The function builds a
chat request — it sends messages= and reads resp.choices[0].message.content — but then sends
it to the completions endpoint:

client = client if hasattr(client, 'completions') else client.chat
resp = client.completions.create(model=..., messages=[...], ...)

completions.create takes prompt, not messages, so the call raises; the bare except Exception
turns that into one printed line and return None. Because post_process_messages_func=None is what
tells the adapter to miss the cache, the verifier silently degrades into "always call the real LLM" —
the feature added in #669 does nothing, for any client shape.

The repository's own unit test already encodes the intended contract (a client.chat.completions
object) and is red on main:

$ python3 -m pytest tests/unit_tests/processor/test_post.py -q
FAILED tests/unit_tests/processor/test_post.py::test_llm_verifier - AssertionE...
1 failed, 4 passed

The hasattr(client, 'completions') probe also picks the wrong branch for a Mock (attributes are
auto-created) and for openai>=1.0 (which has both chat and completions), and it never matches
the pre-1.0 module-level ChatCompletion that examples/processor/llm_verifier_example.py uses.

Fix

Route the chat-shaped request to a chat endpoint, keeping both client generations the docs use:

• client.chat.completions.create when the client exposes .chat (openai>=1.0, langchain-style clients);
• client.ChatCompletion.create for the legacy module client;
• client.completions.create only as the last resort, as before.

No behaviour changes for callers, and the except is left in place so a real API error still degrades
to a cache miss rather than raising inside the post-processor.

Verification

Same test file, two trees, base pinned at c59fb3a6152a4458b2a070ca183b61c4b614095f:

### base tree source is pristine upstream
$ git -C base-wt diff -- gptcache/processor/post.py     # empty

=== BASE (upstream source + this PR's tests) ===
3 failed, 5 passed in 0.75s
  test_llm_verifier                                     # the repo's existing test
  test_llm_verifier_uses_the_chat_endpoint              # added here
  test_llm_verifier_supports_legacy_module_client       # added here
AttributeError: '_LegacyClient' object has no attribute 'chat'   # post.py:144

=== HEAD (this fix + the same tests) ===
8 passed in 0.54s

cwd /private/tmp/r16-parent/{base-wt,gptcache}, PYTEST_DISABLE_PLUGIN_AUTOLOAD=1,
python -m pytest tests/unit_tests/processor/test_post.py -q -o addopts="", base rc=1 / head rc=0,
and gptcache.processor.post.__file__ was printed in each run to prove the base run imported the
upstream copy, not the patched one.

New tests use hand-written fakes instead of a bare Mock so the endpoint actually under test is
observable: _ChatClient sets completions = None, so calling it raises instead of silently
returning a Mock, and _LegacyClient has no .chat at all.

Two notes so nobody chases a phantom: tests/pytest.ini adds --html=... --self-contained-html, so
a bare pytest errors unless pytest-html is installed or -o addopts="" is used; and
tests/unit_tests/processor/test_context.py fails to collect in my local venv with
openai.lib._old_api.APIRemoved because that venv has openai>=1.0 — that is an environment
artifact, not something this patch touches.

Checklist

• bug fix, and the user-visible effect ("semantic verification silently never fires") is stated above
• tests added to the existing unit test module, no existing assertion weakened
• git diff limited to gptcache/processor/post.py and tests/unit_tests/processor/test_post.py

[sre-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: feiiiiii5
To complete the pull request process, please assign xiaofan-luan after the PR has been reviewed.
You can assign the PR to them by writing /assign @xiaofan-luan in a comment when ready.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment