Merge branch 'pr-772' into devin/1785108025-claude-prompt-caching

tests/test_inputs.py

@@ -55,6 +55,110 @@ def test_child_initial_input_no_consecutive_same_role(parent_history: list[Any])
     assert all(prev != nxt for prev, nxt in pairwise(roles))
 
 
+def _cache_points(model_name: str) -> Any:
+    extra = make_model_settings(None, model_name=model_name).extra_args or {}
+    return extra.get("cache_control_injection_points")
+
+
+@pytest.mark.parametrize(
+    "model_name",
+    [
+        "bedrock/global.anthropic.claude-opus-4-8",
+        "anthropic/claude-sonnet-4-5",
+        "openrouter/anthropic/claude-3.5-sonnet",
+    ],
+)
+def test_make_model_settings_enables_prompt_cache_for_claude(model_name: str) -> None:
+    points = _cache_points(model_name)
+    assert points == [
+        {"location": "message", "role": "system"},
+        {"location": "tool_config"},
+        {"location": "message", "index": -1},
+    ]
+
+
+@pytest.mark.parametrize("model_name", ["gpt-5", "vertex_ai/gemini-2.5-pro", "openai/o3"])
+def test_make_model_settings_no_prompt_cache_for_non_claude(model_name: str) -> None:
+    # No injection points for non-Claude models: the LiteLLM cache hook never
+    # fires, so this stays a strict no-op (won't emit cache_control to strict
+    # OpenAI-compatible endpoints).
+    assert make_model_settings(None, model_name=model_name).extra_args is None
+
+
+def test_no_prompt_cache_for_unmapped_bedrock_claude_model(monkeypatch: Any) -> None:
+    """A BEDROCK Claude route LiteLLM has NOT mapped (a new release, or any model
+    when LiteLLM can't refresh its model map and falls back to a stale local
+    copy) must run UNCACHED, not crash. Bedrock's Converse API rejects the
+    unknown field outright (ValidationException 'cache_control_injection_points:
+    Extra inputs are not permitted'); LiteLLM only strips the marker for models
+    it recognises as cache-capable, so an unmapped model would 500 the first
+    call and fail the whole run."""
+    import litellm
+
+    unmapped = "bedrock/global.anthropic.claude-brand-new-9"
+    # Simulate a model LiteLLM doesn't know: no cost-map entry, checker says no.
+    monkeypatch.setattr(litellm, "model_cost", {}, raising=False)
+    if getattr(getattr(litellm, "utils", None), "supports_prompt_caching", None):
+        monkeypatch.setattr(litellm.utils, "supports_prompt_caching", lambda *_a, **_k: False)
+
+    # Bedrock Claude by name, but unmapped → no injection points, no crash.
+    assert make_model_settings(None, model_name=unmapped).extra_args is None
+
+
+def test_prompt_cache_kept_for_non_bedrock_claude_even_if_unmapped(monkeypatch: Any) -> None:
+    """Non-Bedrock Claude routes must KEEP caching-by-family even when LiteLLM
+    can't confirm support — those providers tolerate/ignore the marker (or
+    LiteLLM maps them under keys we don't resolve, e.g. OpenRouter), so gating
+    them on confirmed support would DISABLE caching for capable models — a
+    regression. Only Bedrock hard-rejects, so only Bedrock is guarded."""
+    import litellm
+
+    monkeypatch.setattr(litellm, "model_cost", {}, raising=False)
+    if getattr(getattr(litellm, "utils", None), "supports_prompt_caching", None):
+        monkeypatch.setattr(litellm.utils, "supports_prompt_caching", lambda *_a, **_k: False)
+
+    # Even with LiteLLM knowing nothing, an Anthropic-native / OpenRouter Claude
+    # still gets the injection points.
+    for model in ("anthropic/claude-brand-new-9", "openrouter/anthropic/claude-brand-new"):
+        assert _cache_points(model) == [
+            {"location": "message", "role": "system"},
+            {"location": "tool_config"},
+            {"location": "message", "index": -1},
+        ]
+
+
+def test_conversation_tail_breakpoint_moves_with_appended_transcript() -> None:
+    """The tail breakpoint's premise, end-to-end: LiteLLM's own message-injection
+    logic must place the cache_control on the LAST message for both a short and a
+    long transcript — i.e. it tracks the growing (append-only) tail rather than a
+    fixed position — so the immutable prefix-so-far is cached and re-read next
+    turn.
+
+    Driven through the hook's static ``_apply_message_injections`` primitive
+    (stable across LiteLLM versions) rather than the prompt-manager entrypoint
+    (whose signature drifts).
+    """
+    hook_mod = pytest.importorskip("litellm.integrations.anthropic_cache_control_hook")
+    apply = hook_mod.AnthropicCacheControlHook._apply_message_injections
+    points = _cache_points("bedrock/global.anthropic.claude-opus-4-8")
+    msg_points = [p for p in points if p.get("location") == "message"]
+
+    def last_msg_cache_control(n_turns: int) -> Any:
+        messages: list[dict[str, Any]] = [{"role": "system", "content": "stable prompt"}]
+        for i in range(n_turns):
+            messages.append({"role": "assistant", "content": f"turn {i} action"})
+            messages.append({"role": "user", "content": f"turn {i} tool result"})
+        processed = apply(msg_points, messages, 4)
+        last = processed[-1]
+        content = last.get("content")
+        if isinstance(content, list):
+            return content[-1].get("cache_control")
+        return last.get("cache_control")
+
+    assert last_msg_cache_control(2) == {"type": "ephemeral"}
+    assert last_msg_cache_control(20) == {"type": "ephemeral"}
+
+
 def test_build_root_task_empty_config() -> None:
     assert build_root_task({}) == ""