Add 1-hour cache write pricing for us-gov Haiku 4.5 (#28574)

* fix(thinking): handle None thinking param in is_thinking_enabled (#28598)

Squash-merged by litellm-agent from Terrajlz's PR.

* feat(helm): support tpl rendering in podAnnotations (#28609)

Squash-merged by litellm-agent from devauxbr's PR.

* Add 1-hour cache write pricing for us-gov Haiku 4.5

AWS Bedrock GovCloud applies a +20% premium over global Anthropic
rates. Global Haiku 4.5 5m/1h cache write is $1.25 / $2.00 per MTok;
us-gov is therefore $1.50 / $2.40 per MTok (the 5m rate was already
correct in litellm; the 1h field was missing).

Adds `cache_creation_input_token_cost_above_1hr: 2.4e-06` to the
two us-gov Haiku 4.5 entries in both pricing JSON files:

  - bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0
  - bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0

New parametrized regression test
tests/test_litellm/test_bedrock_usgov_haiku_1hr_cache.py pins both
entries and enforces the 1.6x 5m-to-1h ratio invariant matching the
pattern used by the existing Bedrock and Vertex 1h-cache tests.

Companion to the us-gov Sonnet 4.5 pricing fix.

* chore: trigger shin-agent re-eval on retargeted staging base

* chore: trigger shin-agent re-eval against updated Greptile state

---------

Co-authored-by: Terrajlz <info@jouleselectrictech.com>
Co-authored-by: Bruno Devaux <devaux.br@gmail.com>
Co-authored-by: Sameer Kankute <sameer@berri.ai>

litellm/model_prices_and_context_window_backup.json

@@ -41420,6 +41420,7 @@
     },
     "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": {
         "cache_creation_input_token_cost": 1.5e-06,
+        "cache_creation_input_token_cost_above_1hr": 2.4e-06,
         "cache_read_input_token_cost": 1.2e-07,
         "input_cost_per_token": 1.2e-06,
         "litellm_provider": "bedrock",
@@ -41442,6 +41443,7 @@
     },
     "bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0": {
         "cache_creation_input_token_cost": 1.5e-06,
+        "cache_creation_input_token_cost_above_1hr": 2.4e-06,
         "cache_read_input_token_cost": 1.2e-07,
         "input_cost_per_token": 1.2e-06,
         "litellm_provider": "bedrock",

model_prices_and_context_window.json

@@ -41304,6 +41304,7 @@
     },
     "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": {
         "cache_creation_input_token_cost": 1.5e-06,
+        "cache_creation_input_token_cost_above_1hr": 2.4e-06,
         "cache_read_input_token_cost": 1.2e-07,
         "input_cost_per_token": 1.2e-06,
         "litellm_provider": "bedrock",
@@ -41326,6 +41327,7 @@
     },
     "bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0": {
         "cache_creation_input_token_cost": 1.5e-06,
+        "cache_creation_input_token_cost_above_1hr": 2.4e-06,
         "cache_read_input_token_cost": 1.2e-07,
         "input_cost_per_token": 1.2e-06,
         "litellm_provider": "bedrock",

tests/test_litellm/test_bedrock_usgov_haiku_1hr_cache.py

@@ -0,0 +1,47 @@
+"""
+Validate that AWS GovCloud (Bedrock us-gov-*) Haiku 4.5 entries carry
+the 1-hour cache write tier.
+
+AWS Bedrock GovCloud pricing applies a +20% premium over global
+Anthropic rates. Global Haiku 4.5 1h cache write is $2.00/MTok; us-gov
+is therefore $2.40/MTok — exactly 1.6x the 5-minute rate of $1.50/MTok.
+
+Source: https://aws.amazon.com/bedrock/pricing/
+"""
+
+import json
+import os
+
+import pytest
+
+
+@pytest.fixture(scope="module")
+def model_data():
+    json_path = os.path.join(
+        os.path.dirname(__file__), "../../model_prices_and_context_window.json"
+    )
+    with open(json_path) as f:
+        return json.load(f)
+
+
+HAIKU_USGOV_KEYS = [
+    "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0",
+    "bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0",
+]
+
+
+@pytest.mark.parametrize("model_key", HAIKU_USGOV_KEYS)
+def test_usgov_haiku_4_5_1hr_cache_write(model_data, model_key):
+    assert model_key in model_data, f"Missing model entry: {model_key}"
+    info = model_data[model_key]
+    assert (
+        info["cache_creation_input_token_cost"] == 1.5e-06
+    ), f"{model_key}: 5m cache write should be $1.50/MTok"
+    assert (
+        info["cache_creation_input_token_cost_above_1hr"] == 2.4e-06
+    ), f"{model_key}: 1h cache write should be $2.40/MTok"
+    ratio = (
+        info["cache_creation_input_token_cost_above_1hr"]
+        / info["cache_creation_input_token_cost"]
+    )
+    assert abs(ratio - 1.6) < 1e-9, f"{model_key}: 1h/5m ratio is {ratio}, expected 1.6"