greycloud 0.4.0__tar.gz → 0.4.2__tar.gz

{greycloud-0.4.0/greycloud.egg-info → greycloud-0.4.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: greycloud
-Version: 0.4.0
+Version: 0.4.2
 Summary: Reusable Google GenAI/Vertex AI Client Module
 Author: GreyCloud Contributors
 License: MIT
@@ -21,7 +21,7 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: google-genai>=1.59.0
+Requires-Dist: google-genai>=2.31.0
 Requires-Dist: google-auth>=2.0.0
 Requires-Dist: google-cloud-storage>=3.8.0
 Requires-Dist: aiolimiter>=1.1.0
@@ -177,7 +177,7 @@ config = GreyCloudConfig(
     project_id="your-project-id",
     location="us-central1",
     # Default model is a Gemini 3 flash model; you can override if desired.
-    model="gemini-3-flash-preview",
+    model="gemini-3.8-flash",
 )
 
 # Create client

{greycloud-0.4.0 → greycloud-0.4.2}/README.md

@@ -135,7 +135,7 @@ config = GreyCloudConfig(
     project_id="your-project-id",
     location="us-central1",
     # Default model is a Gemini 3 flash model; you can override if desired.
-    model="gemini-3-flash-preview",
+    model="gemini-3.8-flash",
 )
 
 # Create client

{greycloud-0.4.0 → greycloud-0.4.2}/greycloud/__init__.py

@@ -13,7 +13,7 @@ from .cache import GreyCloudCache
 from .rate_limiter import VertexRateLimiter
 from .async_client import GreyCloudAsyncClient
 
-__version__ = "0.4.0"
+__version__ = "0.4.2"
 __all__ = [
     "GreyCloudConfig",
     "GreyCloudClient",

{greycloud-0.4.0 → greycloud-0.4.2}/greycloud/cache.py

@@ -27,8 +27,8 @@ MIN_TOKENS_BY_MODEL = {
     "gemini-2.0": 2048,
     "gemini-2.5-flash": 1024,
     "gemini-2.5-pro": 4096,
-    "gemini-3.5-flash": 1024,
-    "gemini-3.1-pro": 4096,
+    "gemini-3.8-flash": 1024,
+    "gemini-4-argon": 4096,
 }
 
 # Default minimum if model not recognized

{greycloud-0.4.0 → greycloud-0.4.2}/greycloud/config.py

@@ -48,7 +48,7 @@ class GreyCloudConfig:
     # Model configuration
     # Default to a generally available Gemini 3 flash model.
     # You can override this per-request when calling GreyCloudClient.
-    model: str = "gemini-3-flash-preview"
+    model: str = "gemini-3.8-flash"
     endpoint: str = "https://aiplatform.googleapis.com"
     api_version: str = "v1"
 

{greycloud-0.4.0 → greycloud-0.4.2}/greycloud/rate_limiter.py

@@ -44,7 +44,16 @@ class VertexRateLimiter:
             Whatever the underlying API call raises.
         """
         await self._rpm_limiter.acquire()
-        await self._tpm_limiter.acquire(prompt_tokens_est)
+        # Clamp the TPM acquisition to the bucket capacity: a single request
+        # cannot be split, so one whose estimate exceeds the per-minute budget
+        # must still go through (it waits for a fresh window at worst) rather
+        # than being categorically rejected. Without the clamp, any request
+        # estimated above `tpm` — e.g. long-context prompts carrying a large
+        # injected knowledge block — raises aiolimiter's
+        # "Can't acquire more than the maximum capacity" ValueError on every
+        # attempt and the client retries until the caller gives up. The
+        # server's real TPM quota is enforced by Vertex itself.
+        await self._tpm_limiter.acquire(min(prompt_tokens_est, self.tpm))
         async with self._semaphore:
             return await api_call_coro
 

{greycloud-0.4.0 → greycloud-0.4.2/greycloud.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: greycloud
-Version: 0.4.0
+Version: 0.4.2
 Summary: Reusable Google GenAI/Vertex AI Client Module
 Author: GreyCloud Contributors
 License: MIT
@@ -21,7 +21,7 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: google-genai>=1.59.0
+Requires-Dist: google-genai>=2.31.0
 Requires-Dist: google-auth>=2.0.0
 Requires-Dist: google-cloud-storage>=3.8.0
 Requires-Dist: aiolimiter>=1.1.0
@@ -177,7 +177,7 @@ config = GreyCloudConfig(
     project_id="your-project-id",
     location="us-central1",
     # Default model is a Gemini 3 flash model; you can override if desired.
-    model="gemini-3-flash-preview",
+    model="gemini-3.8-flash",
 )
 
 # Create client

{greycloud-0.4.0 → greycloud-0.4.2}/greycloud.egg-info/requires.txt

@@ -1,4 +1,4 @@
-google-genai>=1.59.0
+google-genai>=2.31.0
 google-auth>=2.0.0
 google-cloud-storage>=3.8.0
 aiolimiter>=1.1.0

{greycloud-0.4.0 → greycloud-0.4.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "greycloud"
-version = "0.4.0"
+version = "0.4.2"
 description = "Reusable Google GenAI/Vertex AI Client Module"
 readme = "README.md"
 # google-genai >= 1.48.0 requires Python >= 3.10.
@@ -28,7 +28,7 @@ classifiers = [
 
 dependencies = [
     # Use a modern google-genai that aligns with Python >= 3.10.
-    "google-genai>=1.59.0",
+    "google-genai>=2.31.0",
     "google-auth>=2.0.0",
     "google-cloud-storage>=3.8.0",
     "aiolimiter>=1.1.0",

{greycloud-0.4.0 → greycloud-0.4.2}/tests/test_batch.py

@@ -227,7 +227,7 @@ class TestGreyCloudBatch:
         """Test batch requests upload"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Test"}]}],
                 config=types.GenerateContentConfig(temperature=0.2),
             )
@@ -261,7 +261,7 @@ class TestGreyCloudBatch:
                 obj = json.loads(lines[0])
                 assert "request" in obj
                 req = obj["request"]
-                assert req["model"] == "gemini-3.1-pro-preview"
+                assert req["model"] == "gemini-4-argon"
                 assert "contents" in req
                 # Config should be serialized as generationConfig (not "config")
                 assert "generationConfig" in req
@@ -274,7 +274,7 @@ class TestGreyCloudBatch:
         """Test that JSONL serialization uses camelCase field names"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Test"}]}],
                 config=types.GenerateContentConfig(
                     temperature=0.5,
@@ -325,7 +325,7 @@ class TestGreyCloudBatch:
         """Test that systemInstruction is serialized as top-level request field"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Hello"}]}],
                 config=types.GenerateContentConfig(
                     system_instruction="You are a helpful assistant.",
@@ -368,7 +368,7 @@ class TestGreyCloudBatch:
         thinking = types.ThinkingConfig(thinking_budget=2048)
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Think"}]}],
                 config=types.GenerateContentConfig(
                     temperature=0.5,
@@ -404,7 +404,7 @@ class TestGreyCloudBatch:
         """Test that None config values are omitted from JSONL"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Test"}]}],
                 config=types.GenerateContentConfig(temperature=0.5),
             )
@@ -437,7 +437,7 @@ class TestGreyCloudBatch:
         """Test that InlinedRequest.metadata is NOT forwarded to JSONL"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Test"}]}],
                 metadata={"request_id": "123"},
             )
@@ -469,7 +469,7 @@ class TestGreyCloudBatch:
         """Test batch job creation"""
         batch_requests = [
             types.InlinedRequest(
-                model="gemini-3.1-pro-preview",
+                model="gemini-4-argon",
                 contents=[{"role": "user", "parts": [{"text": "Test"}]}],
                 config=types.GenerateContentConfig(),
             )

{greycloud-0.4.0 → greycloud-0.4.2}/tests/test_cache.py

@@ -27,14 +27,14 @@ class TestGetMinTokensForModel:
         assert get_min_tokens_for_model("gemini-2.5-pro") == 4096
         assert get_min_tokens_for_model("gemini-2.5-pro-preview") == 4096
 
-    def test_gemini_3_flash(self):
-        """Gemini 3 Flash requires 1024 tokens"""
-        assert get_min_tokens_for_model("gemini-3.5-flash") == 1024
+    def test_gemini_3_8_flash(self):
+        """Gemini 3.8 Flash requires 1024 tokens"""
+        assert get_min_tokens_for_model("gemini-3.8-flash") == 1024
 
-    def test_gemini_3_pro(self):
-        """Gemini 3 Pro requires 4096 tokens"""
-        assert get_min_tokens_for_model("gemini-3.1-pro") == 4096
-        assert get_min_tokens_for_model("gemini-3.1-pro-preview") == 4096
+    def test_gemini_4_argon(self):
+        """Gemini 4 Argon requires 4096 tokens"""
+        assert get_min_tokens_for_model("gemini-4-argon") == 4096
+        assert get_min_tokens_for_model("gemini-4-argon-preview") == 4096
 
     def test_gemini_2_0(self):
         """Gemini 2.0 requires 2048 tokens"""
@@ -47,7 +47,7 @@ class TestGetMinTokensForModel:
     def test_case_insensitive(self):
         """Model name matching is case insensitive"""
         assert get_min_tokens_for_model("GEMINI-2.5-FLASH") == 1024
-        assert get_min_tokens_for_model("gemini-3.1-pro") == 4096
+        assert get_min_tokens_for_model("gemini-4-argon") == 4096
 
 
 class TestGreyCloudCacheInit:
@@ -580,7 +580,7 @@ class TestGreyCloudCacheGetCacheInfo:
         mock_cache = MagicMock(spec=types.CachedContent)
         mock_cache.name = "cachedContents/test-cache"
         mock_cache.display_name = "Test Cache"
-        mock_cache.model = "gemini-3.5-flash"
+        mock_cache.model = "gemini-3.8-flash"
         mock_cache.create_time = datetime.datetime(2024, 1, 1, 12, 0, 0)
         mock_cache.expire_time = datetime.datetime(2024, 1, 1, 13, 0, 0)
 
@@ -592,7 +592,7 @@ class TestGreyCloudCacheGetCacheInfo:
 
         assert info["name"] == "cachedContents/test-cache"
         assert info["display_name"] == "Test Cache"
-        assert info["model"] == "gemini-3.5-flash"
+        assert info["model"] == "gemini-3.8-flash"
         assert info["total_token_count"] == 50000
 
 

{greycloud-0.4.0 → greycloud-0.4.2}/tests/test_config.py

@@ -55,7 +55,7 @@ class TestGreyCloudConfig:
         config = GreyCloudConfig(project_id="test-project")
 
         # Defaults
-        assert config.model == "gemini-3-flash-preview"
+        assert config.model == "gemini-3.8-flash"
         assert config.temperature == 1.0
         assert config.top_p == 0.95
         assert config.max_output_tokens == 65535

{greycloud-0.4.0 → greycloud-0.4.2}/tests/test_init.py

@@ -16,7 +16,7 @@ class TestPackageInit:
 
     def test_version(self):
         """Test package version"""
-        assert __version__ == "0.4.0"
+        assert __version__ == "0.4.2"
 
     def test_all_exports(self):
         """Test __all__ exports"""

{greycloud-0.4.0 → greycloud-0.4.2}/tests/test_rate_limiter.py

@@ -79,3 +79,42 @@ class TestVertexRateLimiterEstimateTokens:
     def test_estimate_empty_string(self):
         """Empty string returns 0"""
         assert VertexRateLimiter.estimate_tokens("") == 0
+
+    @pytest.mark.asyncio
+    async def test_tpm_acquisition_clamped_to_capacity(self, monkeypatch):
+        """A request whose token estimate exceeds the TPM bucket capacity must
+        still proceed (clamped to capacity, waiting for a fresh window at
+        worst) instead of raising aiolimiter's 'Can't acquire more than the
+        maximum capacity' ValueError — a single request cannot be split, and
+        the server's real TPM quota is enforced by Vertex itself. Regression
+        for BlackSheep's long-context prompts (injected knowledge block plus
+        case history estimated above the 250K/min default bucket)."""
+        limiter = VertexRateLimiter(rpm=60, tpm=250_000, max_concurrency=1)
+        mock_coro = AsyncMock(return_value="done")
+
+        # Estimate far above capacity: must not raise.
+        result = await limiter.call_with_limits(500_000, mock_coro())
+        assert result == "done"
+        mock_coro.assert_awaited_once()
+
+        # The bucket was fully consumed by the clamped acquisition: a second
+        # request in the same window must wait for capacity to replenish
+        # (not fail), proving the clamp actually acquired.
+        # The clamped acquisition amount: patch the bucket class to record
+        # the requested amount, asserting the request acquired exactly the
+        # bucket capacity, not the raw estimate.
+        import aiolimiter
+
+        acquired = []
+        real_acquire = aiolimiter.AsyncLimiter.acquire
+
+        async def spy_acquire(self, amount=1):
+            acquired.append(amount)
+            await real_acquire(self, amount)
+
+        monkeypatch.setattr(aiolimiter.AsyncLimiter, "acquire", spy_acquire)
+        limiter2 = VertexRateLimiter(rpm=60, tpm=250_000, max_concurrency=1)
+        await limiter2.call_with_limits(500_000, AsyncMock(return_value="x")())
+        # acquired[0] is the RPM bucket's acquire(1); the TPM acquisition
+        # (clamped to capacity, not the 500K estimate) comes last.
+        assert acquired[-1] == 250_000