llm-gemini 0.17__tar.gz → 0.18.1__tar.gz

{llm_gemini-0.17 → llm_gemini-0.18.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llm-gemini
-Version: 0.17
+Version: 0.18.1
 Summary: LLM plugin to access Google's Gemini family of models
 Author: Simon Willison
 License: Apache-2.0
@@ -66,6 +66,7 @@ llm "A joke about a pelican and a walrus"
 
 Other models are:
 
+- `gemini-2.5-flash-preview-04-17` - Gemini 2.5 Flash preview
 - `gemini-2.5-pro-exp-03-25` - free experimental release of Gemini 2.5 Pro
 - `gemini-2.5-pro-preview-03-25` - paid preview of Gemini 2.5 Pro
 - `gemma-3-27b-it` - [Gemma 3](https://blog.google/technology/developers/gemma-3/) 27B

{llm_gemini-0.17 → llm_gemini-0.18.1}/README.md

@@ -43,6 +43,7 @@ llm "A joke about a pelican and a walrus"
 
 Other models are:
 
+- `gemini-2.5-flash-preview-04-17` - Gemini 2.5 Flash preview
 - `gemini-2.5-pro-exp-03-25` - free experimental release of Gemini 2.5 Pro
 - `gemini-2.5-pro-preview-03-25` - paid preview of Gemini 2.5 Pro
 - `gemma-3-27b-it` - [Gemma 3](https://blog.google/technology/developers/gemma-3/) 27B

{llm_gemini-0.17 → llm_gemini-0.18.1}/llm_gemini.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llm-gemini
-Version: 0.17
+Version: 0.18.1
 Summary: LLM plugin to access Google's Gemini family of models
 Author: Simon Willison
 License: Apache-2.0
@@ -66,6 +66,7 @@ llm "A joke about a pelican and a walrus"
 
 Other models are:
 
+- `gemini-2.5-flash-preview-04-17` - Gemini 2.5 Flash preview
 - `gemini-2.5-pro-exp-03-25` - free experimental release of Gemini 2.5 Pro
 - `gemini-2.5-pro-preview-03-25` - paid preview of Gemini 2.5 Pro
 - `gemma-3-27b-it` - [Gemma 3](https://blog.google/technology/developers/gemma-3/) 27B

{llm_gemini-0.17 → llm_gemini-0.18.1}/llm_gemini.py

@@ -36,6 +36,23 @@ GOOGLE_SEARCH_MODELS = {
     "gemini-1.5-flash-002",
     "gemini-2.0-flash-exp",
     "gemini-2.0-flash",
+    "gemini-2.5-pro-preview-03-25",
+    "gemini-2.5-pro-exp-03-25",
+}
+
+# Older Google models used google_search_retrieval instead of google_search
+GOOGLE_SEARCH_MODELS_USING_SEARCH_RETRIEVAL = {
+    "gemini-1.5-pro-latest",
+    "gemini-1.5-flash-latest",
+    "gemini-1.5-pro-001",
+    "gemini-1.5-flash-001",
+    "gemini-1.5-pro-002",
+    "gemini-1.5-flash-002",
+    "gemini-2.0-flash-exp",
+}
+
+THINKING_BUDGET_MODELS = {
+    "gemini-2.5-flash-preview-04-17",
 }
 
 
@@ -70,17 +87,22 @@ def register_models(register):
         "gemini-2.5-pro-exp-03-25",
         # 4th April 2025 (paid):
         "gemini-2.5-pro-preview-03-25",
+        # 17th April 2025:
+        "gemini-2.5-flash-preview-04-17",
     ]:
         can_google_search = model_id in GOOGLE_SEARCH_MODELS
+        can_thinking_budget = model_id in THINKING_BUDGET_MODELS
         register(
             GeminiPro(
                 model_id,
                 can_google_search=can_google_search,
+                can_thinking_budget=can_thinking_budget,
                 can_schema="flash-thinking" not in model_id,
             ),
             AsyncGeminiPro(
                 model_id,
                 can_google_search=can_google_search,
+                can_thinking_budget=can_thinking_budget,
                 can_schema="flash-thinking" not in model_id,
             ),
         )
@@ -208,12 +230,27 @@ class _SharedGemini:
             default=None,
         )
 
-    def __init__(self, model_id, can_google_search=False, can_schema=False):
+    class OptionsWithThinkingBudget(OptionsWithGoogleSearch):
+        thinking_budget: Optional[int] = Field(
+            description="Indicates the thinking budget in tokens. Set to 0 to disable.",
+            default=None,
+        )
+
+    def __init__(
+        self,
+        model_id,
+        can_google_search=False,
+        can_thinking_budget=False,
+        can_schema=False,
+    ):
         self.model_id = model_id
         self.can_google_search = can_google_search
         self.supports_schema = can_schema
         if can_google_search:
             self.Options = self.OptionsWithGoogleSearch
+        self.can_thinking_budget = can_thinking_budget
+        if can_thinking_budget:
+            self.Options = self.OptionsWithThinkingBudget
 
     def build_messages(self, prompt, conversation):
         messages = []
@@ -262,14 +299,27 @@ class _SharedGemini:
         if prompt.options and prompt.options.code_execution:
             body["tools"] = [{"codeExecution": {}}]
         if prompt.options and self.can_google_search and prompt.options.google_search:
-            body["tools"] = [{"google_search": {}}]
+            tool_name = (
+                "google_search_retrieval"
+                if self.model_id in GOOGLE_SEARCH_MODELS_USING_SEARCH_RETRIEVAL
+                else "google_search"
+            )
+            body["tools"] = [{tool_name: {}}]
         if prompt.system:
             body["systemInstruction"] = {"parts": [{"text": prompt.system}]}
 
+        generation_config = {}
+
         if prompt.schema:
-            body["generationConfig"] = {
-                "response_mime_type": "application/json",
-                "response_schema": cleanup_schema(copy.deepcopy(prompt.schema)),
+            generation_config.update(
+                {
+                    "response_mime_type": "application/json",
+                    "response_schema": cleanup_schema(copy.deepcopy(prompt.schema)),
+                }
+            )
+        if self.can_thinking_budget and prompt.options.thinking_budget is not None:
+            generation_config["thinking_config"] = {
+                "thinking_budget": prompt.options.thinking_budget
             }
 
         config_map = {
@@ -279,16 +329,17 @@ class _SharedGemini:
             "top_k": "topK",
         }
         if prompt.options and prompt.options.json_object:
-            body["generationConfig"] = {"response_mime_type": "application/json"}
+            generation_config["response_mime_type"] = "application/json"
 
         if any(
             getattr(prompt.options, key, None) is not None for key in config_map.keys()
         ):
-            generation_config = {}
             for key, other_key in config_map.items():
                 config_value = getattr(prompt.options, key, None)
                 if config_value is not None:
                     generation_config[other_key] = config_value
+
+        if generation_config:
             body["generationConfig"] = generation_config
 
         return body
@@ -458,9 +509,12 @@ def register_commands(cli):
     def models(key):
         "List of Gemini models pulled from their API"
         key = llm.get_key(key, "gemini", "LLM_GEMINI_KEY")
-        response = httpx.get(
-            f"https://generativelanguage.googleapis.com/v1beta/models?key={key}",
-        )
+        if not key:
+            raise click.ClickException(
+                "You must set the LLM_GEMINI_KEY environment variable or use --key"
+            )
+        url = f"https://generativelanguage.googleapis.com/v1beta/models"
+        response = httpx.get(url, headers={"x-goog-api-key": key})
         response.raise_for_status()
         click.echo(json.dumps(response.json()["models"], indent=2))
 

{llm_gemini-0.17 → llm_gemini-0.18.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "llm-gemini"
-version = "0.17"
+version = "0.18.1"
 description = "LLM plugin to access Google's Gemini family of models"
 readme = "README.md"
 authors = [{name = "Simon Willison"}]

{llm_gemini-0.17 → llm_gemini-0.18.1}/tests/test_gemini.py

@@ -1,4 +1,6 @@
+from click.testing import CliRunner
 import llm
+from llm.cli import cli
 import nest_asyncio
 import json
 import os
@@ -210,3 +212,22 @@ def test_cleanup_schema(schema, expected):
     # Use a deep copy so the original test data remains unchanged.
     result = cleanup_schema(schema)
     assert result == expected
+
+
+@pytest.mark.vcr
+def test_cli_gemini_models(tmpdir, monkeypatch):
+    user_dir = tmpdir / "llm.datasette.io"
+    user_dir.mkdir()
+    monkeypatch.setenv("LLM_USER_PATH", str(user_dir))
+    # With no key set should error nicely
+    runner = CliRunner()
+    result = runner.invoke(cli, ["gemini", "models"])
+    assert result.exit_code == 1
+    assert (
+        "Error: You must set the LLM_GEMINI_KEY environment variable or use --key\n"
+        == result.output
+    )
+    # Try again with --key
+    result2 = runner.invoke(cli, ["gemini", "models", "--key", GEMINI_API_KEY])
+    assert result2.exit_code == 0
+    assert "gemini-1.5-flash-latest" in result2.output