lm-deluge 0.0.88__tar.gz → 0.0.89__tar.gz

{lm_deluge-0.0.88/src/lm_deluge.egg-info → lm_deluge-0.0.89}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.88
+Version: 0.0.89
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10

{lm_deluge-0.0.88 → lm_deluge-0.0.89}/pyproject.toml

@@ -3,7 +3,7 @@ requires = ["setuptools", "wheel"]
 
 [project]
 name = "lm_deluge"
-version = "0.0.88"
+version = "0.0.89"
 authors = [{ name = "Benjamin Anderson", email = "ben@trytaylor.ai" }]
 description = "Python utility for using LLM API models."
 readme = "README.md"

{lm_deluge-0.0.88 → lm_deluge-0.0.89}/src/lm_deluge/api_requests/gemini.py

@@ -52,6 +52,7 @@ async def _build_gemini_request(
 
     # Handle reasoning models (thinking)
     is_gemini_3 = "gemini-3" in model.name.lower()
+    is_gemini_3_flash = "gemini-3-flash" in model.name.lower()
     if is_gemini_3:
         # gemini3 MUST think
         if not sampling_params.reasoning_effort:
@@ -62,13 +63,24 @@ async def _build_gemini_request(
             if effort_key == "xhigh":
                 maybe_warn("WARN_XHIGH_TO_HIGH", model_name=model.name)
                 effort_key = "high"
-            level_map = {
-                "none": "low",
-                "minimal": "low",
-                "low": "low",
-                "medium": "high",  # change when supported
-                "high": "high",
-            }
+            if is_gemini_3_flash:
+                # Flash supports minimal, low, medium, high
+                level_map = {
+                    "none": "low",
+                    "minimal": "minimal",
+                    "low": "low",
+                    "medium": "medium",
+                    "high": "high",
+                }
+            else:
+                # Pro only supports low, high
+                level_map = {
+                    "none": "low",
+                    "minimal": "low",
+                    "low": "low",
+                    "medium": "high",
+                    "high": "high",
+                }
             effort = level_map[effort_key]
         thinking_config = {"thinkingLevel": effort}
         request_json["generationConfig"]["thinkingConfig"] = thinking_config

{lm_deluge-0.0.88 → lm_deluge-0.0.89}/src/lm_deluge/models/google.py

@@ -153,6 +153,19 @@ GOOGLE_MODELS = {
         # Note: >200k tokens pricing is $4/$18 per million
         "reasoning_model": True,
     },
+    "gemini-3-flash-preview": {
+        "id": "gemini-3-flash-preview",
+        "name": "gemini-3-flash-preview",
+        "api_base": "https://generativelanguage.googleapis.com/v1alpha",
+        "api_key_env_var": "GEMINI_API_KEY",
+        "supports_json": True,
+        "supports_logprobs": False,
+        "api_spec": "gemini",
+        "input_cost": 0.5,
+        "cached_input_cost": 0.125,  # estimated
+        "output_cost": 3.0,
+        "reasoning_model": True,
+    },
     # Gemini 2.5 Computer Use model
     "gemini-2.5-computer-use": {
         "id": "gemini-2.5-computer-use",

{lm_deluge-0.0.88 → lm_deluge-0.0.89/src/lm_deluge.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.88
+Version: 0.0.89
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10