PyPI - lm-deluge - Versions diffs - 0.0.26__tar.gz → 0.0.27__tar.gz - Mend

lm-deluge 0.0.26tar.gz → 0.0.27tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of lm-deluge might be problematic. Click here for more details.

Files changed (62) hide show

{lm_deluge-0.0.26/src/lm_deluge.egg-info → lm_deluge-0.0.27}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.26
+Version: 0.0.27
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10
@@ -274,7 +274,7 @@ We support all models in `src/lm_deluge/models.py`. Vertex support is not planne
 ## Feature Support
-We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
+We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Passing `None` (or the string `"none"`) disables Gemini thoughts entirely. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
 ## Built‑in tools

{lm_deluge-0.0.26 → lm_deluge-0.0.27}/README.md RENAMED Viewed

@@ -247,7 +247,7 @@ We support all models in `src/lm_deluge/models.py`. Vertex support is not planne
 ## Feature Support
-We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
+We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Passing `None` (or the string `"none"`) disables Gemini thoughts entirely. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
 ## Built‑in tools

{lm_deluge-0.0.26 → lm_deluge-0.0.27}/pyproject.toml RENAMED Viewed

@@ -3,7 +3,7 @@ requires = ["setuptools", "wheel"]
 [project]
 name = "lm_deluge"
-version = "0.0.26"
+version = "0.0.27"
 authors = [{ name = "Benjamin Anderson", email = "ben@trytaylor.ai" }]
 description = "Python utility for using LLM API models."
 readme = "README.md"

{lm_deluge-0.0.26 → lm_deluge-0.0.27}/src/lm_deluge/api_requests/gemini.py RENAMED Viewed

@@ -37,14 +37,17 @@ async def _build_gemini_request(
     # Handle reasoning models (thinking)
     if model.reasoning_model:
-        request_json["generationConfig"]["thinkingConfig"] = {"includeThoughts": True}
-        if sampling_params.reasoning_effort and "flash" in model.id:
-            budget = {"low": 1024, "medium": 4096, "high": 16384}.get(
-                sampling_params.reasoning_effort
-            )
-            request_json["generationConfig"]["thinkingConfig"]["thinkingBudget"] = (
-                budget
-            )
+        thinking_config = None
+        effort = sampling_params.reasoning_effort
+        if effort is None or effort == "none":
+            # Explicitly disable thoughts when no effort is requested
+            thinking_config = {"includeThoughts": False, "thinkingBudget": 0}
+        else:
+            thinking_config = {"includeThoughts": True}
+            if effort in {"low", "medium", "high"} and "flash" in model.id:
+                budget = {"low": 1024, "medium": 4096, "high": 16384}[effort]
+                thinking_config["thinkingBudget"] = budget
+        request_json["generationConfig"]["thinkingConfig"] = thinking_config
     else:
         if sampling_params.reasoning_effort:

{lm_deluge-0.0.26 → lm_deluge-0.0.27/src/lm_deluge.egg-info}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.26
+Version: 0.0.27
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10
@@ -274,7 +274,7 @@ We support all models in `src/lm_deluge/models.py`. Vertex support is not planne
 ## Feature Support
-We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
+We support structured outputs via `json_mode` parameter provided to `SamplingParams`. Structured outputs with a schema are planned. Reasoning models are supported via the `reasoning_effort` parameter, which is translated to a thinking budget for Claude/Gemini. Passing `None` (or the string `"none"`) disables Gemini thoughts entirely. Image models are supported. We support tool use as documented above. We support logprobs for OpenAI models that return them.
 ## Built‑in tools