lm-deluge 0.0.25__py3-none-any.whl → 0.0.27__py3-none-any.whl

lm_deluge/api_requests/gemini.py

@@ -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/file.py

@@ -1,3 +1,4 @@
+from functools import cached_property
 import os
 import io
 import requests
@@ -68,13 +69,13 @@ class File:
             return encoded
         return f"data:{self._mime()};base64,{encoded}"
 
-    @property
+    @cached_property
     def fingerprint(self) -> str:
         # Hash the file contents for fingerprinting
         file_bytes = self._bytes()
         return xxhash.xxh64(file_bytes).hexdigest()
 
-    @property
+    @cached_property
     def size(self) -> int:
         """Return file size in bytes."""
         return len(self._bytes())

lm_deluge/request_context.py

@@ -1,4 +1,5 @@
 from dataclasses import dataclass, field
+from functools import cached_property
 from typing import Any, Callable
 
 from .config import SamplingParams
@@ -39,14 +40,18 @@ class RequestContext:
 
     # Computed properties
     cache_key: str = field(init=False)
-    num_tokens: int = field(init=False)
+    # num_tokens: int = field(init=False)
 
-    def __post_init__(self):
-        # Compute cache key from prompt fingerprint
-        self.cache_key = self.prompt.fingerprint
+    # def __post_init__(self):
+    #     # Compute cache key from prompt fingerprint
+    #     # self.cache_key = self.prompt.fingerprint
 
-        # Compute token count
-        self.num_tokens = self.prompt.count_tokens(self.sampling_params.max_new_tokens)
+    #     # Compute token count
+    #     self.num_tokens =
+
+    @cached_property
+    def num_tokens(self):
+        return self.prompt.count_tokens(self.sampling_params.max_new_tokens)
 
     def maybe_callback(self, response, tracker):
         if not self.callback:

{lm_deluge-0.0.25.dist-info → lm_deluge-0.0.27.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.25
+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-0.0.25.dist-info → lm_deluge-0.0.27.dist-info}/RECORD

@@ -6,12 +6,12 @@ lm_deluge/client.py,sha256=nAGMwdUPDVx-x23hZF6U5Yhug6Zf5FT27RHj_kj8nZk,26369
 lm_deluge/config.py,sha256=H1tQyJDNHGFuwxqQNL5Z-CjWAC0luHSBA3iY_pxmACM,932
 lm_deluge/embed.py,sha256=CO-TOlC5kOTAM8lcnicoG4u4K664vCBwHF1vHa-nAGg,13382
 lm_deluge/errors.py,sha256=oHjt7YnxWbh-eXMScIzov4NvpJMo0-2r5J6Wh5DQ1tk,209
-lm_deluge/file.py,sha256=zQH1STMjCG9pczO7Fk9Jw0_0Pj_8CogcdIxTe4J4AJw,5414
+lm_deluge/file.py,sha256=ggOXcAOI-LAykXLb3N3WfLlHdBjQIJtW9ltmfKCIsNQ,5466
 lm_deluge/gemini_limits.py,sha256=V9mpS9JtXYz7AY6OuKyQp5TuIMRH1BVv9YrSNmGmHNA,1569
 lm_deluge/image.py,sha256=Qpa0k5yXfrpSaHzVUwW_TEn7yEgmwzYGL17Sa7-KhSA,7729
 lm_deluge/models.py,sha256=3vgI1BlfT4_Higev25QhhXJufQvsI6pd0yjF9YL0crA,49812
 lm_deluge/prompt.py,sha256=cfwzCAmT-1K0v7SfEMUrxpBkJGgf7IFlWfNLJrCcoBM,37025
-lm_deluge/request_context.py,sha256=0X-5m8BKn51rnnjzGDDXqbuSUEFGjdayirQjbvPcjMI,2425
+lm_deluge/request_context.py,sha256=o33LSEwnK6YPhZeulUoSE_VrdKCXiCQa0tjjixK2K6M,2540
 lm_deluge/rerank.py,sha256=-NBAJdHz9OB-SWWJnHzkFmeVO4wR6lFV7Vw-SxG7aVo,11457
 lm_deluge/tool.py,sha256=_coOKB9nPNVZoseMRumRyQ8BMR7_d0IlstzMHNT69JY,15732
 lm_deluge/tracker.py,sha256=-EkFDAklh5mclIFR-5SthAwNL4p1yKS8LUN7rhpOVPQ,9266
@@ -21,7 +21,7 @@ lm_deluge/api_requests/anthropic.py,sha256=Vi2iGQw6LGPW8d4X489Jg-dlEye529jf8i5Ys
 lm_deluge/api_requests/base.py,sha256=EVHNFtlttKbN7Tt1MnLaO-NjvKHPSV5CqlRv-OnpVAE,5593
 lm_deluge/api_requests/bedrock.py,sha256=p6leW5stnvb406lwbmVFmfTGxdEI-t0GfaTExpol3qk,10900
 lm_deluge/api_requests/common.py,sha256=BZ3vRO5TB669_UsNKugkkuFSzoLHOYJIKt4nV4sf4vc,422
-lm_deluge/api_requests/gemini.py,sha256=UeA4gsIePsIu6IQQ7izR359LLQQTi0Qky_ykjH1XQf4,7461
+lm_deluge/api_requests/gemini.py,sha256=W4NjQ0buBsdS7RYpzDahrXNQWMzDHRMLNRSphCOmIqg,7685
 lm_deluge/api_requests/mistral.py,sha256=S_LpOfCGbCVEROH_od3P-tYeNYTKFMamMTL-c_wFCBI,4597
 lm_deluge/api_requests/openai.py,sha256=jb_pBGqSbqs7SvbA45Odu3JkwS3jsKn-p3hG8-qkYbc,21509
 lm_deluge/api_requests/response.py,sha256=FtkVYk_rDH93Kj9pqbB-l7a4dQHzVr6ivKL9khYKLbs,5966
@@ -48,8 +48,8 @@ lm_deluge/util/logprobs.py,sha256=UkBZakOxWluaLqHrjARu7xnJ0uCHVfLGHJdnYlEcutk,11
 lm_deluge/util/spatial.py,sha256=BsF_UKhE-x0xBirc-bV1xSKZRTUhsOBdGqsMKme20C8,4099
 lm_deluge/util/validation.py,sha256=hz5dDb3ebvZrZhnaWxOxbNSVMI6nmaOODBkk0htAUhs,1575
 lm_deluge/util/xml.py,sha256=Ft4zajoYBJR3HHCt2oHwGfymGLdvp_gegVmJ-Wqk4Ck,10547
-lm_deluge-0.0.25.dist-info/licenses/LICENSE,sha256=uNNXGXPCw2TC7CUs7SEBkA-Mz6QBQFWUUEWDMgEs1dU,1058
-lm_deluge-0.0.25.dist-info/METADATA,sha256=I-AUU79sc9sxjRV8UAPVWtlmzg10voZ2irH-bLkh_9E,12978
-lm_deluge-0.0.25.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-lm_deluge-0.0.25.dist-info/top_level.txt,sha256=hqU-TJX93yBwpgkDtYcXyLr3t7TLSCCZ_reytJjwBaE,10
-lm_deluge-0.0.25.dist-info/RECORD,,
+lm_deluge-0.0.27.dist-info/licenses/LICENSE,sha256=uNNXGXPCw2TC7CUs7SEBkA-Mz6QBQFWUUEWDMgEs1dU,1058
+lm_deluge-0.0.27.dist-info/METADATA,sha256=vKjAfWLWBC9_fgSvYXlQQMxgUEf1E3n57Y4kZBy9_7I,13053
+lm_deluge-0.0.27.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+lm_deluge-0.0.27.dist-info/top_level.txt,sha256=hqU-TJX93yBwpgkDtYcXyLr3t7TLSCCZ_reytJjwBaE,10
+lm_deluge-0.0.27.dist-info/RECORD,,