inferencesh 0.2.16__py3-none-any.whl → 0.2.17__py3-none-any.whl

inferencesh/models/llm.py

@@ -228,6 +228,7 @@ class ResponseTransformer:
     def __init__(self, output_cls: type[LLMOutput] = LLMOutput):
         self.state = ResponseState()
         self.output_cls = output_cls
+        self.timing = None  # Will be set by stream_generate
     
     def clean_text(self, text: str) -> str:
         """Clean common tokens from the text and apply model-specific cleaning.
@@ -264,10 +265,17 @@ class ResponseTransformer:
             text: Cleaned text to process for reasoning
         """
         # Default implementation for <think> style reasoning
-        if "<think>" in text:
+        if "<think>" in text and not self.state.state_changes["reasoning_started"]:
             self.state.state_changes["reasoning_started"] = True
-        if "</think>" in text:
+            if self.timing:
+                self.timing.start_reasoning()
+        
+        if "</think>" in text and not self.state.state_changes["reasoning_ended"]:
             self.state.state_changes["reasoning_ended"] = True
+            if self.timing:
+                # Estimate token count from character count (rough approximation)
+                token_count = len(self.state.buffer.split("<think>")[1].split("</think>")[0]) // 4
+                self.timing.end_reasoning(token_count)
         
         if "<think>" in self.state.buffer:
             parts = self.state.buffer.split("</think>", 1)
@@ -381,6 +389,9 @@ def stream_generate(
     }
 
     with timing_context() as timing:
+        # Set timing context in transformer
+        transformer.timing = timing
+        
         def generation_thread():
             nonlocal thread_exception, usage_stats
             try:

{inferencesh-0.2.16.dist-info → inferencesh-0.2.17.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: inferencesh
-Version: 0.2.16
+Version: 0.2.17
 Summary: inference.sh Python SDK
 Author: Inference Shell Inc.
 Author-email: "Inference Shell Inc." <hello@inference.sh>

{inferencesh-0.2.16.dist-info → inferencesh-0.2.17.dist-info}/RECORD

@@ -2,13 +2,13 @@ inferencesh/__init__.py,sha256=WdADtOhfa3HDOunoE9HLFCTFlXRykYstBIH1FpyWvj8,613
 inferencesh/models/__init__.py,sha256=FDwcdtT6c4hbRitymjmN-hZMlQa8RbKSftkZZyjtUXA,536
 inferencesh/models/base.py,sha256=4gZQRi8J7y9U6PrGD9pRIehd1MJVJAqGakPQDs2AKFM,3251
 inferencesh/models/file.py,sha256=5xnpypcRahM1YcEjj64rv9g2gTimxrZb41YT4r440hU,7393
-inferencesh/models/llm.py,sha256=8wZATOPTYNgtzl34vKeQaCXDbpyTMCzA4jOGrxbo5L4,18380
+inferencesh/models/llm.py,sha256=ndRFWbnAD1-BQFTTEb2YShGP-gjVd7w80ItT4XFV1_U,18983
 inferencesh/utils/__init__.py,sha256=-xiD6uo2XzcrPAWFb_fUbaimmnW4KFKc-8IvBzaxNd4,148
 inferencesh/utils/download.py,sha256=7n5twvoNYDcFnKJyefImaj2YfzRI7vddQw4usZbj38c,1521
 inferencesh/utils/storage.py,sha256=E4J8emd4eFKdmdDgAqzz3TpaaDd3n0l8gYlMHuY8yIU,519
-inferencesh-0.2.16.dist-info/licenses/LICENSE,sha256=OsgqEWIh2el_QMj0y8O1A5Q5Dl-dxqqYbFE6fszuR4s,1086
-inferencesh-0.2.16.dist-info/METADATA,sha256=wCILA3L4dmdzeKDEc3oNKH1etYwwYZjsCC7bfzepCa0,2757
-inferencesh-0.2.16.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-inferencesh-0.2.16.dist-info/entry_points.txt,sha256=6IC-fyozAqW3ljsMLGCXxJ0_ui2Jb-2fLHtoH1RTnEE,45
-inferencesh-0.2.16.dist-info/top_level.txt,sha256=TSMHg3T1ThMl1HGAWmzBClwOYH1ump5neof9BfHIwaA,12
-inferencesh-0.2.16.dist-info/RECORD,,
+inferencesh-0.2.17.dist-info/licenses/LICENSE,sha256=OsgqEWIh2el_QMj0y8O1A5Q5Dl-dxqqYbFE6fszuR4s,1086
+inferencesh-0.2.17.dist-info/METADATA,sha256=gOHSF2qhb_XT_tXlHqp5HAVblDDZzOjM9kGcbp99JQ4,2757
+inferencesh-0.2.17.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+inferencesh-0.2.17.dist-info/entry_points.txt,sha256=6IC-fyozAqW3ljsMLGCXxJ0_ui2Jb-2fLHtoH1RTnEE,45
+inferencesh-0.2.17.dist-info/top_level.txt,sha256=TSMHg3T1ThMl1HGAWmzBClwOYH1ump5neof9BfHIwaA,12
+inferencesh-0.2.17.dist-info/RECORD,,