unique_toolkit 0.8.8__tar.gz → 0.8.9__tar.gz

{unique_toolkit-0.8.8 → unique_toolkit-0.8.9}/CHANGELOG.md

@@ -5,6 +5,9 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.8.9] - 2025-08-15
+- Reduce input token limits for `ANTHROPIC_CLAUDE_3_7_SONNET_THINKING`, `ANTHROPIC_CLAUDE_3_7_SONNET`, `ANTHROPIC_CLAUDE_OPUS_4` and `ANTHROPIC_CLAUDE_SONNET_4` to 180_000 from 200_000
+
 ## [0.8.8] - 2025-08-11
 - Make chat service openai stream response openai compatible 
 - Make `ChatMessage` openai compatible

{unique_toolkit-0.8.8 → unique_toolkit-0.8.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 0.8.8
+Version: 0.8.9
 Summary: 
 License: Proprietary
 Author: Martin Fadler
@@ -114,6 +114,9 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.8.9] - 2025-08-15
+- Reduce input token limits for `ANTHROPIC_CLAUDE_3_7_SONNET_THINKING`, `ANTHROPIC_CLAUDE_3_7_SONNET`, `ANTHROPIC_CLAUDE_OPUS_4` and `ANTHROPIC_CLAUDE_SONNET_4` to 180_000 from 200_000
+
 ## [0.8.8] - 2025-08-11
 - Make chat service openai stream response openai compatible 
 - Make `ChatMessage` openai compatible

{unique_toolkit-0.8.8 → unique_toolkit-0.8.9}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "unique_toolkit"
-version = "0.8.8"
+version = "0.8.9"
 description = ""
 authors = [
     "Martin Fadler <martin.fadler@unique.ch>",

{unique_toolkit-0.8.8 → unique_toolkit-0.8.9}/unique_toolkit/language_model/infos.py

@@ -507,7 +507,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-3-7-sonnet",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=128_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=128_000,
                     ),
                     info_cutoff_at=date(2024, 10, 31),
                     published_at=date(2025, 2, 24),
@@ -525,7 +527,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-3-7-sonnet-thinking",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=128_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=128_000,
                     ),
                     info_cutoff_at=date(2024, 10, 31),
                     published_at=date(2025, 2, 24),
@@ -543,7 +547,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-sonnet-4",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=64_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=64_000,
                     ),
                     info_cutoff_at=date(2025, 3, 1),
                     published_at=date(2025, 5, 1),
@@ -561,7 +567,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-opus-4",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=32_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=32_000,
                     ),
                     info_cutoff_at=date(2025, 3, 1),
                     published_at=date(2025, 5, 1),