code-puppy 0.0.90__tar.gz → 0.0.92__tar.gz

{code_puppy-0.0.90 → code_puppy-0.0.92}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-puppy
-Version: 0.0.90
+Version: 0.0.92
 Summary: Code generation agent
 Author: Michael Pfaffenberger
 License: MIT
@@ -27,7 +27,6 @@ Requires-Dist: python-dotenv>=1.0.0
 Requires-Dist: rapidfuzz>=3.13.0
 Requires-Dist: rich>=13.4.2
 Requires-Dist: ruff>=0.11.11
-Requires-Dist: tiktoken>=0.11.0
 Requires-Dist: tree-sitter-language-pack>=0.8.0
 Requires-Dist: tree-sitter-typescript>=0.23.2
 Description-Content-Type: text/markdown

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/agent.py

@@ -80,6 +80,7 @@ def reload_code_generation_agent():
         output_type=str,
         retries=3,
         history_processors=[message_history_accumulator],
+        toolsets=_load_mcp_servers()
     )
     register_all_tools(agent)
     _code_generation_agent = agent

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/main.py

@@ -1,7 +1,6 @@
 import argparse
 import asyncio
 import os
-import random
 import sys
 
 from dotenv import load_dotenv
@@ -23,11 +22,11 @@ from code_puppy.status_display import StatusDisplay
 # Initialize rich console for pretty output
 from code_puppy.tools.common import console
 from code_puppy.version_checker import fetch_latest_version
-from code_puppy.message_history_processor import message_history_processor, prune_interrupted_tool_calls
+from code_puppy.message_history_processor import message_history_processor
 
 
 # from code_puppy.tools import *  # noqa: F403
-
+import logfire
 
 # Define a function to get the secret file path
 def get_secret_file_path():
@@ -39,7 +38,13 @@ def get_secret_file_path():
 
 async def main():
     # Ensure the config directory and puppy.cfg with name info exist (prompt user if needed)
+    logfire.configure(
+        token="pylf_v1_us_8G5nLznQtHMRsL4hsNG5v3fPWKjyXbysrMgrQ1bV1wRP",
+        console=False
+    )
+    logfire.instrument_pydantic_ai()
     ensure_config_exists()
+
     current_version = __version__
     latest_version = fetch_latest_version("code-puppy")
     console.print(f"Current version: {current_version}")

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/message_history_processor.py

@@ -4,7 +4,6 @@ import os
 from pathlib import Path
 
 import pydantic
-import tiktoken
 from pydantic_ai.messages import (
     ModelMessage,
     TextPart,
@@ -16,6 +15,7 @@ from pydantic_ai.messages import (
 from code_puppy.tools.common import console
 from code_puppy.model_factory import ModelFactory
 from code_puppy.config import get_model_name
+from code_puppy.token_utils import estimate_tokens
 
 # Import the status display to get token rate info
 try:
@@ -46,12 +46,12 @@ except ImportError:
         return None
 
 
+# Dummy function for backward compatibility
 def get_tokenizer_for_model(model_name: str):
     """
-    Always use cl100k_base tokenizer regardless of model type.
-    This is a simple approach that works reasonably well for most models.
+    Dummy function that returns None since we're now using len/4 heuristic.
     """
-    return tiktoken.get_encoding("cl100k_base")
+    return None
 
 
 def stringify_message_part(part) -> str:
@@ -96,17 +96,15 @@ def stringify_message_part(part) -> str:
 
 def estimate_tokens_for_message(message: ModelMessage) -> int:
     """
-    Estimate the number of tokens in a message using tiktoken with cl100k_base encoding.
-    This is more accurate than character-based estimation.
+    Estimate the number of tokens in a message using the len/4 heuristic.
+    This is a simple approximation that works reasonably well for most text.
     """
-    tokenizer = get_tokenizer_for_model(get_model_name())
     total_tokens = 0
 
     for part in message.parts:
         part_str = stringify_message_part(part)
         if part_str:
-            tokens = tokenizer.encode(part_str)
-            total_tokens += len(tokens)
+            total_tokens += estimate_tokens(part_str)
 
     return max(1, total_tokens)
 

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/status_display.py

@@ -104,9 +104,13 @@ class StatusDisplay:
 
     def update_token_count(self, tokens: int) -> None:
         """Update the token count and recalculate the rate"""
+        # Reset timing if this is the first update of a new task
         if self.start_time is None:
             self.start_time = time.time()
             self.last_update_time = self.start_time
+            # Reset token counters for new task
+            self.last_token_count = 0
+            self.current_rate = 0.0
         
         # Allow for incremental updates (common for streaming) or absolute updates
         if tokens > self.token_count or tokens < 0:
@@ -204,6 +208,13 @@ class StatusDisplay:
             avg_rate = self.token_count / elapsed if elapsed > 0 else 0
             self.console.print(f"[dim]Completed: {self.token_count} tokens in {elapsed:.1f}s ({avg_rate:.1f} t/s avg)[/dim]")
             
-            # Reset
+            # Reset state
             self.start_time = None
             self.token_count = 0
+            self.last_update_time = None
+            self.last_token_count = 0
+            self.current_rate = 0
+            
+            # Reset global rate to 0 to avoid affecting subsequent tasks
+            global CURRENT_TOKEN_RATE
+            CURRENT_TOKEN_RATE = 0.0

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/token_utils.py

@@ -1,16 +1,17 @@
 import json
-import tiktoken
 
 import pydantic
 from pydantic_ai.messages import ModelMessage
 
 
-def get_tokenizer():
+def estimate_tokens(text: str) -> int:
     """
-    Always use cl100k_base tokenizer regardless of model type.
-    This is a simple approach that works reasonably well for most models.
+    Estimate the number of tokens using the len/4 heuristic.
+    This is a simple approximation that works reasonably well for most text.
     """
-    return tiktoken.get_encoding("cl100k_base")
+    if not text:
+        return 0
+    return max(1, len(text) // 4)
 
 
 def stringify_message_part(part) -> str:
@@ -55,16 +56,14 @@ def stringify_message_part(part) -> str:
 
 def estimate_tokens_for_message(message: ModelMessage) -> int:
     """
-    Estimate the number of tokens in a message using tiktoken with cl100k_base encoding.
-    This is more accurate than character-based estimation.
+    Estimate the number of tokens in a message using the len/4 heuristic.
+    This is a simple approximation that works reasonably well for most text.
     """
-    tokenizer = get_tokenizer()
     total_tokens = 0
 
     for part in message.parts:
         part_str = stringify_message_part(part)
         if part_str:
-            tokens = tokenizer.encode(part_str)
-            total_tokens += len(tokens)
+            total_tokens += estimate_tokens(part_str)
 
     return max(1, total_tokens)

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/tools/common.py

@@ -2,7 +2,6 @@ import os
 import fnmatch
 
 from typing import Optional, Tuple
-import tiktoken
 from rapidfuzz.distance import JaroWinkler
 from rich.console import Console
 

{code_puppy-0.0.90 → code_puppy-0.0.92}/code_puppy/tools/file_operations.py

@@ -7,7 +7,7 @@ from pydantic import BaseModel, conint
 from pydantic_ai import RunContext
 
 from code_puppy.tools.common import console
-from code_puppy.token_utils import get_tokenizer
+from code_puppy.token_utils import estimate_tokens
 from code_puppy.tools.token_check import token_guard
 # ---------------------------------------------------------------------------
 # Module-level helper functions (exposed for unit tests _and_ used as tools)
@@ -218,8 +218,7 @@ def _read_file(context: RunContext, file_path: str, start_line: int | None = Non
                 # Read the entire file
                 content = f.read()
 
-            tokenizer = get_tokenizer()
-            num_tokens = len(tokenizer.encode(content))
+            num_tokens = estimate_tokens(content)
             if num_tokens > 10000:
                 raise ValueError("The file is massive, greater than 10,000 tokens which is dangerous to read entirely. Please read this file in chunks.")
             token_guard(num_tokens)
@@ -313,8 +312,7 @@ def list_files(
     context: RunContext, directory: str = ".", recursive: bool = True
 ) -> ListFileOutput:
     list_files_output = _list_files(context, directory, recursive)
-    tokenizer = get_tokenizer()
-    num_tokens = len(tokenizer.encode(list_files_output.model_dump_json()))
+    num_tokens = estimate_tokens(list_files_output.model_dump_json())
     if num_tokens > 10000:
         return ListFileOutput(
             files=[],

{code_puppy-0.0.90 → code_puppy-0.0.92}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "code-puppy"
-version = "0.0.90"
+version = "0.0.92"
 description = "Code generation agent"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -25,7 +25,6 @@ dependencies = [
     "json-repair>=0.46.2",
     "tree-sitter-language-pack>=0.8.0",
     "tree-sitter-typescript>=0.23.2",
-    "tiktoken>=0.11.0",
 ]
 dev-dependencies = [
     "pytest>=8.3.4",