patchpal 0.22.2__tar.gz → 0.22.4__tar.gz

{patchpal-0.22.2/patchpal.egg-info → patchpal-0.22.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: patchpal
-Version: 0.22.2
+Version: 0.22.4
 Summary: An agentic coding and automation assistant, supporting both local and cloud LLMs
 Author: PatchPal Contributors
 License-Expression: Apache-2.0
@@ -30,7 +30,7 @@ Requires-Dist: boto3
 Requires-Dist: pymupdf>=1.23.0
 Requires-Dist: python-docx>=1.0.0
 Requires-Dist: python-pptx>=0.6.0
-Requires-Dist: tree-sitter-language-pack>=0.3.0
+Requires-Dist: tree-sitter-language-pack<1.0.0,>=0.3.0
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-cov>=4.0.0; extra == "dev"

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal/__init__.py

@@ -1,6 +1,6 @@
 """PatchPal - An open-source Claude Code clone implemented purely in Python."""
 
-__version__ = "0.22.2"
+__version__ = "0.22.4"
 
 from patchpal.agent import create_agent, create_react_agent
 from patchpal.cli.autopilot import autopilot_loop

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal/agent/function_calling.py

@@ -479,6 +479,10 @@ class PatchPalAgent:
         self.cumulative_input_tokens = 0
         self.cumulative_output_tokens = 0
 
+        # Track last prompt tokens from most recent API response (includes cache operations)
+        # This is the ACTUAL token count sent to the API, used for accurate context management
+        self.last_prompt_tokens = None
+
         # Track cache-related tokens (for Anthropic/Bedrock models with prompt caching)
         self.cumulative_cache_creation_tokens = 0
         self.cumulative_cache_read_tokens = 0
@@ -953,6 +957,7 @@ It's currently empty (just the template). The file is automatically loaded at se
 
         # Check for compaction BEFORE starting work
         # This ensures we never compact mid-execution and lose tool results
+        # Always estimates current messages to avoid staleness issues (no actual_prompt_tokens)
         if self.enable_auto_compact and self.context_manager.needs_compaction(self.messages):
             self._perform_auto_compaction()
 
@@ -1100,7 +1105,10 @@ It's currently empty (just the template). The file is automatically loaded at se
                 last_prompt_tokens = None  # Track for reactive context management
                 if hasattr(response, "usage") and response.usage:
                     if hasattr(response.usage, "prompt_tokens"):
+                        # LiteLLM already includes cache operations in prompt_tokens for Anthropic/Bedrock
+                        # (see litellm/llms/anthropic/chat/transformation.py)
                         last_prompt_tokens = response.usage.prompt_tokens
+                        self.last_prompt_tokens = last_prompt_tokens  # Store for /status command
                         self.cumulative_input_tokens += response.usage.prompt_tokens
                     if hasattr(response.usage, "completion_tokens"):
                         self.cumulative_output_tokens += response.usage.completion_tokens

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal/cli/interactive.py

@@ -670,7 +670,9 @@ Supported models: Any LiteLLM-supported model
 
             # Handle /status command - show context window usage
             if user_input.lower() in ["status", "/status"]:
-                stats = agent.context_manager.get_usage_stats(agent.messages)
+                stats = agent.context_manager.get_usage_stats(
+                    agent.messages, actual_prompt_tokens=agent.last_prompt_tokens
+                )
 
                 print("\n" + "=" * 70)
                 print("\033[1;36mContext Window Status\033[0m")

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal/cli/sandbox.py

@@ -686,9 +686,6 @@ DESCRIPTION:
     - Pre-built image with patchpal installed (fast startup)
     - Auto-mounts ~/.patchpal for custom tools, config, and memory
     - Custom tools work automatically (from ~/.patchpal/tools/ and <repo>/.patchpal/tools/)
-    - Auto-sets OLLAMA_CONTEXT_LENGTH for Ollama models:
-      * 8192 for regular models (agents)
-      * 32768 for reasoning models (gpt-oss, deepseek-r1, qwq, qwen)
 
     Recommended for autopilot mode and high-risk operations.
 

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal/context.py

@@ -321,30 +321,31 @@ Be comprehensive but concise. The goal is to continue work seamlessly without lo
     ) -> bool:
         """Check if context window needs compaction.
 
-        Supports both reactive (preferred) and proactive (fallback) approaches:
-        - Reactive: Use actual_prompt_tokens from latest API response
-        - Proactive: Estimate tokens if actual_prompt_tokens not available
-
-        The reactive approach is preferred as it uses actual token counts from the LLM,
-        avoiding the need for tiktoken or other estimation methods.
+        ALWAYS estimates current messages to avoid staleness issues when predicting
+        whether the NEXT API call will overflow. Using actual_prompt_tokens from a
+        previous call can cause false negatives when large messages are added between
+        the last API call and the compaction check.
+
+        Example of staleness bug (fixed):
+        - Previous API call: 120K tokens (60% usage)
+        - User pastes huge changelog: +90K tokens
+        - Total: 210K tokens (exceeds 200K limit)
+        - Bug: If we used actual_prompt_tokens=120K, we'd think we're at 60%
+        - Fix: Always re-estimate to see the 210K total
+
+        The actual_prompt_tokens parameter is kept for API compatibility but ignored
+        for compaction decisions. Use get_usage_stats() for display purposes where
+        actual tokens are appropriate (staleness OK for showing recent stats).
 
         Args:
             messages: Current message history
-            actual_prompt_tokens: Optional actual prompt token count from latest API response
+            actual_prompt_tokens: IGNORED - kept for API compatibility only
 
         Returns:
             True if compaction is needed
         """
-        # Reactive approach (preferred): use actual token counts from API response
-        if actual_prompt_tokens is not None:
-            # Add output reserve to account for response tokens
-            total_tokens = actual_prompt_tokens + self.output_reserve
-            usage_ratio = total_tokens / self.context_limit
-            return usage_ratio >= self.COMPACT_THRESHOLD
-
-        # Proactive approach (fallback): estimate tokens when API data not available
-        # This uses character-based estimation (3 chars per token) which works
-        # reliably without requiring tiktoken or network access
+        # ALWAYS estimate current messages - never use stale actual_prompt_tokens
+        # This ensures we detect large message additions that happen between API calls
         # Note: Dynamic date/time message adds ~30 tokens on each LLM call
         system_tokens = self.estimator.estimate_tokens(self.system_prompt)
         datetime_tokens = 30  # Approximate size of dynamic date/time message
@@ -355,15 +356,37 @@ Be comprehensive but concise. The goal is to continue work seamlessly without lo
         usage_ratio = total_tokens / self.context_limit
         return usage_ratio >= self.COMPACT_THRESHOLD
 
-    def get_usage_stats(self, messages: List[Dict[str, Any]]) -> Dict[str, Any]:
+    def get_usage_stats(
+        self, messages: List[Dict[str, Any]], actual_prompt_tokens: int = None
+    ) -> Dict[str, Any]:
         """Get current context usage statistics.
 
         Args:
             messages: Current message history
+            actual_prompt_tokens: Optional actual prompt tokens from latest API response (includes cache operations)
 
         Returns:
             Dict with usage statistics
         """
+        # If we have actual prompt tokens from API (includes cache writes/reads), use those
+        if actual_prompt_tokens is not None:
+            total_tokens = actual_prompt_tokens + self.output_reserve
+            # For display purposes, estimate system vs message breakdown
+            system_tokens = self.estimator.estimate_tokens(self.system_prompt)
+            datetime_tokens = 30
+            message_tokens = actual_prompt_tokens - system_tokens - datetime_tokens
+
+            return {
+                "system_tokens": system_tokens + datetime_tokens,
+                "message_tokens": max(0, message_tokens),  # Ensure non-negative
+                "output_reserve": self.output_reserve,
+                "total_tokens": total_tokens,
+                "context_limit": self.context_limit,
+                "usage_ratio": total_tokens / self.context_limit,
+                "usage_percent": int((total_tokens / self.context_limit) * 100),
+            }
+
+        # Fallback to estimation when actual tokens not available
         system_tokens = self.estimator.estimate_tokens(self.system_prompt)
         datetime_tokens = 30  # Approximate size of dynamic date/time message
         message_tokens = self.estimator.estimate_messages_tokens(messages)

{patchpal-0.22.2 → patchpal-0.22.4/patchpal.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: patchpal
-Version: 0.22.2
+Version: 0.22.4
 Summary: An agentic coding and automation assistant, supporting both local and cloud LLMs
 Author: PatchPal Contributors
 License-Expression: Apache-2.0
@@ -30,7 +30,7 @@ Requires-Dist: boto3
 Requires-Dist: pymupdf>=1.23.0
 Requires-Dist: python-docx>=1.0.0
 Requires-Dist: python-pptx>=0.6.0
-Requires-Dist: tree-sitter-language-pack>=0.3.0
+Requires-Dist: tree-sitter-language-pack<1.0.0,>=0.3.0
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-cov>=4.0.0; extra == "dev"

{patchpal-0.22.2 → patchpal-0.22.4}/patchpal.egg-info/requires.txt

@@ -9,7 +9,7 @@ boto3
 pymupdf>=1.23.0
 python-docx>=1.0.0
 python-pptx>=0.6.0
-tree-sitter-language-pack>=0.3.0
+tree-sitter-language-pack<1.0.0,>=0.3.0
 
 [dev]
 pytest>=7.0.0

{patchpal-0.22.2 → patchpal-0.22.4}/pyproject.toml

@@ -36,7 +36,7 @@ dependencies = [
     "pymupdf>=1.23.0",
     "python-docx>=1.0.0",
     "python-pptx>=0.6.0",
-    "tree-sitter-language-pack>=0.3.0",
+    "tree-sitter-language-pack>=0.3.0,<1.0.0",
 ]
 
 [project.optional-dependencies]