code-puppy 0.0.165__tar.gz → 0.0.167__tar.gz

{code_puppy-0.0.165 → code_puppy-0.0.167}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-puppy
-Version: 0.0.165
+Version: 0.0.167
 Summary: Code generation agent
 Project-URL: repository, https://github.com/mpfaffenberger/code_puppy
 Project-URL: HomePage, https://github.com/mpfaffenberger/code_puppy
@@ -33,6 +33,7 @@ Requires-Dist: rapidfuzz>=3.13.0
 Requires-Dist: rich>=13.4.2
 Requires-Dist: ripgrep>=14.1.0
 Requires-Dist: ruff>=0.11.11
+Requires-Dist: tenacity>=8.2.0
 Requires-Dist: termcolor>=3.1.0
 Requires-Dist: textual-dev>=1.7.0
 Requires-Dist: textual>=5.0.0

code_puppy-0.0.167/code_puppy/http_utils.py

@@ -0,0 +1,225 @@
+"""
+HTTP utilities module for code-puppy.
+
+This module provides functions for creating properly configured HTTP clients.
+"""
+
+import os
+import socket
+from typing import Dict, Optional, Union
+
+import httpx
+import requests
+from tenacity import retry_if_exception_type, stop_after_attempt, wait_exponential
+
+try:
+    from pydantic_ai.retries import (
+        AsyncTenacityTransport,
+        RetryConfig,
+        TenacityTransport,
+        wait_retry_after,
+    )
+except ImportError:
+    # Fallback if pydantic_ai.retries is not available
+    AsyncTenacityTransport = None
+    RetryConfig = None
+    TenacityTransport = None
+    wait_retry_after = None
+
+try:
+    from .reopenable_async_client import ReopenableAsyncClient
+except ImportError:
+    ReopenableAsyncClient = None
+
+try:
+    from .messaging import emit_info
+except ImportError:
+    # Fallback if messaging system is not available
+    def emit_info(content: str, **metadata):
+        pass  # No-op if messaging system is not available
+
+
+def get_cert_bundle_path() -> str:
+    # First check if SSL_CERT_FILE environment variable is set
+    ssl_cert_file = os.environ.get("SSL_CERT_FILE")
+    if ssl_cert_file and os.path.exists(ssl_cert_file):
+        return ssl_cert_file
+
+
+def create_client(
+    timeout: int = 180,
+    verify: Union[bool, str] = None,
+    headers: Optional[Dict[str, str]] = None,
+    retry_status_codes: tuple = (429, 502, 503, 504),
+) -> httpx.Client:
+    if verify is None:
+        verify = get_cert_bundle_path()
+
+    # If retry components are available, create a client with retry transport
+    if TenacityTransport and RetryConfig and wait_retry_after:
+        def should_retry_status(response):
+            """Raise exceptions for retryable HTTP status codes."""
+            if response.status_code in retry_status_codes:
+                emit_info(f"HTTP retry: Retrying request due to status code {response.status_code}")
+                response.raise_for_status()
+
+        transport = TenacityTransport(
+            config=RetryConfig(
+                retry=lambda e: isinstance(e, httpx.HTTPStatusError) and e.response.status_code in retry_status_codes,
+                wait=wait_retry_after(
+                    fallback_strategy=wait_exponential(multiplier=1, max=60),
+                    max_wait=300
+                ),
+                stop=stop_after_attempt(5),
+                reraise=True
+            ),
+            validate_response=should_retry_status
+        )
+        
+        return httpx.Client(transport=transport, verify=verify, headers=headers or {}, timeout=timeout)
+    else:
+        # Fallback to regular client if retry components are not available
+        return httpx.Client(verify=verify, headers=headers or {}, timeout=timeout)
+
+
+def create_async_client(
+    timeout: int = 180,
+    verify: Union[bool, str] = None,
+    headers: Optional[Dict[str, str]] = None,
+    retry_status_codes: tuple = (429, 502, 503, 504),
+) -> httpx.AsyncClient:
+    if verify is None:
+        verify = get_cert_bundle_path()
+
+    # If retry components are available, create a client with retry transport
+    if AsyncTenacityTransport and RetryConfig and wait_retry_after:
+        def should_retry_status(response):
+            """Raise exceptions for retryable HTTP status codes."""
+            if response.status_code in retry_status_codes:
+                emit_info(f"HTTP retry: Retrying request due to status code {response.status_code}")
+                response.raise_for_status()
+
+        transport = AsyncTenacityTransport(
+            config=RetryConfig(
+                retry=lambda e: isinstance(e, httpx.HTTPStatusError) and e.response.status_code in retry_status_codes,
+                wait=wait_retry_after(
+                    fallback_strategy=wait_exponential(multiplier=1, max=60),
+                    max_wait=300
+                ),
+                stop=stop_after_attempt(5),
+                reraise=True
+            ),
+            validate_response=should_retry_status
+        )
+        
+        return httpx.AsyncClient(transport=transport, verify=verify, headers=headers or {}, timeout=timeout)
+    else:
+        # Fallback to regular client if retry components are not available
+        return httpx.AsyncClient(verify=verify, headers=headers or {}, timeout=timeout)
+
+
+def create_requests_session(
+    timeout: float = 5.0,
+    verify: Union[bool, str] = None,
+    headers: Optional[Dict[str, str]] = None,
+) -> requests.Session:
+    session = requests.Session()
+
+    if verify is None:
+        verify = get_cert_bundle_path()
+
+    session.verify = verify
+
+    if headers:
+        session.headers.update(headers or {})
+
+    return session
+
+
+def create_auth_headers(
+    api_key: str, header_name: str = "Authorization"
+) -> Dict[str, str]:
+    return {header_name: f"Bearer {api_key}"}
+
+
+def resolve_env_var_in_header(headers: Dict[str, str]) -> Dict[str, str]:
+    resolved_headers = {}
+
+    for key, value in headers.items():
+        if isinstance(value, str):
+            try:
+                expanded = os.path.expandvars(value)
+                resolved_headers[key] = expanded
+            except Exception:
+                resolved_headers[key] = value
+        else:
+            resolved_headers[key] = value
+
+    return resolved_headers
+
+
+def create_reopenable_async_client(
+    timeout: int = 180,
+    verify: Union[bool, str] = None,
+    headers: Optional[Dict[str, str]] = None,
+    retry_status_codes: tuple = (429, 502, 503, 504),
+) -> Union[ReopenableAsyncClient, httpx.AsyncClient]:
+    if verify is None:
+        verify = get_cert_bundle_path()
+
+    # If retry components are available, create a client with retry transport
+    if AsyncTenacityTransport and RetryConfig and wait_retry_after:
+        def should_retry_status(response):
+            """Raise exceptions for retryable HTTP status codes."""
+            if response.status_code in retry_status_codes:
+                emit_info(f"HTTP retry: Retrying request due to status code {response.status_code}")
+                response.raise_for_status()
+
+        transport = AsyncTenacityTransport(
+            config=RetryConfig(
+                retry=lambda e: isinstance(e, httpx.HTTPStatusError) and e.response.status_code in retry_status_codes,
+                wait=wait_retry_after(
+                    fallback_strategy=wait_exponential(multiplier=1, max=60),
+                    max_wait=300
+                ),
+                stop=stop_after_attempt(5),
+                reraise=True
+            ),
+            validate_response=should_retry_status
+        )
+        
+        if ReopenableAsyncClient is not None:
+            return ReopenableAsyncClient(
+                transport=transport, verify=verify, headers=headers or {}, timeout=timeout
+            )
+        else:
+            # Fallback to regular AsyncClient if ReopenableAsyncClient is not available
+            return httpx.AsyncClient(transport=transport, verify=verify, headers=headers or {}, timeout=timeout)
+    else:
+        # Fallback to regular clients if retry components are not available
+        if ReopenableAsyncClient is not None:
+            return ReopenableAsyncClient(
+                verify=verify, headers=headers or {}, timeout=timeout
+            )
+        else:
+            # Fallback to regular AsyncClient if ReopenableAsyncClient is not available
+            return httpx.AsyncClient(verify=verify, headers=headers or {}, timeout=timeout)
+
+
+def is_cert_bundle_available() -> bool:
+    cert_path = get_cert_bundle_path()
+    return os.path.exists(cert_path) and os.path.isfile(cert_path)
+
+
+def find_available_port(start_port=8090, end_port=9010, host="127.0.0.1"):
+    for port in range(start_port, end_port + 1):
+        try:
+            # Try to bind to the port to check if it's available
+            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as sock:
+                sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+                sock.bind((host, port))
+                return port
+        except OSError:
+            # Port is in use, try the next one
+            continue
+    return None

{code_puppy-0.0.165 → code_puppy-0.0.167}/code_puppy/mcp/managed_server.py

@@ -226,11 +226,9 @@ class ManagedMCPServer:
                     http_kwargs["timeout"] = config["timeout"]
                 if "read_timeout" in config:
                     http_kwargs["read_timeout"] = config["read_timeout"]
-                if "http_client" in config:
-                    http_kwargs["http_client"] = config["http_client"]
-                elif config.get("headers"):
+                if "headers" in config:
+                    http_kwargs["headers"] = config.get("headers")
                     # Create HTTP client if headers are provided but no client specified
-                    http_kwargs["http_client"] = self._get_http_client()
 
                 self._pydantic_server = MCPServerStreamableHTTP(
                     **http_kwargs, process_tool_call=process_tool_call

{code_puppy-0.0.165 → code_puppy-0.0.167}/code_puppy/mcp/server_registry_catalog.py

@@ -791,18 +791,17 @@ MCP_SERVER_REGISTRY: List[MCPServerTemplate] = [
         description="Search and retrieve documentation from multiple sources with AI-powered context understanding",
         category="Documentation",
         tags=["documentation", "search", "context", "ai", "knowledge", "docs", "cloud"],
-        type="stdio",
+        type="http",
         config={
-            "timeout": 30,
-            "command": "npx",
-            "args": ["-y", "@upstash/context7-mcp", "--api-key", "$CONTEXT7_API_KEY"],
+            "url": "https://mcp.context7.com/mcp",
+             "headers": {
+                "Authorization": "Bearer $CONTEXT7_API_KEY"
+             }
         },
         verified=True,
         popular=True,
         requires=MCPServerRequirements(
             environment_vars=["CONTEXT7_API_KEY"],
-            required_tools=["node", "npx"],
-            package_dependencies=["@upstash/context7-mcp"],
         ),
         example_usage="Cloud-based service - no local setup required",
     ),

{code_puppy-0.0.165 → code_puppy-0.0.167}/code_puppy/message_history_processor.py

@@ -1,15 +1,15 @@
 import json
 import queue
-from typing import Any, Dict, List, Set, Tuple
+from typing import Any, List, Set, Tuple
 
 import pydantic
 from pydantic_ai.messages import ModelMessage, ModelRequest, TextPart, ToolCallPart
 
 from code_puppy.config import (
-    get_compaction_strategy,
-    get_compaction_threshold,
     get_model_name,
     get_protected_token_count,
+    get_compaction_threshold,
+    get_compaction_strategy,
 )
 from code_puppy.messaging import emit_error, emit_info, emit_warning
 from code_puppy.model_factory import ModelFactory
@@ -82,9 +82,7 @@ def estimate_tokens_for_message(message: ModelMessage) -> int:
 
 
 def filter_huge_messages(messages: List[ModelMessage]) -> List[ModelMessage]:
-    # First deduplicate tool returns to clean up any duplicates
-    deduplicated = deduplicate_tool_returns(messages)
-    filtered = [m for m in deduplicated if estimate_tokens_for_message(m) < 50000]
+    filtered = [m for m in messages if estimate_tokens_for_message(m) < 50000]
     pruned = prune_interrupted_tool_calls(filtered)
     return pruned
 
@@ -150,6 +148,81 @@ def split_messages_for_protected_summarization(
     return messages_to_summarize, protected_messages
 
 
+def deduplicate_tool_returns(messages: List[ModelMessage]) -> List[ModelMessage]:
+    """
+    Remove duplicate tool returns while preserving the first occurrence for each tool_call_id.
+
+    This function identifies tool-return parts that share the same tool_call_id and
+    removes duplicates, keeping only the first return for each id. This prevents
+    conversation corruption from duplicate tool_result blocks.
+    """
+    if not messages:
+        return messages
+
+    seen_tool_returns: Set[str] = set()
+    deduplicated: List[ModelMessage] = []
+    removed_count = 0
+
+    for msg in messages:
+        # Check if this message has any parts we need to filter
+        if not hasattr(msg, "parts") or not msg.parts:
+            deduplicated.append(msg)
+            continue
+
+        # Filter parts within this message
+        filtered_parts = []
+        msg_had_duplicates = False
+
+        for part in msg.parts:
+            tool_call_id = getattr(part, "tool_call_id", None)
+            part_kind = getattr(part, "part_kind", None)
+
+            # Check if this is a tool-return part
+            if tool_call_id and part_kind in {
+                "tool-return",
+                "tool-result",
+                "tool_result",
+            }:
+                if tool_call_id in seen_tool_returns:
+                    # This is a duplicate return, skip it
+                    msg_had_duplicates = True
+                    removed_count += 1
+                    continue
+                else:
+                    # First occurrence of this return, keep it
+                    seen_tool_returns.add(tool_call_id)
+                    filtered_parts.append(part)
+            else:
+                # Not a tool return, always keep
+                filtered_parts.append(part)
+
+        # If we filtered out parts, create a new message with filtered parts
+        if msg_had_duplicates and filtered_parts:
+            # Create a new message with the same attributes but filtered parts
+            new_msg = type(msg)(parts=filtered_parts)
+            # Copy over other attributes if they exist
+            for attr_name in dir(msg):
+                if (
+                    not attr_name.startswith("_")
+                    and attr_name != "parts"
+                    and hasattr(msg, attr_name)
+                ):
+                    try:
+                        setattr(new_msg, attr_name, getattr(msg, attr_name))
+                    except (AttributeError, TypeError):
+                        # Skip attributes that can't be set
+                        pass
+            deduplicated.append(new_msg)
+        elif filtered_parts:  # No duplicates but has parts
+            deduplicated.append(msg)
+        # If no parts remain after filtering, drop the entire message
+
+    if removed_count > 0:
+        emit_warning(f"Removed {removed_count} duplicate tool-return part(s)")
+
+    return deduplicated
+
+
 def summarize_messages(
     messages: List[ModelMessage], with_protection=True
 ) -> Tuple[List[ModelMessage], List[ModelMessage]]:
@@ -236,100 +309,21 @@ def get_model_context_length() -> int:
     return int(context_length)
 
 
-def deduplicate_tool_returns(messages: List[ModelMessage]) -> List[ModelMessage]:
-    """
-    Remove duplicate tool returns while preserving the first occurrence for each tool_call_id.
-
-    This function identifies tool-return parts that share the same tool_call_id and
-    removes duplicates, keeping only the first return for each id. This prevents
-    conversation corruption from duplicate tool_result blocks.
-    """
-    if not messages:
-        return messages
-
-    seen_tool_returns: Set[str] = set()
-    deduplicated: List[ModelMessage] = []
-    removed_count = 0
-
-    for msg in messages:
-        # Check if this message has any parts we need to filter
-        if not hasattr(msg, "parts") or not msg.parts:
-            deduplicated.append(msg)
-            continue
-
-        # Filter parts within this message
-        filtered_parts = []
-        msg_had_duplicates = False
-
-        for part in msg.parts:
-            tool_call_id = getattr(part, "tool_call_id", None)
-            part_kind = getattr(part, "part_kind", None)
-
-            # Check if this is a tool-return part
-            if tool_call_id and part_kind in {
-                "tool-return",
-                "tool-result",
-                "tool_result",
-            }:
-                if tool_call_id in seen_tool_returns:
-                    # This is a duplicate return, skip it
-                    msg_had_duplicates = True
-                    removed_count += 1
-                    continue
-                else:
-                    # First occurrence of this return, keep it
-                    seen_tool_returns.add(tool_call_id)
-                    filtered_parts.append(part)
-            else:
-                # Not a tool return, always keep
-                filtered_parts.append(part)
-
-        # If we filtered out parts, create a new message with filtered parts
-        if msg_had_duplicates and filtered_parts:
-            # Create a new message with the same attributes but filtered parts
-            new_msg = type(msg)(parts=filtered_parts)
-            # Copy over other attributes if they exist
-            for attr_name in dir(msg):
-                if (
-                    not attr_name.startswith("_")
-                    and attr_name != "parts"
-                    and hasattr(msg, attr_name)
-                ):
-                    try:
-                        setattr(new_msg, attr_name, getattr(msg, attr_name))
-                    except (AttributeError, TypeError):
-                        # Skip attributes that can't be set
-                        pass
-            deduplicated.append(new_msg)
-        elif filtered_parts:  # No duplicates but has parts
-            deduplicated.append(msg)
-        # If no parts remain after filtering, drop the entire message
-
-    if removed_count > 0:
-        emit_warning(f"Removed {removed_count} duplicate tool-return part(s)")
-
-    return deduplicated
-
-
 def prune_interrupted_tool_calls(messages: List[ModelMessage]) -> List[ModelMessage]:
     """
     Remove any messages that participate in mismatched tool call sequences.
 
     A mismatched tool call id is one that appears in a ToolCall (model/tool request)
-    without a corresponding tool return, or vice versa. We enforce a strict 1:1 ratio
-    between tool calls and tool returns. We preserve original order and only drop
-    messages that contain parts referencing mismatched tool_call_ids.
+    without a corresponding tool return, or vice versa. We preserve original order
+    and only drop messages that contain parts referencing mismatched tool_call_ids.
     """
     if not messages:
         return messages
 
-    # First deduplicate tool returns to clean up any duplicate returns
-    messages = deduplicate_tool_returns(messages)
+    tool_call_ids: Set[str] = set()
+    tool_return_ids: Set[str] = set()
 
-    tool_call_counts: Dict[str, int] = {}
-    tool_return_counts: Dict[str, int] = {}
-
-    # First pass: count occurrences of each tool_call_id for calls vs returns
+    # First pass: collect ids for calls vs returns
     for msg in messages:
         for part in getattr(msg, "parts", []) or []:
             tool_call_id = getattr(part, "tool_call_id", None)
@@ -338,25 +332,11 @@ def prune_interrupted_tool_calls(messages: List[ModelMessage]) -> List[ModelMess
             # Heuristic: if it's an explicit ToolCallPart or has a tool_name/args,
             # consider it a call; otherwise it's a return/result.
             if part.part_kind == "tool-call":
-                tool_call_counts[tool_call_id] = (
-                    tool_call_counts.get(tool_call_id, 0) + 1
-                )
+                tool_call_ids.add(tool_call_id)
             else:
-                tool_return_counts[tool_call_id] = (
-                    tool_return_counts.get(tool_call_id, 0) + 1
-                )
-
-    # Find mismatched tool_call_ids (not exactly 1:1 ratio)
-    all_tool_ids = set(tool_call_counts.keys()) | set(tool_return_counts.keys())
-    mismatched: Set[str] = set()
-
-    for tool_id in all_tool_ids:
-        call_count = tool_call_counts.get(tool_id, 0)
-        return_count = tool_return_counts.get(tool_id, 0)
-        # Enforce strict 1:1 ratio - both must be exactly 1
-        if call_count != 1 or return_count != 1:
-            mismatched.add(tool_id)
+                tool_return_ids.add(tool_call_id)
 
+    mismatched: Set[str] = tool_call_ids.symmetric_difference(tool_return_ids)
     if not mismatched:
         return messages
 
@@ -382,10 +362,7 @@ def prune_interrupted_tool_calls(messages: List[ModelMessage]) -> List[ModelMess
 
 
 def message_history_processor(messages: List[ModelMessage]) -> List[ModelMessage]:
-    # First, deduplicate tool returns to clean up any duplicates
-    messages = deduplicate_tool_returns(messages)
-
-    # Then, prune any interrupted/mismatched tool-call conversations
+    # First, prune any interrupted/mismatched tool-call conversations
     total_current_tokens = sum(estimate_tokens_for_message(msg) for msg in messages)
 
     model_max = get_model_context_length()
@@ -477,8 +454,6 @@ def truncation(
     messages: List[ModelMessage], protected_tokens: int
 ) -> List[ModelMessage]:
     emit_info("Truncating message history to manage token usage")
-    # First deduplicate tool returns to clean up any duplicates
-    messages = deduplicate_tool_returns(messages)
     result = [messages[0]]  # Always keep the first message (system prompt)
     num_tokens = 0
     stack = queue.LifoQueue()
@@ -501,10 +476,6 @@ def truncation(
 
 def message_history_accumulator(messages: List[Any]):
     _message_history = get_message_history()
-
-    # Deduplicate tool returns in current history before processing new messages
-    _message_history = deduplicate_tool_returns(_message_history)
-
     message_history_hashes = set([hash_message(m) for m in _message_history])
     for msg in messages:
         if (
@@ -513,12 +484,6 @@ def message_history_accumulator(messages: List[Any]):
         ):
             _message_history.append(msg)
 
-    # Deduplicate tool returns again after adding new messages to ensure no duplicates
-    _message_history = deduplicate_tool_returns(_message_history)
-
-    # Update the message history with deduplicated messages
-    set_message_history(_message_history)
-
     # Apply message history trimming using the main processor
     # This ensures we maintain global state while still managing context limits
     message_history_processor(_message_history)