vibecore 0.2.0__py3-none-any.whl

vibecore/mcp/server_wrapper.py

@@ -0,0 +1,109 @@
+"""MCP server wrapper for renaming tools with a prefix pattern."""
+
+from typing import TYPE_CHECKING, Any
+
+from agents.mcp import MCPServer
+from agents.run_context import RunContextWrapper
+from mcp.types import CallToolResult, GetPromptResult, ListPromptsResult
+from mcp.types import Tool as MCPTool
+
+if TYPE_CHECKING:
+    from agents import AgentBase
+
+
+class NameOverridingMCPServer(MCPServer):
+    """Wrapper for MCP servers that renames tools with mcp__servername__toolname pattern."""
+
+    def __init__(self, actual_server: MCPServer, use_structured_content: bool = False):
+        """Initialize the wrapper.
+
+        Args:
+            actual_server: The actual MCP server to wrap.
+            use_structured_content: Whether to use structured content.
+        """
+        super().__init__(use_structured_content=use_structured_content)
+        self.actual_server = actual_server
+        # Store the mapping between renamed and original tool names
+        self._tool_name_mapping: dict[str, str] = {}
+
+    @property
+    def name(self) -> str:
+        """Return the name of the wrapped server."""
+        return self.actual_server.name
+
+    async def connect(self) -> None:
+        """Connect to the wrapped server."""
+        await self.actual_server.connect()
+
+    async def cleanup(self) -> None:
+        """Cleanup the wrapped server."""
+        await self.actual_server.cleanup()
+
+    async def list_tools(
+        self,
+        run_context: RunContextWrapper[Any] | None = None,
+        agent: "AgentBase | None" = None,
+    ) -> list[MCPTool]:
+        """List tools with renamed names.
+
+        Args:
+            run_context: The run context.
+            agent: The agent requesting tools.
+
+        Returns:
+            List of tools with renamed names following mcp__servername__toolname pattern.
+        """
+        # Get tools from the actual server
+        tools = await self.actual_server.list_tools(run_context, agent)
+
+        # Rename each tool
+        renamed_tools = []
+        for tool in tools:
+            # Create the new name with the pattern mcp__servername__toolname
+            original_name = tool.name
+            new_name = f"mcp__{self.name}__{original_name}"
+
+            # Store the mapping for call_tool
+            self._tool_name_mapping[new_name] = original_name
+
+            # Create a new tool with the renamed name
+            renamed_tool = MCPTool(
+                name=new_name,
+                description=tool.description,
+                inputSchema=tool.inputSchema,
+            )
+            renamed_tools.append(renamed_tool)
+
+        return renamed_tools
+
+    async def call_tool(self, tool_name: str, arguments: dict[str, Any] | None) -> CallToolResult:
+        """Call a tool using its original name.
+
+        Args:
+            tool_name: The renamed tool name (mcp__servername__toolname).
+            arguments: The tool arguments.
+
+        Returns:
+            The result from calling the tool.
+        """
+        # Map the renamed tool name back to the original
+        original_name = self._tool_name_mapping.get(tool_name)
+        if original_name is None:
+            # If not in mapping, try to extract from pattern
+            if tool_name.startswith(f"mcp__{self.name}__"):
+                # Extract original name from pattern
+                original_name = tool_name[len(f"mcp__{self.name}__") :]
+            else:
+                # Use as-is if not matching our pattern
+                original_name = tool_name
+
+        # Call the tool with the original name
+        return await self.actual_server.call_tool(original_name, arguments)
+
+    async def list_prompts(self) -> ListPromptsResult:
+        """List prompts from the wrapped server."""
+        return await self.actual_server.list_prompts()
+
+    async def get_prompt(self, name: str, arguments: dict[str, Any] | None = None) -> GetPromptResult:
+        """Get a prompt from the wrapped server."""
+        return await self.actual_server.get_prompt(name, arguments)

vibecore/models/__init__.py

@@ -0,0 +1,5 @@
+"""Models package for Vibecore application."""
+
+from .anthropic import AnthropicModel
+
+__all__ = ["AnthropicModel"]

vibecore/models/anthropic.py

@@ -0,0 +1,239 @@
+"""Anthropic model implementation with automatic cache control."""
+
+import json
+import logging
+from typing import Any, Literal, overload
+
+import litellm
+from agents.agent_output import AgentOutputSchemaBase
+from agents.extensions.models.litellm_model import LitellmModel
+from agents.handoffs import Handoff
+from agents.items import TResponseInputItem
+from agents.model_settings import ModelSettings
+from agents.models.interface import ModelTracing
+from agents.tool import Tool
+from agents.tracing.span_data import GenerationSpanData
+from agents.tracing.spans import Span
+from openai import AsyncStream
+from openai.types.chat import ChatCompletionChunk
+from openai.types.responses import Response
+
+# Set up debug logging
+logger = logging.getLogger(__name__)
+
+
+def _transform_messages_for_cache(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    """Transform messages to add cache_control for Anthropic models.
+
+    Caches up to 4 messages in priority order:
+    1. Last message
+    2. Message before last user message (often tool result)
+    3. Message before second-to-last user message (often tool result)
+    4. Last system message
+
+    Args:
+        messages: List of message dictionaries
+
+    Returns:
+        Transformed messages with cache_control added
+    """
+    if not messages:
+        return []
+
+    indices_to_cache = set()
+
+    # 1. Always cache the last message
+    indices_to_cache.add(len(messages) - 1)
+
+    # 2. Find user messages going backwards
+    user_message_indices = []
+    for i in range(len(messages) - 1, -1, -1):
+        if messages[i].get("role") == "user":
+            user_message_indices.append(i)
+
+    # 3. Cache message before the last user message (if exists)
+    # This is often a tool result which contains important context
+    if len(user_message_indices) >= 1 and user_message_indices[0] > 0:
+        indices_to_cache.add(user_message_indices[0] - 1)
+
+    # 4. Cache message before the second-to-last user message (if exists)
+    # This is also often a tool result
+    if len(user_message_indices) >= 2 and user_message_indices[1] > 0:
+        indices_to_cache.add(user_message_indices[1] - 1)
+
+    # 5. Find and cache the last system message
+    for i in range(len(messages) - 1, -1, -1):
+        if messages[i].get("role") == "system":
+            indices_to_cache.add(i)
+            break
+
+    # Transform messages with cache_control only for selected indices
+    transformed = []
+    for i, msg in enumerate(messages):
+        new_msg = msg.copy()
+
+        if i in indices_to_cache:
+            # Add cache_control to this message
+            content = new_msg.get("content")
+
+            if isinstance(content, str):
+                # Only add cache_control if text is not empty
+                if content:
+                    # Convert string content to list format with cache_control
+                    new_msg["content"] = [{"type": "text", "text": content, "cache_control": {"type": "ephemeral"}}]
+                # else: keep empty string as is, don't convert to list format
+            elif isinstance(content, list):
+                # Add cache_control to first text item if not already present
+                new_content = []
+                cache_added = False
+
+                for item in content:
+                    if isinstance(item, dict) and item.get("type") == "text" and not cache_added:
+                        # Only add cache_control if text is not empty
+                        text_content = item.get("text", "")
+                        if text_content and "cache_control" not in item:
+                            # Add cache_control to the first non-empty text item without cache_control
+                            new_item = item.copy()
+                            new_item["cache_control"] = {"type": "ephemeral"}
+                            new_content.append(new_item)
+                            cache_added = True
+                        elif text_content and "cache_control" in item:
+                            # Non-empty item already has cache_control
+                            new_content.append(item)
+                            cache_added = True
+                        else:
+                            # Empty text or already has cache_control - keep as is
+                            new_content.append(item)
+                    else:
+                        new_content.append(item)
+
+                new_msg["content"] = new_content
+
+        transformed.append(new_msg)
+
+    return transformed
+
+
+class AnthropicModel(LitellmModel):
+    """Anthropic model that automatically adds cache_control to messages.
+
+    This implementation minimally overrides the _fetch_response method to intercept
+    and transform messages before they're sent to the Anthropic API.
+
+    The override approach:
+    1. Temporarily replaces litellm.acompletion with an intercepting function
+    2. The interceptor transforms messages only for this specific model
+    3. Calls the parent's _fetch_response which uses the interceptor
+    4. Always restores the original function, even if an error occurs
+
+    This minimal approach ensures compatibility with upstream LitellmModel changes
+    while adding the necessary cache_control functionality for Anthropic models.
+    """
+
+    def __init__(self, model_name: str, base_url: str | None = None, api_key: str | None = None):
+        """Initialize AnthropicModel."""
+        super().__init__(model_name, base_url, api_key)
+        logger.debug(f"AnthropicModel initialized with model: {model_name}")
+
+    @overload
+    async def _fetch_response(
+        self,
+        system_instructions: str | None,
+        input: str | list[TResponseInputItem],
+        model_settings: ModelSettings,
+        tools: list[Tool],
+        output_schema: AgentOutputSchemaBase | None,
+        handoffs: list[Handoff],
+        span: Span[GenerationSpanData],
+        tracing: ModelTracing,
+        stream: Literal[True],
+        prompt: Any | None = None,
+    ) -> tuple[Response, AsyncStream[ChatCompletionChunk]]: ...
+
+    @overload
+    async def _fetch_response(
+        self,
+        system_instructions: str | None,
+        input: str | list[TResponseInputItem],
+        model_settings: ModelSettings,
+        tools: list[Tool],
+        output_schema: AgentOutputSchemaBase | None,
+        handoffs: list[Handoff],
+        span: Span[GenerationSpanData],
+        tracing: ModelTracing,
+        stream: Literal[False],
+        prompt: Any | None = None,
+    ) -> Any: ...  # litellm.ModelResponse
+
+    async def _fetch_response(
+        self,
+        system_instructions: str | None,
+        input: str | list[TResponseInputItem],
+        model_settings: ModelSettings,
+        tools: list[Tool],
+        output_schema: AgentOutputSchemaBase | None,
+        handoffs: list[Handoff],
+        span: Span[GenerationSpanData],
+        tracing: ModelTracing,
+        stream: bool = False,
+        prompt: Any | None = None,
+    ) -> Any | tuple[Response, AsyncStream[ChatCompletionChunk]]:
+        """Override _fetch_response to add cache_control to messages."""
+        # Store the original litellm.acompletion function
+        original_acompletion = litellm.acompletion
+
+        async def _intercepting_acompletion(*args, **kwargs):
+            """Intercept litellm.acompletion calls to transform messages."""
+            # Only transform messages for this Anthropic model
+            if kwargs.get("model") == self.model and "messages" in kwargs:
+                messages = kwargs["messages"]
+                logger.debug(f"Intercepting Anthropic API call for model {self.model} with {len(messages)} messages")
+
+                # Transform messages to add cache_control
+                transformed = _transform_messages_for_cache(messages)
+                kwargs["messages"] = transformed
+
+                # Log transformation for debugging
+                if logger.isEnabledFor(logging.DEBUG):
+                    for i, (orig, trans) in enumerate(zip(messages[:2], transformed[:2], strict=False)):
+                        logger.debug(f"Message {i} transformation:")
+                        logger.debug(f"  Original: {json.dumps(orig, indent=2)}")
+                        logger.debug(f"  Transformed: {json.dumps(trans, indent=2)}")
+
+            # Call the original function with potentially transformed kwargs
+            return await original_acompletion(*args, **kwargs)
+
+        try:
+            # Temporarily replace litellm.acompletion with our intercepting version
+            litellm.acompletion = _intercepting_acompletion
+
+            # Call the parent's implementation, which will use our intercepting function
+            if stream:
+                return await super()._fetch_response(
+                    system_instructions=system_instructions,
+                    input=input,
+                    model_settings=model_settings,
+                    tools=tools,
+                    output_schema=output_schema,
+                    handoffs=handoffs,
+                    span=span,
+                    tracing=tracing,
+                    stream=True,
+                    prompt=prompt,
+                )
+            else:
+                return await super()._fetch_response(
+                    system_instructions=system_instructions,
+                    input=input,
+                    model_settings=model_settings,
+                    tools=tools,
+                    output_schema=output_schema,
+                    handoffs=handoffs,
+                    span=span,
+                    tracing=tracing,
+                    stream=False,
+                    prompt=prompt,
+                )
+        finally:
+            # Always restore the original function
+            litellm.acompletion = original_acompletion

vibecore/prompts/common_system_prompt.txt

@@ -0,0 +1,64 @@
+You are an interactive CLI tool that helps users with any task they need assistance with. Use the instructions below and the tools available to you to assist the user.
+
+# Tone and style
+You should be concise, direct, and to the point. When you run a non-trivial bash command, you should explain what the command does and why you are running it, to make sure the user understands what you are doing (this is especially important when you are running a command that will make changes to the user's system).
+Remember that your output will be displayed on a command line interface. Your responses can use Github-flavored markdown for formatting, and will be rendered in a monospace font using the CommonMark specification.
+Output text to communicate with the user; all text you output outside of tool use is displayed to the user. Only use tools to complete tasks. Never use tools like Bash or code comments as means to communicate with the user during the session.
+If you cannot or will not help the user with something, please do not say why or what it could lead to, since this comes across as preachy and annoying. Please offer helpful alternatives if possible, and otherwise keep your response to 1-2 sentences.
+Only use emojis if the user explicitly requests it. Avoid using emojis in all communication unless asked.
+IMPORTANT: You should minimize output tokens as much as possible while maintaining helpfulness, quality, and accuracy. Only address the specific query or task at hand, avoiding tangential information unless absolutely critical for completing the request. If you can answer in 1-3 sentences or a short paragraph, please do.
+IMPORTANT: You should NOT answer with unnecessary preamble or postamble (such as explaining your code or summarizing your action), unless the user asks you to.
+IMPORTANT: Keep your responses short, since they will be displayed on a command line interface. You MUST answer concisely with fewer than 4 lines (not including tool use or code generation), unless user asks for detail. Answer the user's question directly, without elaboration, explanation, or details. One word answers are best. Avoid introductions, conclusions, and explanations. You MUST avoid text before/after your response, such as "The answer is <answer>.", "Here is the content of the file..." or "Based on the information provided, the answer is..." or "Here is what I will do next...". Here are some examples to demonstrate appropriate verbosity:
+<example>
+user: 2 + 2
+assistant: 4
+</example>
+
+<example>
+user: what is 2+2?
+assistant: 4
+</example>
+
+<example>
+user: is 11 a prime number?
+assistant: Yes
+</example>
+
+<example>
+user: what command should I run to list files in the current directory?
+assistant: ls
+</example>
+
+<example>
+user: what command should I run to find all Python files?
+assistant: find . -name "*.py"
+</example>
+
+<example>
+user: How many golf balls fit inside a jetta?
+assistant: 150000
+</example>
+
+<example>
+user: what files are in the src directory?
+assistant: [runs ls tool on src/ and sees main.py, utils.py, config.py]
+user: which file contains the configuration?
+assistant: src/config.py
+</example>
+
+# Proactiveness
+You are allowed to be proactive, but only when the user asks you to do something. You should strive to strike a balance between:
+1. Doing the right thing when asked, including taking actions and follow-up actions
+2. Not surprising the user with actions you take without asking
+For example, if the user asks you how to approach something, you should do your best to answer their question first, and not immediately jump into taking actions.
+3. Do not add additional code explanation summary unless requested by the user. After working on a file, just stop, rather than providing an explanation of what you did.
+
+# Following conventions
+When making changes to files, first understand the file's code conventions. Mimic code style, use existing libraries and utilities, and follow existing patterns.
+- NEVER assume that a given library is available, even if it is well known. Whenever you write code that uses a library or framework, first check that this codebase already uses the given library. For example, you might look at neighboring files, or check the package.json (or cargo.toml, and so on depending on the language).
+- When you create a new component, first look at existing components to see how they're written; then consider framework choice, naming conventions, typing, and other conventions.
+- When you edit a piece of code, first look at the code's surrounding context (especially its imports) to understand the code's choice of frameworks and libraries. Then consider how to make the given change in a way that is most idiomatic.
+- Always follow security best practices. Never introduce code that exposes or logs secrets and keys. Never commit secrets or keys to the repository.
+
+# Code style
+- IMPORTANT: DO NOT ADD ***ANY*** COMMENTS unless asked

vibecore/session/__init__.py

@@ -0,0 +1,5 @@
+"""Session storage implementations for vibecore."""
+
+from .jsonl_session import JSONLSession
+
+__all__ = ["JSONLSession"]

vibecore/session/file_lock.py

@@ -0,0 +1,127 @@
+"""File locking utilities for thread-safe session operations."""
+
+import asyncio
+import threading
+from collections.abc import AsyncIterator
+from contextlib import asynccontextmanager
+from pathlib import Path
+
+
+class FileLockManager:
+    """Manages thread-based file locks for session files.
+
+    This is a basic implementation using threading locks.
+    In Phase 2, this will be enhanced with OS-level file locking
+    using fcntl (Unix) and msvcrt (Windows).
+    """
+
+    def __init__(self):
+        """Initialize the file lock manager."""
+        # Dictionary mapping file paths to their locks
+        self._locks: dict[str, threading.Lock] = {}
+        # Lock to protect the _locks dictionary itself
+        self._locks_lock = threading.Lock()
+
+    def _get_lock(self, file_path: Path) -> threading.Lock:
+        """Get or create a lock for the specified file path.
+
+        Args:
+            file_path: Path to the file to lock
+
+        Returns:
+            Threading lock for the file
+        """
+        path_str = str(file_path.resolve())
+
+        with self._locks_lock:
+            if path_str not in self._locks:
+                self._locks[path_str] = threading.Lock()
+            return self._locks[path_str]
+
+    @asynccontextmanager
+    async def acquire_lock(
+        self,
+        file_path: Path,
+        timeout: float = 30.0,
+        exclusive: bool = True,
+    ) -> AsyncIterator[None]:
+        """Acquire a lock for the specified file.
+
+        Args:
+            file_path: Path to the file to lock
+            timeout: Maximum time to wait for the lock (in seconds)
+            exclusive: Whether to acquire an exclusive lock (unused in thread-based impl)
+
+        Yields:
+            None when the lock is acquired
+
+        Raises:
+            TimeoutError: If the lock cannot be acquired within the timeout
+        """
+        lock = self._get_lock(file_path)
+
+        # Try to acquire the lock with timeout
+        acquired = await asyncio.to_thread(lock.acquire, timeout=timeout)
+
+        if not acquired:
+            raise TimeoutError(f"Could not acquire lock for {file_path} within {timeout} seconds")
+
+        try:
+            yield
+        finally:
+            # Always release the lock
+            lock.release()
+
+    def cleanup_lock(self, file_path: Path) -> None:
+        """Remove the lock for a file that no longer exists.
+
+        This should be called after deleting a file to avoid memory leaks.
+
+        Args:
+            file_path: Path to the file whose lock should be removed
+        """
+        path_str = str(file_path.resolve())
+
+        with self._locks_lock:
+            if path_str in self._locks:
+                del self._locks[path_str]
+
+
+# Global lock manager instance
+_lock_manager = FileLockManager()
+
+
+@asynccontextmanager
+async def acquire_file_lock(
+    file_path: Path,
+    timeout: float = 30.0,
+    exclusive: bool = True,
+) -> AsyncIterator[None]:
+    """Acquire a lock for the specified file.
+
+    This is a convenience function that uses the global lock manager.
+
+    Args:
+        file_path: Path to the file to lock
+        timeout: Maximum time to wait for the lock (in seconds)
+        exclusive: Whether to acquire an exclusive lock
+
+    Yields:
+        None when the lock is acquired
+
+    Raises:
+        TimeoutError: If the lock cannot be acquired within the timeout
+    """
+    async with _lock_manager.acquire_lock(file_path, timeout, exclusive):
+        yield
+
+
+def cleanup_file_lock(file_path: Path) -> None:
+    """Remove the lock for a file that no longer exists.
+
+    This is a convenience function that uses the global lock manager.
+
+    Args:
+        file_path: Path to the file whose lock should be removed
+    """
+    _lock_manager.cleanup_lock(file_path)