code-puppy 0.0.320__tar.gz → 0.0.338__tar.gz

{code_puppy-0.0.320 → code_puppy-0.0.338}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-puppy
-Version: 0.0.320
+Version: 0.0.338
 Summary: Code generation agent
 Project-URL: repository, https://github.com/mpfaffenberger/code_puppy
 Project-URL: HomePage, https://github.com/mpfaffenberger/code_puppy
@@ -34,6 +34,7 @@ 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: termflow-md>=0.1.6
 Requires-Dist: uvicorn>=0.30.0
 Description-Content-Type: text/markdown
 
@@ -100,66 +101,32 @@ uvx code-puppy -i
 
 ### UV (Recommended)
 
+#### macOS / Linux
+
 ```bash
 # Install UV if you don't have it
 curl -LsSf https://astral.sh/uv/install.sh | sh
 
-# Set UV to always use managed Python (one-time setup)
-echo 'export UV_MANAGED_PYTHON=1' >> ~/.zshrc  # or ~/.bashrc
-source ~/.zshrc  # or ~/.bashrc
-
-# Install and run code-puppy
-uvx code-puppy -i
-```
-
-UV will automatically download the latest compatible Python version (3.11+) if your system doesn't have one.
-
-### pip (Alternative)
-
-```bash
-pip install code-puppy
+uvx code-puppy
 ```
 
-*Note: pip installation requires your system Python to be 3.11 or newer.*
+#### Windows
 
-### Permanent Python Management
+On Windows, we recommend installing code-puppy as a global tool for the best experience with keyboard shortcuts (Ctrl+C/Ctrl+X cancellation):
 
-To make UV always use managed Python versions (recommended):
-
-```bash
-# Set environment variable permanently
-echo 'export UV_MANAGED_PYTHON=1' >> ~/.zshrc  # or ~/.bashrc
-source ~/.zshrc  # or ~/.bashrc
+```powershell
+# Install UV if you don't have it (run in PowerShell as Admin)
+powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
 
-# Now all UV commands will prefer managed Python installations
-uvx code-puppy  # No need for --managed-python flag anymore
+uvx code-puppy
 ```
 
-### Verifying Python Version
+## Changelog (By Kittylog!)
 
-```bash
-# Check which Python UV will use
-uv python find
-
-# Or check the current project's Python
-uv run python --version
-```
+[📋 View the full changelog on Kittylog](https://kittylog.app/c/mpfaffenberger/code_puppy)
 
 ## Usage
 
-### Custom Commands
-Create markdown files in `.claude/commands/`, `.github/prompts/`, or `.agents/commands/` to define custom slash commands. The filename becomes the command name and the content runs as a prompt.
-
-```bash
-# Create a custom command
-echo "# Code Review
-
-Please review this code for security issues." > .claude/commands/review.md
-
-# Use it in Code Puppy
-/review with focus on authentication
-```
-
 ### Adding Models from models.dev 🆕
 
 While there are several models configured right out of the box from providers like Synthetic, Cerebras, OpenAI, Google, and Anthropic, Code Puppy integrates with [models.dev](https://models.dev) to let you browse and add models from **65+ providers** with a single command:
@@ -227,6 +194,18 @@ The following environment variables control DBOS behavior:
 - `DBOS_SYSTEM_DATABASE_URL`: Database URL used by DBOS. Can point to a local SQLite file or a Postgres instance. Example: `postgresql://postgres:dbos@localhost:5432/postgres`. Default: `dbos_store.sqlite` file in the config directory.
 - `DBOS_APP_VERSION`: If set, Code Puppy uses it as the [DBOS application version](https://docs.dbos.dev/architecture#application-and-workflow-versions) and automatically tries to recover pending workflows for this version. Default: Code Puppy version + Unix timestamp in millisecond (disable automatic recovery).
 
+### Custom Commands
+Create markdown files in `.claude/commands/`, `.github/prompts/`, or `.agents/commands/` to define custom slash commands. The filename becomes the command name and the content runs as a prompt.
+
+```bash
+# Create a custom command
+echo "# Code Review
+
+Please review this code for security issues." > .claude/commands/review.md
+
+# Use it in Code Puppy
+/review with focus on authentication
+```
 
 ## Requirements
 
@@ -246,9 +225,6 @@ For examples and more information about agent rules, visit [https://agent.md](ht
 
 Use the `/mcp` command to manage MCP (list, start, stop, status, etc.)
 
-Watch this video for examples! https://www.youtube.com/watch?v=1t1zEetOqlo
-
-
 ## Round Robin Model Distribution
 
 Code Puppy supports **Round Robin model distribution** to help you overcome rate limits and distribute load across multiple AI models. This feature automatically cycles through configured models with each request, maximizing your API usage while staying within rate limits.

{code_puppy-0.0.320 → code_puppy-0.0.338}/README.md

@@ -61,66 +61,32 @@ uvx code-puppy -i
 
 ### UV (Recommended)
 
+#### macOS / Linux
+
 ```bash
 # Install UV if you don't have it
 curl -LsSf https://astral.sh/uv/install.sh | sh
 
-# Set UV to always use managed Python (one-time setup)
-echo 'export UV_MANAGED_PYTHON=1' >> ~/.zshrc  # or ~/.bashrc
-source ~/.zshrc  # or ~/.bashrc
-
-# Install and run code-puppy
-uvx code-puppy -i
-```
-
-UV will automatically download the latest compatible Python version (3.11+) if your system doesn't have one.
-
-### pip (Alternative)
-
-```bash
-pip install code-puppy
+uvx code-puppy
 ```
 
-*Note: pip installation requires your system Python to be 3.11 or newer.*
+#### Windows
 
-### Permanent Python Management
+On Windows, we recommend installing code-puppy as a global tool for the best experience with keyboard shortcuts (Ctrl+C/Ctrl+X cancellation):
 
-To make UV always use managed Python versions (recommended):
-
-```bash
-# Set environment variable permanently
-echo 'export UV_MANAGED_PYTHON=1' >> ~/.zshrc  # or ~/.bashrc
-source ~/.zshrc  # or ~/.bashrc
+```powershell
+# Install UV if you don't have it (run in PowerShell as Admin)
+powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
 
-# Now all UV commands will prefer managed Python installations
-uvx code-puppy  # No need for --managed-python flag anymore
+uvx code-puppy
 ```
 
-### Verifying Python Version
+## Changelog (By Kittylog!)
 
-```bash
-# Check which Python UV will use
-uv python find
-
-# Or check the current project's Python
-uv run python --version
-```
+[📋 View the full changelog on Kittylog](https://kittylog.app/c/mpfaffenberger/code_puppy)
 
 ## Usage
 
-### Custom Commands
-Create markdown files in `.claude/commands/`, `.github/prompts/`, or `.agents/commands/` to define custom slash commands. The filename becomes the command name and the content runs as a prompt.
-
-```bash
-# Create a custom command
-echo "# Code Review
-
-Please review this code for security issues." > .claude/commands/review.md
-
-# Use it in Code Puppy
-/review with focus on authentication
-```
-
 ### Adding Models from models.dev 🆕
 
 While there are several models configured right out of the box from providers like Synthetic, Cerebras, OpenAI, Google, and Anthropic, Code Puppy integrates with [models.dev](https://models.dev) to let you browse and add models from **65+ providers** with a single command:
@@ -188,6 +154,18 @@ The following environment variables control DBOS behavior:
 - `DBOS_SYSTEM_DATABASE_URL`: Database URL used by DBOS. Can point to a local SQLite file or a Postgres instance. Example: `postgresql://postgres:dbos@localhost:5432/postgres`. Default: `dbos_store.sqlite` file in the config directory.
 - `DBOS_APP_VERSION`: If set, Code Puppy uses it as the [DBOS application version](https://docs.dbos.dev/architecture#application-and-workflow-versions) and automatically tries to recover pending workflows for this version. Default: Code Puppy version + Unix timestamp in millisecond (disable automatic recovery).
 
+### Custom Commands
+Create markdown files in `.claude/commands/`, `.github/prompts/`, or `.agents/commands/` to define custom slash commands. The filename becomes the command name and the content runs as a prompt.
+
+```bash
+# Create a custom command
+echo "# Code Review
+
+Please review this code for security issues." > .claude/commands/review.md
+
+# Use it in Code Puppy
+/review with focus on authentication
+```
 
 ## Requirements
 
@@ -207,9 +185,6 @@ For examples and more information about agent rules, visit [https://agent.md](ht
 
 Use the `/mcp` command to manage MCP (list, start, stop, status, etc.)
 
-Watch this video for examples! https://www.youtube.com/watch?v=1t1zEetOqlo
-
-
 ## Round Robin Model Distribution
 
 Code Puppy supports **Round Robin model distribution** to help you overcome rate limits and distribute load across multiple AI models. This feature automatically cycles through configured models with each request, maximizing your API usage while staying within rate limits.

{code_puppy-0.0.320 → code_puppy-0.0.338}/code_puppy/agents/base_agent.py

@@ -8,7 +8,18 @@ import threading
 import uuid
 from abc import ABC, abstractmethod
 from collections.abc import AsyncIterable
-from typing import Any, Callable, Dict, List, Optional, Sequence, Set, Tuple, Union
+from typing import (
+    Any,
+    Callable,
+    Dict,
+    List,
+    Optional,
+    Sequence,
+    Set,
+    Tuple,
+    Type,
+    Union,
+)
 
 import mcp
 import pydantic
@@ -1230,6 +1241,74 @@ class BaseAgent(ABC):
             self._mcp_servers = mcp_servers
         return self._code_generation_agent
 
+    def _create_agent_with_output_type(self, output_type: Type[Any]) -> PydanticAgent:
+        """Create a temporary agent configured with a custom output_type.
+
+        This is used when structured output is requested via run_with_mcp.
+        The agent is created fresh with the same configuration as the main agent
+        but with the specified output_type instead of str.
+
+        Args:
+            output_type: The Pydantic model or type for structured output.
+
+        Returns:
+            A configured PydanticAgent (or DBOSAgent wrapper) with the custom output_type.
+        """
+        from code_puppy.model_utils import prepare_prompt_for_model
+        from code_puppy.tools import register_tools_for_agent
+
+        model_name = self.get_model_name()
+        models_config = ModelFactory.load_config()
+        model, resolved_model_name = self._load_model_with_fallback(
+            model_name, models_config, str(uuid.uuid4())
+        )
+
+        instructions = self.get_system_prompt()
+        puppy_rules = self.load_puppy_rules()
+        if puppy_rules:
+            instructions += f"\n{puppy_rules}"
+
+        mcp_servers = getattr(self, "_mcp_servers", []) or []
+        model_settings = make_model_settings(resolved_model_name)
+
+        prepared = prepare_prompt_for_model(
+            model_name, instructions, "", prepend_system_to_user=False
+        )
+        instructions = prepared.instructions
+
+        global _reload_count
+        _reload_count += 1
+
+        if get_use_dbos():
+            temp_agent = PydanticAgent(
+                model=model,
+                instructions=instructions,
+                output_type=output_type,
+                retries=3,
+                toolsets=[],
+                history_processors=[self.message_history_accumulator],
+                model_settings=model_settings,
+            )
+            agent_tools = self.get_available_tools()
+            register_tools_for_agent(temp_agent, agent_tools)
+            dbos_agent = DBOSAgent(
+                temp_agent, name=f"{self.name}-structured-{_reload_count}"
+            )
+            return dbos_agent
+        else:
+            temp_agent = PydanticAgent(
+                model=model,
+                instructions=instructions,
+                output_type=output_type,
+                retries=3,
+                toolsets=mcp_servers,
+                history_processors=[self.message_history_accumulator],
+                model_settings=model_settings,
+            )
+            agent_tools = self.get_available_tools()
+            register_tools_for_agent(temp_agent, agent_tools)
+            return temp_agent
+
     # It's okay to decorate it with DBOS.step even if not using DBOS; the decorator is a no-op in that case.
     @DBOS.step()
     def message_history_accumulator(self, ctx: RunContext, messages: List[Any]):
@@ -1260,17 +1339,20 @@ class BaseAgent(ABC):
     ) -> None:
         """Handle streaming events from the agent run.
 
-        This method processes streaming events and emits TextPart and ThinkingPart
-        content with styled banners as they stream in.
+        This method processes streaming events and emits TextPart, ThinkingPart,
+        and ToolCallPart content with styled banners/tokens as they stream in.
 
         Args:
             ctx: The run context.
             events: Async iterable of streaming events (PartStartEvent, PartDeltaEvent, etc.).
         """
         from pydantic_ai import PartDeltaEvent, PartStartEvent
-        from pydantic_ai.messages import TextPartDelta, ThinkingPartDelta
+        from pydantic_ai.messages import (
+            TextPartDelta,
+            ThinkingPartDelta,
+            ToolCallPartDelta,
+        )
         from rich.console import Console
-        from rich.markdown import Markdown
         from rich.markup import escape
 
         from code_puppy.messaging.spinner import pause_all_spinners
@@ -1284,29 +1366,36 @@ class BaseAgent(ABC):
             # Fallback if console not set (shouldn't happen in normal use)
             console = Console()
 
-        # Track which part indices we're currently streaming (for Text/Thinking parts)
+        # Track which part indices we're currently streaming (for Text/Thinking/Tool parts)
         streaming_parts: set[int] = set()
         thinking_parts: set[int] = (
             set()
         )  # Track which parts are thinking (for dim style)
         text_parts: set[int] = set()  # Track which parts are text
+        tool_parts: set[int] = set()  # Track which parts are tool calls
         banner_printed: set[int] = set()  # Track if banner was already printed
-        text_buffer: dict[int, list[str]] = {}  # Buffer text for final markdown render
-        token_count: dict[int, int] = {}  # Track token count per text part
+        token_count: dict[int, int] = {}  # Track token count per text/tool part
         did_stream_anything = False  # Track if we streamed any content
 
+        # Termflow streaming state for text parts
+        from termflow import Parser as TermflowParser
+        from termflow import Renderer as TermflowRenderer
+
+        termflow_parsers: dict[int, TermflowParser] = {}
+        termflow_renderers: dict[int, TermflowRenderer] = {}
+        termflow_line_buffers: dict[int, str] = {}  # Buffer incomplete lines
+
         def _print_thinking_banner() -> None:
             """Print the THINKING banner with spinner pause and line clear."""
             nonlocal did_stream_anything
-            import sys
             import time
 
             from code_puppy.config import get_banner_color
 
             pause_all_spinners()
             time.sleep(0.1)  # Delay to let spinner fully clear
-            sys.stdout.write("\r\x1b[K")  # Clear line
-            sys.stdout.flush()
+            # Clear line and print newline before banner
+            console.print(" " * 50, end="\r")
             console.print()  # Newline before banner
             # Bold banner with configurable color and lightning bolt
             thinking_color = get_banner_color("thinking")
@@ -1316,21 +1405,19 @@ class BaseAgent(ABC):
                 ),
                 end="",
             )
-            sys.stdout.flush()
             did_stream_anything = True
 
         def _print_response_banner() -> None:
             """Print the AGENT RESPONSE banner with spinner pause and line clear."""
             nonlocal did_stream_anything
-            import sys
             import time
 
             from code_puppy.config import get_banner_color
 
             pause_all_spinners()
             time.sleep(0.1)  # Delay to let spinner fully clear
-            sys.stdout.write("\r\x1b[K")  # Clear line
-            sys.stdout.flush()
+            # Clear line and print newline before banner
+            console.print(" " * 50, end="\r")
             console.print()  # Newline before banner
             response_color = get_banner_color("agent_response")
             console.print(
@@ -1338,7 +1425,6 @@ class BaseAgent(ABC):
                     f"[bold white on {response_color}] AGENT RESPONSE [/bold white on {response_color}]"
                 )
             )
-            sys.stdout.flush()
             did_stream_anything = True
 
         async for event in events:
@@ -1357,12 +1443,25 @@ class BaseAgent(ABC):
                 elif isinstance(part, TextPart):
                     streaming_parts.add(event.index)
                     text_parts.add(event.index)
-                    text_buffer[event.index] = []  # Initialize buffer
-                    token_count[event.index] = 0  # Initialize token counter
-                    # Buffer initial content if present
+                    # Initialize termflow streaming for this text part
+                    termflow_parsers[event.index] = TermflowParser()
+                    termflow_renderers[event.index] = TermflowRenderer(
+                        output=console.file, width=console.width
+                    )
+                    termflow_line_buffers[event.index] = ""
+                    # Handle initial content if present
                     if part.content and part.content.strip():
-                        text_buffer[event.index].append(part.content)
-                        token_count[event.index] += 1
+                        _print_response_banner()
+                        banner_printed.add(event.index)
+                        termflow_line_buffers[event.index] = part.content
+                elif isinstance(part, ToolCallPart):
+                    streaming_parts.add(event.index)
+                    tool_parts.add(event.index)
+                    token_count[event.index] = 0  # Initialize token counter
+                    # Track tool name for display
+                    banner_printed.add(
+                        event.index
+                    )  # Use banner_printed to track if we've shown tool info
 
             # PartDeltaEvent - stream the content as it arrives
             elif isinstance(event, PartDeltaEvent):
@@ -1370,23 +1469,29 @@ class BaseAgent(ABC):
                     delta = event.delta
                     if isinstance(delta, (TextPartDelta, ThinkingPartDelta)):
                         if delta.content_delta:
-                            # For text parts, show token counter then render at end
+                            # For text parts, stream markdown with termflow
                             if event.index in text_parts:
-                                import sys
-
                                 # Print banner on first content
                                 if event.index not in banner_printed:
                                     _print_response_banner()
                                     banner_printed.add(event.index)
-                                # Accumulate text for final markdown render
-                                text_buffer[event.index].append(delta.content_delta)
-                                token_count[event.index] += 1
-                                # Update token counter in place (single line)
-                                count = token_count[event.index]
-                                sys.stdout.write(
-                                    f"\r\x1b[K  ⏳ Receiving... {count} tokens"
+
+                                # Add content to line buffer
+                                termflow_line_buffers[event.index] += (
+                                    delta.content_delta
                                 )
-                                sys.stdout.flush()
+
+                                # Process complete lines
+                                parser = termflow_parsers[event.index]
+                                renderer = termflow_renderers[event.index]
+                                buffer = termflow_line_buffers[event.index]
+
+                                while "\n" in buffer:
+                                    line, buffer = buffer.split("\n", 1)
+                                    events_to_render = parser.parse_line(line)
+                                    renderer.render_all(events_to_render)
+
+                                termflow_line_buffers[event.index] = buffer
                             else:
                                 # For thinking parts, stream immediately (dim)
                                 if event.index not in banner_printed:
@@ -1394,44 +1499,72 @@ class BaseAgent(ABC):
                                     banner_printed.add(event.index)
                                 escaped = escape(delta.content_delta)
                                 console.print(f"[dim]{escaped}[/dim]", end="")
+                    elif isinstance(delta, ToolCallPartDelta):
+                        # For tool calls, count chunks received
+                        token_count[event.index] += 1
+                        # Get tool name if available
+                        tool_name = getattr(delta, "tool_name_delta", "")
+                        count = token_count[event.index]
+                        # Display with tool wrench icon and tool name
+                        if tool_name:
+                            console.print(
+                                f"  🔧 Calling {tool_name}... {count} chunks   ",
+                                end="\r",
+                            )
+                        else:
+                            console.print(
+                                f"  🔧 Calling tool... {count} chunks   ",
+                                end="\r",
+                            )
 
             # PartEndEvent - finish the streaming with a newline
             elif isinstance(event, PartEndEvent):
                 if event.index in streaming_parts:
-                    # For text parts, clear counter line and render markdown
+                    # For text parts, finalize termflow rendering
                     if event.index in text_parts:
-                        import sys
-
-                        # Clear the token counter line
-                        sys.stdout.write("\r\x1b[K")
-                        sys.stdout.flush()
-                        # Render the final markdown nicely
-                        if event.index in text_buffer:
-                            try:
-                                final_content = "".join(text_buffer[event.index])
-                                if final_content.strip():
-                                    console.print(Markdown(final_content))
-                            except Exception:
-                                pass
-                            del text_buffer[event.index]
-                        # Clean up token count
-                        token_count.pop(event.index, None)
+                        # Render any remaining buffered content
+                        if event.index in termflow_parsers:
+                            parser = termflow_parsers[event.index]
+                            renderer = termflow_renderers[event.index]
+                            remaining = termflow_line_buffers.get(event.index, "")
+
+                            # Parse and render any remaining partial line
+                            if remaining.strip():
+                                events_to_render = parser.parse_line(remaining)
+                                renderer.render_all(events_to_render)
+
+                            # Finalize the parser to close any open blocks
+                            final_events = parser.finalize()
+                            renderer.render_all(final_events)
+
+                            # Clean up termflow state
+                            del termflow_parsers[event.index]
+                            del termflow_renderers[event.index]
+                            del termflow_line_buffers[event.index]
+                    # For tool parts, clear the chunk counter line
+                    elif event.index in tool_parts:
+                        # Clear the chunk counter line by printing spaces and returning
+                        console.print(" " * 50, end="\r")
                     # For thinking parts, just print newline
                     elif event.index in banner_printed:
                         console.print()  # Final newline after streaming
+
+                    # Clean up token count
+                    token_count.pop(event.index, None)
                     # Clean up all tracking sets
                     streaming_parts.discard(event.index)
                     thinking_parts.discard(event.index)
                     text_parts.discard(event.index)
+                    tool_parts.discard(event.index)
                     banner_printed.discard(event.index)
 
-                    # Resume spinner if next part is NOT text/thinking (avoid race condition)
-                    # If next part is a tool call or None, it's safe to resume
+                    # Resume spinner if next part is NOT text/thinking/tool (avoid race condition)
+                    # If next part is None or handled differently, it's safe to resume
                     # Note: spinner itself handles blank line before appearing
                     from code_puppy.messaging.spinner import resume_all_spinners
 
                     next_kind = getattr(event, "next_part_kind", None)
-                    if next_kind not in ("text", "thinking"):
+                    if next_kind not in ("text", "thinking", "tool-call"):
                         resume_all_spinners()
 
         # Spinner is resumed in PartEndEvent when appropriate (based on next_part_kind)
@@ -1590,6 +1723,7 @@ class BaseAgent(ABC):
         *,
         attachments: Optional[Sequence[BinaryContent]] = None,
         link_attachments: Optional[Sequence[Union[ImageUrl, DocumentUrl]]] = None,
+        output_type: Optional[Type[Any]] = None,
         **kwargs,
     ) -> Any:
         """Run the agent with MCP servers, attachments, and full cancellation support.
@@ -1598,10 +1732,13 @@ class BaseAgent(ABC):
             prompt: Primary user prompt text (may be empty when attachments present).
             attachments: Local binary payloads (e.g., dragged images) to include.
             link_attachments: Remote assets (image/document URLs) to include.
+            output_type: Optional Pydantic model or type for structured output.
+                When provided, creates a temporary agent configured to return
+                this type instead of the default string output.
             **kwargs: Additional arguments forwarded to `pydantic_ai.Agent.run`.
 
         Returns:
-            The agent's response.
+            The agent's response (typed according to output_type if specified).
 
         Raises:
             asyncio.CancelledError: When execution is cancelled by user.
@@ -1625,6 +1762,11 @@ class BaseAgent(ABC):
         pydantic_agent = (
             self._code_generation_agent or self.reload_code_generation_agent()
         )
+
+        # If a custom output_type is specified, create a temporary agent with that type
+        if output_type is not None:
+            pydantic_agent = self._create_agent_with_output_type(output_type)
+
         # Handle claude-code and chatgpt-codex models: prepend system prompt to first user message
         from code_puppy.model_utils import is_chatgpt_codex_model, is_claude_code_model
 
@@ -1821,7 +1963,12 @@ class BaseAgent(ABC):
         def graceful_sigint_handler(_sig, _frame):
             # When using keyboard-based cancel, SIGINT should be a no-op
             # (just show a hint to user about the configured cancel key)
+            # Also reset terminal to prevent bricking on Windows+uvx
             from code_puppy.keymap import get_cancel_agent_display_name
+            from code_puppy.terminal_utils import reset_windows_terminal_full
+
+            # Reset terminal state first to prevent bricking
+            reset_windows_terminal_full()
 
             cancel_key = get_cancel_agent_display_name()
             emit_info(f"Use {cancel_key} to cancel the agent task.")