autodebug 0.5.1__tar.gz

autodebug-0.5.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 TEEJORAM
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

autodebug-0.5.1/PKG-INFO

@@ -0,0 +1,251 @@
+Metadata-Version: 2.1
+Name: autodebug
+Version: 0.5.1
+Summary: Let an AI agent debug your program with the debugger of VS Code, Cursor or Visual Studio
+Requires-Python: >=3.11
+License: MIT
+Classifier: Development Status :: 4 - Beta
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Software Development :: Debuggers
+Project-URL: Documentation, https://teejoram.github.io/AutoDebug/
+Project-URL: Source, https://github.com/TEEJORAM/AutoDebug
+Project-URL: Issues, https://github.com/TEEJORAM/AutoDebug/issues
+Description-Content-Type: text/markdown
+
+# AutoDebug
+
+AutoDebug lets an AI agent debug your program with the debugger of **VS Code**,
+**Cursor** or **Visual Studio 2022/2026**: it builds, sets breakpoints, runs,
+steps, reads variables, and writes a root-cause report into your project.
+Nothing runs until you click **Approve** in the IDE.
+
+Agents: Cursor's agent, Claude Code, GitHub Copilot (VS Code and Visual
+Studio), and any other agent that supports MCP or has a terminal.
+
+**Documentation:** [teejoram.github.io/AutoDebug](https://teejoram.github.io/AutoDebug/)
+
+## What you need
+
+- **Python 3.11 or newer** and **Git**.
+- One IDE, with the debugger for your language:
+
+| IDE | C++ | Python |
+|---|---|---|
+| VS Code | **C/C++** or **CodeLLDB** extension | **Python** and **Python Debugger** extensions |
+| Cursor | **CodeLLDB** extension (Microsoft's C/C++ debugger does not run in Cursor) | **Python** and **Python Debugger** extensions |
+| Visual Studio 2022 or 2026 | the **Desktop development with C++** workload | not supported yet |
+
+## Connect (about 2 minutes)
+
+**1. Install AutoDebug.** If you use Visual Studio, close it first.
+
+Windows (PowerShell):
+
+```powershell
+py -m pip install git+https://github.com/TEEJORAM/AutoDebug.git
+py -m autodebug setup
+```
+
+macOS / Linux:
+
+```sh
+python3 -m venv ~/.autodebug
+~/.autodebug/bin/python -m pip install git+https://github.com/TEEJORAM/AutoDebug.git
+~/.autodebug/bin/python -m autodebug setup
+```
+
+`setup` finds every IDE and agent on the computer and connects each one: it
+installs the AutoDebug extension (VS Code, Cursor, Visual Studio) and adds the
+`autodebug` MCP server (VS Code, Visual Studio, Claude Code). In Cursor the
+extension registers the server itself.
+
+**2. Do the one step `setup` prints for your IDE:**
+
+| IDE | Step |
+|---|---|
+| VS Code | Command Palette → **Developer: Reload Window**. If VS Code asks whether to trust or start the **autodebug** server, allow it. |
+| Cursor | Command Palette → **Developer: Reload Window**, in every Cursor window. The extension registers the server with Cursor itself: **Cursor Settings → Customize → MCPs** shows **extension-autodebug**, already switched on. |
+| Visual Studio | Start it and open your solution. Set the startup project and the **Debug** configuration. For Copilot in Visual Studio only: in the chat, **Tools** → tick **autodebug**. |
+| Claude Code | Nothing. Start a new session. |
+
+**3. Check:** `py -m autodebug doctor` (macOS / Linux: `~/.autodebug/bin/python -m autodebug doctor`).
+It lists the connected IDEs, the open IDE windows and what to fix.
+
+**Optional, per project:** `py -m autodebug setup C:\path\to\project` also
+writes an AutoDebug section into the project's `AGENTS.md`. Agents read it and
+use the terminal when their app hides the MCP tools (see below).
+
+## Which agent debugs in which IDE
+
+The agent and the IDE are separate. Any connected agent can debug in any IDE
+window that runs the AutoDebug extension:
+
+- **Cursor's agent** debugs in Cursor, VS Code or Visual Studio.
+- **Claude Code** debugs in VS Code, Cursor or Visual Studio.
+- **Copilot** debugs in VS Code (agent mode), or in Visual Studio (agent mode).
+
+When windows of more than one IDE are open, name the IDE in the prompt, for
+example "use the Visual Studio window".
+
+## Debug
+
+Open the project in the IDE and ask the agent, for example:
+
+> The app crashes when I delete the last item in a list. Use the autodebug
+> tools: build the Debug configuration, stop where the item is removed, find
+> the root cause and write a report.
+
+1. The IDE shows the **Approve** dialog with the goal, the build, the program,
+   the breakpoints and the limits. Click **Approve** within 2 minutes.
+2. The agent builds, stops at the breakpoints, steps and traces.
+3. The report is written to `autodebug-reports/` in the project folder.
+
+Try it first on [examples/AutoDebugDemo](https://github.com/TEEJORAM/AutoDebug/blob/main/examples/AutoDebugDemo/README.md):
+the same bug in C++ and Python.
+
+## When the agent cannot see the autodebug tools
+
+Some apps drop MCP tools from a chat without an error. The agent can then run
+the same tools in its terminal, in the same debug session:
+
+```powershell
+py -m autodebug run list_targets ide=visualstudio
+py -m autodebug run help
+```
+
+`autodebug setup <project folder>` writes the exact command for your computer
+into the project's `AGENTS.md`, so the agent finds it by itself.
+
+## Troubleshooting
+
+| What you see | What to do |
+|---|---|
+| The agent has no autodebug tools | Cursor: reload the window and start a new chat; **Settings → Customize → MCPs** should show **extension-autodebug** with a green dot. `autodebug doctor` flags an old `mcp.json` entry that gets in the way. Visual Studio Copilot: **Tools** → tick **autodebug**. Any agent: it can use `autodebug run` (above). |
+| "No VS Code, Cursor or Visual Studio window is running the extension" | Open the project in the IDE. VS Code / Cursor: **Developer: Reload Window**; the status bar shows **AutoDebug**. Visual Studio: View → Output → **AutoDebug** says it is listening. |
+| No Approve dialog | Switch to the IDE window: the dialog opens there (in Visual Studio it stays on top of other windows). The agent reports "no answer within 2 minutes" when nobody answered. |
+| `NOT_BUILT` | The program has not been built. The agent retries with the build task it names. |
+| Visual Studio is open during setup | Close every Visual Studio window, then run `setup` again. |
+| No debugger for type "…" | Install that debugger extension in the IDE (table above) and reload the window. |
+
+## Update
+
+```powershell
+py -m pip install --force-reinstall git+https://github.com/TEEJORAM/AutoDebug.git
+py -m autodebug setup
+```
+
+## Commands
+
+| Command | Does |
+|---|---|
+| `autodebug setup` | Connect every IDE and agent found. `setup cursor`, `vscode`, `visualstudio` or `claude` connects one; add a project folder for project-only entries and `AGENTS.md` |
+| `autodebug doctor [folder]` | Check the setup, the open IDE windows and the AutoDebug service |
+| `autodebug run <tool> key=value …` | Call a debugging tool from a terminal; `run help` lists them |
+| `autodebug serve` | The MCP server; the IDE or agent starts it |
+
+On Windows write `py -m autodebug`; on macOS / Linux `~/.autodebug/bin/python -m autodebug`.
+
+## Agent tools
+
+| Tool | What it does |
+|---|---|
+| `list_targets` | IDE windows, their launch configurations and build tasks |
+| `start_session` | Approval, build, breakpoints, launch or attach; waits for the first stop |
+| `wait` · `pause` · `stop_session` | Follow, interrupt or end the session. `wait` can `ask_developer` to reproduce a GUI step |
+| `step` | Over, into or out, 1–20 times, with the variables that changed |
+| `continue` | Run to the next stop, or past other breakpoints to a line |
+| `inspect` · `stack` · `threads` | Variables (including nested paths), the call stack, and threads |
+| `breakpoints` · `exceptions` | Source and function breakpoints; exception filters |
+| `trace` | Collect variables every time a line runs |
+| `write_report` | Root-cause report in `autodebug-reports/` with a timeline of the session |
+
+The agent cannot evaluate expressions or change variables, and secret-looking
+values are hidden from it. AutoDebug listens on 127.0.0.1 only, with a token.
+How it works: [docs/design.md](https://github.com/TEEJORAM/AutoDebug/blob/main/docs/design.md).
+
+AutoDebug is released under the [MIT License](https://github.com/TEEJORAM/AutoDebug/blob/main/LICENSE).
+
+## Large C++ programs
+
+**Build with symbols.** Debug the `Debug` or `RelWithDebInfo` build; optimised
+builds without debug info show wrong lines and missing variables. Let the
+build run first through the launch configuration's `preLaunchTask` or the
+agent's `build_task`. A long compile gets `build_minutes` (default 30, up to
+600), separate from the debugging time limit.
+
+**Launch configurations.** For example, in `.vscode/launch.json`:
+
+```json
+{
+  "name": "Debug app (CodeLLDB)",
+  "type": "lldb",
+  "request": "launch",
+  "program": "${workspaceFolder}/build/bin/app",
+  "cwd": "${workspaceFolder}",
+  "preLaunchTask": "build",
+  "windows": { "program": "${workspaceFolder}/build/bin/Debug/app.exe" }
+},
+{
+  "name": "Attach to app (CodeLLDB)",
+  "type": "lldb",
+  "request": "attach",
+  "pid": "${command:pickMyProcess}"
+}
+```
+
+The Approve dialog shows the whole configuration, the build commands with
+everything they depend on, and the environment. Configurations that prompt for
+input or debug another computer are refused.
+
+**Attach to a running application.** Start the app yourself, get to the state
+just before the bug, and have the agent attach with `process_id`. When the
+session ends the debugger detaches and the app keeps running.
+
+**What helps in a big codebase:**
+
+- **Slow starts.** Loading symbols for a large binary can take minutes; the
+  agent keeps calling `wait`.
+- **Shared libraries and plugins.** Breakpoints in them bind when the library
+  loads; until then they show as not bound yet, with the debugger's reason.
+- **Code outside the workspace.** Pass `source_folders` for third-party
+  libraries or an SDK that live outside the project folder.
+- **Threads.** `threads` lists them and selects one for `stack`, `inspect` and
+  `step`.
+- **Deep stacks.** `stack` pages through up to 200 frames at a time.
+- **Objects and containers.** `inspect` with `depth` 1–3 opens structs, pointers
+  and containers. Names can be paths: `scene->objects[2].transform`.
+- **Exceptions.** `exceptions` lists the debugger's exception filters (CodeLLDB
+  offers C++ throw and catch) and turns them on. A stop on an exception reports
+  which one.
+- **GUI reproduction.** `wait` with `ask_developer` shows you a note such as
+  "Click Delete on the last item". You do it, and the program stops at the
+  breakpoint.
+- **Embedded Python.** If the application embeds Python, start debugpy inside
+  it and attach with a debugpy configuration (the sample has
+  **Attach to Python (port 5678)**).
+- **Function breakpoints.** Break on `Document::removeItem` without knowing the file.
+
+Values the agent reads are sent to the AI model, like any other tool result.
+Debug only programs and data you are allowed to share with it.
+
+## Develop
+
+```sh
+PYTHONPATH=src python -m unittest discover -s tests
+```
+
+All of AutoDebug's logic is Python (`src/autodebug`). The IDE extension is a
+small JavaScript relay (`src/autodebug/extension/relay.js`) for VS Code and
+Cursor, and a C# relay (`src/autodebug/visualstudio`) for Visual Studio. The
+relays show the Approve dialog and pass debugger requests and events through.
+
+To work on AutoDebug itself, clone the repository and install it editable:
+`py -m pip install -e .` (Windows) or `python3 -m pip install -e .`.
+
+Most tests run with Python alone. The relay tests need Node.js. The
+real-debugger tests also need `pip install debugpy` (Python) and gdb 14 or later
+with g++ (C++, Linux and macOS). See [CLAUDE.md](https://github.com/TEEJORAM/AutoDebug/blob/main/CLAUDE.md) for the rules.

autodebug-0.5.1/README.md

@@ -0,0 +1,233 @@
+# AutoDebug
+
+AutoDebug lets an AI agent debug your program with the debugger of **VS Code**,
+**Cursor** or **Visual Studio 2022/2026**: it builds, sets breakpoints, runs,
+steps, reads variables, and writes a root-cause report into your project.
+Nothing runs until you click **Approve** in the IDE.
+
+Agents: Cursor's agent, Claude Code, GitHub Copilot (VS Code and Visual
+Studio), and any other agent that supports MCP or has a terminal.
+
+**Documentation:** [teejoram.github.io/AutoDebug](https://teejoram.github.io/AutoDebug/)
+
+## What you need
+
+- **Python 3.11 or newer** and **Git**.
+- One IDE, with the debugger for your language:
+
+| IDE | C++ | Python |
+|---|---|---|
+| VS Code | **C/C++** or **CodeLLDB** extension | **Python** and **Python Debugger** extensions |
+| Cursor | **CodeLLDB** extension (Microsoft's C/C++ debugger does not run in Cursor) | **Python** and **Python Debugger** extensions |
+| Visual Studio 2022 or 2026 | the **Desktop development with C++** workload | not supported yet |
+
+## Connect (about 2 minutes)
+
+**1. Install AutoDebug.** If you use Visual Studio, close it first.
+
+Windows (PowerShell):
+
+```powershell
+py -m pip install git+https://github.com/TEEJORAM/AutoDebug.git
+py -m autodebug setup
+```
+
+macOS / Linux:
+
+```sh
+python3 -m venv ~/.autodebug
+~/.autodebug/bin/python -m pip install git+https://github.com/TEEJORAM/AutoDebug.git
+~/.autodebug/bin/python -m autodebug setup
+```
+
+`setup` finds every IDE and agent on the computer and connects each one: it
+installs the AutoDebug extension (VS Code, Cursor, Visual Studio) and adds the
+`autodebug` MCP server (VS Code, Visual Studio, Claude Code). In Cursor the
+extension registers the server itself.
+
+**2. Do the one step `setup` prints for your IDE:**
+
+| IDE | Step |
+|---|---|
+| VS Code | Command Palette → **Developer: Reload Window**. If VS Code asks whether to trust or start the **autodebug** server, allow it. |
+| Cursor | Command Palette → **Developer: Reload Window**, in every Cursor window. The extension registers the server with Cursor itself: **Cursor Settings → Customize → MCPs** shows **extension-autodebug**, already switched on. |
+| Visual Studio | Start it and open your solution. Set the startup project and the **Debug** configuration. For Copilot in Visual Studio only: in the chat, **Tools** → tick **autodebug**. |
+| Claude Code | Nothing. Start a new session. |
+
+**3. Check:** `py -m autodebug doctor` (macOS / Linux: `~/.autodebug/bin/python -m autodebug doctor`).
+It lists the connected IDEs, the open IDE windows and what to fix.
+
+**Optional, per project:** `py -m autodebug setup C:\path\to\project` also
+writes an AutoDebug section into the project's `AGENTS.md`. Agents read it and
+use the terminal when their app hides the MCP tools (see below).
+
+## Which agent debugs in which IDE
+
+The agent and the IDE are separate. Any connected agent can debug in any IDE
+window that runs the AutoDebug extension:
+
+- **Cursor's agent** debugs in Cursor, VS Code or Visual Studio.
+- **Claude Code** debugs in VS Code, Cursor or Visual Studio.
+- **Copilot** debugs in VS Code (agent mode), or in Visual Studio (agent mode).
+
+When windows of more than one IDE are open, name the IDE in the prompt, for
+example "use the Visual Studio window".
+
+## Debug
+
+Open the project in the IDE and ask the agent, for example:
+
+> The app crashes when I delete the last item in a list. Use the autodebug
+> tools: build the Debug configuration, stop where the item is removed, find
+> the root cause and write a report.
+
+1. The IDE shows the **Approve** dialog with the goal, the build, the program,
+   the breakpoints and the limits. Click **Approve** within 2 minutes.
+2. The agent builds, stops at the breakpoints, steps and traces.
+3. The report is written to `autodebug-reports/` in the project folder.
+
+Try it first on [examples/AutoDebugDemo](https://github.com/TEEJORAM/AutoDebug/blob/main/examples/AutoDebugDemo/README.md):
+the same bug in C++ and Python.
+
+## When the agent cannot see the autodebug tools
+
+Some apps drop MCP tools from a chat without an error. The agent can then run
+the same tools in its terminal, in the same debug session:
+
+```powershell
+py -m autodebug run list_targets ide=visualstudio
+py -m autodebug run help
+```
+
+`autodebug setup <project folder>` writes the exact command for your computer
+into the project's `AGENTS.md`, so the agent finds it by itself.
+
+## Troubleshooting
+
+| What you see | What to do |
+|---|---|
+| The agent has no autodebug tools | Cursor: reload the window and start a new chat; **Settings → Customize → MCPs** should show **extension-autodebug** with a green dot. `autodebug doctor` flags an old `mcp.json` entry that gets in the way. Visual Studio Copilot: **Tools** → tick **autodebug**. Any agent: it can use `autodebug run` (above). |
+| "No VS Code, Cursor or Visual Studio window is running the extension" | Open the project in the IDE. VS Code / Cursor: **Developer: Reload Window**; the status bar shows **AutoDebug**. Visual Studio: View → Output → **AutoDebug** says it is listening. |
+| No Approve dialog | Switch to the IDE window: the dialog opens there (in Visual Studio it stays on top of other windows). The agent reports "no answer within 2 minutes" when nobody answered. |
+| `NOT_BUILT` | The program has not been built. The agent retries with the build task it names. |
+| Visual Studio is open during setup | Close every Visual Studio window, then run `setup` again. |
+| No debugger for type "…" | Install that debugger extension in the IDE (table above) and reload the window. |
+
+## Update
+
+```powershell
+py -m pip install --force-reinstall git+https://github.com/TEEJORAM/AutoDebug.git
+py -m autodebug setup
+```
+
+## Commands
+
+| Command | Does |
+|---|---|
+| `autodebug setup` | Connect every IDE and agent found. `setup cursor`, `vscode`, `visualstudio` or `claude` connects one; add a project folder for project-only entries and `AGENTS.md` |
+| `autodebug doctor [folder]` | Check the setup, the open IDE windows and the AutoDebug service |
+| `autodebug run <tool> key=value …` | Call a debugging tool from a terminal; `run help` lists them |
+| `autodebug serve` | The MCP server; the IDE or agent starts it |
+
+On Windows write `py -m autodebug`; on macOS / Linux `~/.autodebug/bin/python -m autodebug`.
+
+## Agent tools
+
+| Tool | What it does |
+|---|---|
+| `list_targets` | IDE windows, their launch configurations and build tasks |
+| `start_session` | Approval, build, breakpoints, launch or attach; waits for the first stop |
+| `wait` · `pause` · `stop_session` | Follow, interrupt or end the session. `wait` can `ask_developer` to reproduce a GUI step |
+| `step` | Over, into or out, 1–20 times, with the variables that changed |
+| `continue` | Run to the next stop, or past other breakpoints to a line |
+| `inspect` · `stack` · `threads` | Variables (including nested paths), the call stack, and threads |
+| `breakpoints` · `exceptions` | Source and function breakpoints; exception filters |
+| `trace` | Collect variables every time a line runs |
+| `write_report` | Root-cause report in `autodebug-reports/` with a timeline of the session |
+
+The agent cannot evaluate expressions or change variables, and secret-looking
+values are hidden from it. AutoDebug listens on 127.0.0.1 only, with a token.
+How it works: [docs/design.md](https://github.com/TEEJORAM/AutoDebug/blob/main/docs/design.md).
+
+AutoDebug is released under the [MIT License](https://github.com/TEEJORAM/AutoDebug/blob/main/LICENSE).
+
+## Large C++ programs
+
+**Build with symbols.** Debug the `Debug` or `RelWithDebInfo` build; optimised
+builds without debug info show wrong lines and missing variables. Let the
+build run first through the launch configuration's `preLaunchTask` or the
+agent's `build_task`. A long compile gets `build_minutes` (default 30, up to
+600), separate from the debugging time limit.
+
+**Launch configurations.** For example, in `.vscode/launch.json`:
+
+```json
+{
+  "name": "Debug app (CodeLLDB)",
+  "type": "lldb",
+  "request": "launch",
+  "program": "${workspaceFolder}/build/bin/app",
+  "cwd": "${workspaceFolder}",
+  "preLaunchTask": "build",
+  "windows": { "program": "${workspaceFolder}/build/bin/Debug/app.exe" }
+},
+{
+  "name": "Attach to app (CodeLLDB)",
+  "type": "lldb",
+  "request": "attach",
+  "pid": "${command:pickMyProcess}"
+}
+```
+
+The Approve dialog shows the whole configuration, the build commands with
+everything they depend on, and the environment. Configurations that prompt for
+input or debug another computer are refused.
+
+**Attach to a running application.** Start the app yourself, get to the state
+just before the bug, and have the agent attach with `process_id`. When the
+session ends the debugger detaches and the app keeps running.
+
+**What helps in a big codebase:**
+
+- **Slow starts.** Loading symbols for a large binary can take minutes; the
+  agent keeps calling `wait`.
+- **Shared libraries and plugins.** Breakpoints in them bind when the library
+  loads; until then they show as not bound yet, with the debugger's reason.
+- **Code outside the workspace.** Pass `source_folders` for third-party
+  libraries or an SDK that live outside the project folder.
+- **Threads.** `threads` lists them and selects one for `stack`, `inspect` and
+  `step`.
+- **Deep stacks.** `stack` pages through up to 200 frames at a time.
+- **Objects and containers.** `inspect` with `depth` 1–3 opens structs, pointers
+  and containers. Names can be paths: `scene->objects[2].transform`.
+- **Exceptions.** `exceptions` lists the debugger's exception filters (CodeLLDB
+  offers C++ throw and catch) and turns them on. A stop on an exception reports
+  which one.
+- **GUI reproduction.** `wait` with `ask_developer` shows you a note such as
+  "Click Delete on the last item". You do it, and the program stops at the
+  breakpoint.
+- **Embedded Python.** If the application embeds Python, start debugpy inside
+  it and attach with a debugpy configuration (the sample has
+  **Attach to Python (port 5678)**).
+- **Function breakpoints.** Break on `Document::removeItem` without knowing the file.
+
+Values the agent reads are sent to the AI model, like any other tool result.
+Debug only programs and data you are allowed to share with it.
+
+## Develop
+
+```sh
+PYTHONPATH=src python -m unittest discover -s tests
+```
+
+All of AutoDebug's logic is Python (`src/autodebug`). The IDE extension is a
+small JavaScript relay (`src/autodebug/extension/relay.js`) for VS Code and
+Cursor, and a C# relay (`src/autodebug/visualstudio`) for Visual Studio. The
+relays show the Approve dialog and pass debugger requests and events through.
+
+To work on AutoDebug itself, clone the repository and install it editable:
+`py -m pip install -e .` (Windows) or `python3 -m pip install -e .`.
+
+Most tests run with Python alone. The relay tests need Node.js. The
+real-debugger tests also need `pip install debugpy` (Python) and gdb 14 or later
+with g++ (C++, Linux and macOS). See [CLAUDE.md](https://github.com/TEEJORAM/AutoDebug/blob/main/CLAUDE.md) for the rules.

autodebug-0.5.1/build_backend.py

@@ -0,0 +1,105 @@
+"""Standard-library PEP 517 backend: builds AutoDebug without downloading anything."""
+
+from __future__ import annotations
+
+import base64
+import csv
+import hashlib
+import io
+import tarfile
+import tomllib
+import zipfile
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent
+PROJECT = tomllib.loads((ROOT / "pyproject.toml").read_text(encoding="utf-8"))["project"]
+NAME, VERSION = PROJECT["name"], PROJECT["version"]
+DIST_INFO = f"{NAME}-{VERSION}.dist-info"
+PACKAGE = ROOT / "src" / "autodebug"
+
+
+def _package_files() -> list[Path]:
+    """Python modules plus the extension sources the wheel must carry."""
+    files = [p for p in PACKAGE.rglob("*.py") if "__pycache__" not in p.parts]
+    files += [p for p in (PACKAGE / "extension").glob("*") if p.suffix in (".js", ".json", ".md")]
+    return sorted(files)
+
+
+def _metadata() -> dict[str, bytes]:
+    scripts = "\n".join(f"{name} = {target}" for name, target in PROJECT.get("scripts", {}).items())
+    headers = [f"Metadata-Version: 2.1", f"Name: {NAME}", f"Version: {VERSION}", f"Summary: {PROJECT['description']}",
+               f"Requires-Python: {PROJECT['requires-python']}"]
+    if PROJECT.get("license"):
+        headers.append(f"License: {PROJECT['license']}")
+    headers += [f"Classifier: {c}" for c in PROJECT.get("classifiers", [])]
+    headers += [f"Project-URL: {label}, {url}" for label, url in PROJECT.get("urls", {}).items()]
+    readme = (ROOT / PROJECT["readme"]).read_text(encoding="utf-8") if PROJECT.get("readme") else ""
+    if readme:
+        headers.append("Description-Content-Type: text/markdown")
+    return {
+        "METADATA": ("\n".join(headers) + "\n\n" + readme).encode("utf-8"),
+        "WHEEL": b"Wheel-Version: 1.0\nGenerator: autodebug\nRoot-Is-Purelib: true\nTag: py3-none-any\n",
+        "entry_points.txt": f"[console_scripts]\n{scripts}\n".encode(),
+    }
+
+
+def _wheel(directory: str, editable: bool) -> str:
+    filename = f"{NAME}-{VERSION}-py3-none-any.whl"
+    records = []
+    with zipfile.ZipFile(Path(directory) / filename, "w", zipfile.ZIP_DEFLATED) as wheel:
+        def add(name: str, data: bytes) -> None:
+            wheel.writestr(name, data)
+            digest = base64.urlsafe_b64encode(hashlib.sha256(data).digest()).rstrip(b"=").decode()
+            records.append((name, f"sha256={digest}", str(len(data))))
+
+        if editable:
+            add(f"__editable__.{NAME}.pth", f"{ROOT / 'src'}\n".encode())
+        else:
+            for path in _package_files():
+                add(path.relative_to(ROOT / "src").as_posix(), path.read_bytes())
+        for name, data in _metadata().items():
+            add(f"{DIST_INFO}/{name}", data)
+        record = io.StringIO()
+        csv.writer(record, lineterminator="\n").writerows([*records, (f"{DIST_INFO}/RECORD", "", "")])
+        wheel.writestr(f"{DIST_INFO}/RECORD", record.getvalue())
+    return filename
+
+
+def get_requires_for_build_wheel(config_settings=None):
+    return []
+
+
+get_requires_for_build_editable = get_requires_for_build_sdist = get_requires_for_build_wheel
+
+
+def prepare_metadata_for_build_wheel(metadata_directory, config_settings=None):
+    folder = Path(metadata_directory) / DIST_INFO
+    folder.mkdir(parents=True, exist_ok=True)
+    for name, data in _metadata().items():
+        (folder / name).write_bytes(data)
+    return DIST_INFO
+
+
+prepare_metadata_for_build_editable = prepare_metadata_for_build_wheel
+
+
+def build_wheel(wheel_directory, config_settings=None, metadata_directory=None):
+    return _wheel(wheel_directory, editable=False)
+
+
+def build_editable(wheel_directory, config_settings=None, metadata_directory=None):
+    return _wheel(wheel_directory, editable=True)
+
+
+def build_sdist(sdist_directory, config_settings=None):
+    filename = f"{NAME}-{VERSION}.tar.gz"
+    prefix = f"{NAME}-{VERSION}"
+    with tarfile.open(Path(sdist_directory) / filename, "w:gz") as archive:
+        metadata = _metadata()["METADATA"]
+        info = tarfile.TarInfo(f"{prefix}/PKG-INFO")
+        info.size = len(metadata)
+        archive.addfile(info, io.BytesIO(metadata))
+        for path in [ROOT / "pyproject.toml", ROOT / "build_backend.py", ROOT / "README.md", ROOT / "LICENSE",
+                     *_package_files()]:
+            archive.add(path, arcname=f"{prefix}/{path.relative_to(ROOT).as_posix()}", recursive=False)
+    return filename

autodebug-0.5.1/pyproject.toml

@@ -0,0 +1,30 @@
+[project]
+name = "autodebug"
+version = "0.5.1"
+description = "Let an AI agent debug your program with the debugger of VS Code, Cursor or Visual Studio"
+requires-python = ">=3.11"
+readme = "README.md"
+license = "MIT"
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "License :: OSI Approved :: MIT License",
+    "Environment :: Console",
+    "Intended Audience :: Developers",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3",
+    "Topic :: Software Development :: Debuggers",
+]
+dependencies = []
+
+[project.urls]
+Documentation = "https://teejoram.github.io/AutoDebug/"
+Source = "https://github.com/TEEJORAM/AutoDebug"
+Issues = "https://github.com/TEEJORAM/AutoDebug/issues"
+
+[project.scripts]
+autodebug = "autodebug.cli:main"
+
+[build-system]
+requires = []
+build-backend = "build_backend"
+backend-path = ["."]

autodebug-0.5.1/src/autodebug/__init__.py

@@ -0,0 +1,3 @@
+"""AutoDebug: let an AI agent debug a program with the IDE's own debugger."""
+
+__version__ = "0.5.1"

autodebug-0.5.1/src/autodebug/__main__.py

@@ -0,0 +1,11 @@
+if not __package__:  # run as `python <checkout>/src/autodebug`: no install or PYTHONPATH needed
+    import os
+    import sys
+
+    here = os.path.dirname(os.path.abspath(__file__))
+    sys.path[:] = [p for p in sys.path if os.path.abspath(p or ".") != here]
+    sys.path.insert(0, os.path.dirname(here))
+
+from autodebug.cli import main
+
+raise SystemExit(main())