agentic-python-coder 3.4.1__tar.gz → 3.4.3__tar.gz

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.5
 Name: agentic-python-coder
-Version: 3.4.1
+Version: 3.4.3
 Summary: A lightweight Python coding agent that writes, executes, and iterates on code through natural language instructions
 Author: Stefan Szeider
 License: Apache-2.0
@@ -11,12 +11,14 @@ Classifier: Intended Audience :: Developers
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Programming Language :: Python :: 3.15
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Software Development :: Code Generators
-Requires-Python: <3.14,>=3.13
+Requires-Python: >=3.13
 Requires-Dist: ipykernel>=7.3.0
 Requires-Dist: jupyter-client>=8.9.1
-Requires-Dist: mcp>=1.28.1
+Requires-Dist: mcp<2,>=1.28.1
 Requires-Dist: openai>=2.45.0
 Requires-Dist: python-dotenv>=1.2.2
 Requires-Dist: pyyaml>=6.0.3
@@ -33,7 +35,7 @@ Description-Content-Type: text/markdown
 
 # Agentic Python Coder
 
-[![Python 3.13](https://img.shields.io/badge/python-3.13-blue.svg)](https://www.python.org/downloads/)
+[![Python 3.13](https://img.shields.io/badge/python-3.13%2B-blue.svg)](https://www.python.org/downloads/)
 [![License: Apache 2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 [![MCP Compatible](https://img.shields.io/badge/MCP-Compatible-green.svg)](https://modelcontextprotocol.io/)
 [![UV](https://img.shields.io/badge/Packaged%20with-UV-purple)](https://github.com/astral-sh/uv)
@@ -51,14 +53,14 @@ For details on architecture and constraint modelling applications, see [[Szeider
 
 ### Prerequisites
 
-- Python 3.13
+- Python 3.13 or newer
 - UV package manager: `curl -LsSf https://astral.sh/uv/install.sh | sh`
 
 ### For the Coding Agent
 
 ```bash
-# Install as CLI tool
-uv tool install agentic-python-coder
+# Install as CLI tool (--python 3.14: ortools, used by cpmpy, has no 3.15 wheels yet)
+uv tool install --python 3.14 agentic-python-coder
 
 # Set up OpenRouter API key
 mkdir -p ~/.config/coder
@@ -97,7 +99,7 @@ Add to your Claude Desktop MCP configuration:
   "mcpServers": {
     "ipython": {
       "command": "uvx",
-      "args": ["--from", "agentic-python-coder", "ipython_mcp"]
+      "args": ["--python", "3.14", "--from", "agentic-python-coder", "ipython_mcp"]
     }
   }
 }
@@ -132,7 +134,7 @@ coder -i
 | `--version`, `-V` | Show version and exit |
 | `--init [TEMPLATE]` | Initialize example templates (cpmpy, clingo, regex, or all) |
 | `--task`, `-t FILE` | Load task from markdown file |
-| `--model MODEL` | Model name or JSON file (default: sonnet45) |
+| `--model MODEL` | Model name or JSON file (default: sonnet55) |
 | `--project`, `-p FILE` | Project template for domain-specific prompts |
 | `--with PACKAGE` | Add packages dynamically (repeatable) |
 | `--dir`, `-d DIR` | Working directory |
@@ -146,21 +148,25 @@ coder -i
 
 ```bash
 # Built-in models (versioned names)
-coder --model sonnet45 "task"   # Claude Sonnet 4.5 (default)
+coder --model sonnet45 "task"   # Claude Sonnet 4.5
 coder --model sonnet46 "task"   # Claude Sonnet 4.6
 coder --model sonnet5 "task"    # Claude Sonnet 5
 coder --model opus45 "task"     # Claude Opus 4.5
 coder --model opus48 "task"     # Claude Opus 4.8
+coder --model sonnet55 "task"   # Claude Sonnet 5.5 (default)
+coder --model opus55 "task"     # Claude Opus 5.5
 coder --model deepseek31 "task" # DeepSeek v3.1
-coder --model grok41 "task"     # X.AI Grok 4.1
+coder --model grok47 "task"     # X.AI Grok 4.7
 coder --model qwen3 "task"      # Qwen3 Coder
 coder --model gemini25 "task"     # Gemini Pro 2.5
-coder --model gemini3pro "task"    # Gemini 3 Pro Preview
 coder --model gemini31 "task"      # Gemini 3.1 Pro Preview
 coder --model gemini3flash "task"  # Gemini 3 Flash Preview (fast, low-cost)
 coder --model gpt52 "task"         # GPT-5.2
 coder --model gpt56sol "task"      # GPT-5.6 Sol
 coder --model gpt56terra "task"    # GPT-5.6 Terra
+coder --model gpt6sol "task"       # GPT-6 Sol
+coder --model gpt6astra "task"     # GPT-6 Astra
+coder --model gemini4argon "task"  # Gemini 4 Argon
 
 # Custom model (JSON file)
 coder --model ./mymodel.json "task"
@@ -217,7 +223,7 @@ import agentic_python_coder as coder
 messages, stats, log_path = coder.solve_task(
     "Write a fibonacci function",
     working_directory="/tmp/workspace",
-    model="sonnet45",
+    model="sonnet55",
     quiet=True,
 )
 
@@ -236,7 +242,7 @@ from agentic_python_coder import solve_task
 messages, stats, log_path = solve_task(
     task="Your task description",
     working_directory=".",           # Where to run and save files
-    model=None,                      # Model name: "sonnet45", "opus45", or JSON file
+    model=None,                      # Model name: "sonnet55", "opus55", or JSON file
     system_prompt=None,              # Custom system prompt (string)
     system_prompt_path=None,         # Path to system prompt file
     project_prompt=None,             # Domain-specific context
@@ -300,9 +306,9 @@ agent = create_coding_agent(
 ```python
 from agentic_python_coder import get_openrouter_llm, list_available_models
 
-llm = get_openrouter_llm(model="sonnet45")
+llm = get_openrouter_llm(model="sonnet55")
 print(list_available_models())
-# ['deepseek31', 'gemini25', 'gemini3flash', 'gemini3pro', 'gpt52', 'grok41', 'opus45', 'qwen3', 'sonnet45']
+# ['deepseek31', 'gemini25', 'gemini31', 'gemini3flash', 'gpt52', ...]
 ```
 
 ---
@@ -320,7 +326,7 @@ Add to your MCP settings (e.g., `~/.claude/claude_desktop_config.json` or projec
   "mcpServers": {
     "ipython": {
       "command": "uvx",
-      "args": ["--from", "agentic-python-coder", "ipython_mcp"]
+      "args": ["--python", "3.14", "--from", "agentic-python-coder", "ipython_mcp"]
     }
   }
 }

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/README.md

@@ -1,6 +1,6 @@
 # Agentic Python Coder
 
-[![Python 3.13](https://img.shields.io/badge/python-3.13-blue.svg)](https://www.python.org/downloads/)
+[![Python 3.13](https://img.shields.io/badge/python-3.13%2B-blue.svg)](https://www.python.org/downloads/)
 [![License: Apache 2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 [![MCP Compatible](https://img.shields.io/badge/MCP-Compatible-green.svg)](https://modelcontextprotocol.io/)
 [![UV](https://img.shields.io/badge/Packaged%20with-UV-purple)](https://github.com/astral-sh/uv)
@@ -18,14 +18,14 @@ For details on architecture and constraint modelling applications, see [[Szeider
 
 ### Prerequisites
 
-- Python 3.13
+- Python 3.13 or newer
 - UV package manager: `curl -LsSf https://astral.sh/uv/install.sh | sh`
 
 ### For the Coding Agent
 
 ```bash
-# Install as CLI tool
-uv tool install agentic-python-coder
+# Install as CLI tool (--python 3.14: ortools, used by cpmpy, has no 3.15 wheels yet)
+uv tool install --python 3.14 agentic-python-coder
 
 # Set up OpenRouter API key
 mkdir -p ~/.config/coder
@@ -64,7 +64,7 @@ Add to your Claude Desktop MCP configuration:
   "mcpServers": {
     "ipython": {
       "command": "uvx",
-      "args": ["--from", "agentic-python-coder", "ipython_mcp"]
+      "args": ["--python", "3.14", "--from", "agentic-python-coder", "ipython_mcp"]
     }
   }
 }
@@ -99,7 +99,7 @@ coder -i
 | `--version`, `-V` | Show version and exit |
 | `--init [TEMPLATE]` | Initialize example templates (cpmpy, clingo, regex, or all) |
 | `--task`, `-t FILE` | Load task from markdown file |
-| `--model MODEL` | Model name or JSON file (default: sonnet45) |
+| `--model MODEL` | Model name or JSON file (default: sonnet55) |
 | `--project`, `-p FILE` | Project template for domain-specific prompts |
 | `--with PACKAGE` | Add packages dynamically (repeatable) |
 | `--dir`, `-d DIR` | Working directory |
@@ -113,21 +113,25 @@ coder -i
 
 ```bash
 # Built-in models (versioned names)
-coder --model sonnet45 "task"   # Claude Sonnet 4.5 (default)
+coder --model sonnet45 "task"   # Claude Sonnet 4.5
 coder --model sonnet46 "task"   # Claude Sonnet 4.6
 coder --model sonnet5 "task"    # Claude Sonnet 5
 coder --model opus45 "task"     # Claude Opus 4.5
 coder --model opus48 "task"     # Claude Opus 4.8
+coder --model sonnet55 "task"   # Claude Sonnet 5.5 (default)
+coder --model opus55 "task"     # Claude Opus 5.5
 coder --model deepseek31 "task" # DeepSeek v3.1
-coder --model grok41 "task"     # X.AI Grok 4.1
+coder --model grok47 "task"     # X.AI Grok 4.7
 coder --model qwen3 "task"      # Qwen3 Coder
 coder --model gemini25 "task"     # Gemini Pro 2.5
-coder --model gemini3pro "task"    # Gemini 3 Pro Preview
 coder --model gemini31 "task"      # Gemini 3.1 Pro Preview
 coder --model gemini3flash "task"  # Gemini 3 Flash Preview (fast, low-cost)
 coder --model gpt52 "task"         # GPT-5.2
 coder --model gpt56sol "task"      # GPT-5.6 Sol
 coder --model gpt56terra "task"    # GPT-5.6 Terra
+coder --model gpt6sol "task"       # GPT-6 Sol
+coder --model gpt6astra "task"     # GPT-6 Astra
+coder --model gemini4argon "task"  # Gemini 4 Argon
 
 # Custom model (JSON file)
 coder --model ./mymodel.json "task"
@@ -184,7 +188,7 @@ import agentic_python_coder as coder
 messages, stats, log_path = coder.solve_task(
     "Write a fibonacci function",
     working_directory="/tmp/workspace",
-    model="sonnet45",
+    model="sonnet55",
     quiet=True,
 )
 
@@ -203,7 +207,7 @@ from agentic_python_coder import solve_task
 messages, stats, log_path = solve_task(
     task="Your task description",
     working_directory=".",           # Where to run and save files
-    model=None,                      # Model name: "sonnet45", "opus45", or JSON file
+    model=None,                      # Model name: "sonnet55", "opus55", or JSON file
     system_prompt=None,              # Custom system prompt (string)
     system_prompt_path=None,         # Path to system prompt file
     project_prompt=None,             # Domain-specific context
@@ -267,9 +271,9 @@ agent = create_coding_agent(
 ```python
 from agentic_python_coder import get_openrouter_llm, list_available_models
 
-llm = get_openrouter_llm(model="sonnet45")
+llm = get_openrouter_llm(model="sonnet55")
 print(list_available_models())
-# ['deepseek31', 'gemini25', 'gemini3flash', 'gemini3pro', 'gpt52', 'grok41', 'opus45', 'qwen3', 'sonnet45']
+# ['deepseek31', 'gemini25', 'gemini31', 'gemini3flash', 'gpt52', ...]
 ```
 
 ---
@@ -287,7 +291,7 @@ Add to your MCP settings (e.g., `~/.claude/claude_desktop_config.json` or projec
   "mcpServers": {
     "ipython": {
       "command": "uvx",
-      "args": ["--from", "agentic-python-coder", "ipython_mcp"]
+      "args": ["--python", "3.14", "--from", "agentic-python-coder", "ipython_mcp"]
     }
   }
 }

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/coder/src/agentic_python_coder/agent.py

@@ -19,6 +19,55 @@ from agentic_python_coder.tools import (
 # Default maximum number of steps the agent can take before stopping
 DEFAULT_STEP_LIMIT = 200
 
+# Retries for API error responses the OpenAI SDK parses without raising:
+# OpenRouter intermittently returns HTTP 200 whose body is an error object
+# ({"choices": null, "error": {...}}), or a choice with finish_reason "error"
+API_ERROR_MAX_RETRIES = 3
+API_ERROR_BACKOFF_SECONDS = 5
+
+
+def _response_error(response) -> Optional[str]:
+    """Return a description if the response is an error envelope, else None."""
+    if not response.choices:
+        error = (getattr(response, "model_extra", None) or {}).get("error")
+        if error:
+            return f"no choices, error: {error}"
+        dump = getattr(response, "model_dump_json", None)
+        body = dump() if callable(dump) else repr(response)
+        return f"no choices, body: {body[:500]}"
+    choice = response.choices[0]
+    if choice.finish_reason == "error":
+        error = (getattr(choice, "model_extra", None) or {}).get("error") or (
+            getattr(response, "model_extra", None) or {}
+        ).get("error")
+        return f"finish_reason 'error', error: {error}"
+    return None
+
+
+def _create_with_retry(llm_config: LLMConfig, request_kwargs: dict, quiet: bool):
+    """Call the completions API, retrying error responses.
+
+    Raises RuntimeError when retries are exhausted, so callers see a failure
+    instead of a silent early stop reported as success.
+    """
+    detail = ""
+    for attempt in range(API_ERROR_MAX_RETRIES + 1):
+        response = llm_config.client.chat.completions.create(**request_kwargs)
+        detail = _response_error(response)
+        if detail is None:
+            return response
+        if not quiet:
+            print(
+                f"Warning: API returned an error response "
+                f"(attempt {attempt + 1}/{API_ERROR_MAX_RETRIES + 1}); {detail}"
+            )
+        if attempt < API_ERROR_MAX_RETRIES:
+            time.sleep(API_ERROR_BACKOFF_SECONDS * (attempt + 1))
+    raise RuntimeError(
+        f"API returned error responses on {API_ERROR_MAX_RETRIES + 1} attempts; "
+        f"last: {detail}"
+    )
+
 
 def load_prompt(prompt_path: Path) -> str:
     """Load a prompt from file."""
@@ -246,8 +295,8 @@ def run_agent(
             request_kwargs["tools"] = openai_tools
             request_kwargs["tool_choice"] = "auto"
 
-        # Call API
-        response = agent.llm_config.client.chat.completions.create(**request_kwargs)
+        # Call API (retries transient error responses, raises if they persist)
+        response = _create_with_retry(agent.llm_config, request_kwargs, quiet)
 
         # Track tokens (some providers return usage objects with None fields)
         if response.usage:
@@ -261,9 +310,6 @@ def run_agent(
                 response.usage.total_tokens or 0
             )
 
-        if not response.choices:
-            break
-
         assistant_message = response.choices[0].message
 
         # Check for tool calls

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/coder/src/agentic_python_coder/cli.py

@@ -8,6 +8,8 @@ from pathlib import Path
 from typing import Optional, Dict, Any
 from importlib import resources
 
+from openai import AuthenticationError
+
 from agentic_python_coder.runner import (
     solve_task,
     get_system_prompt_path,
@@ -485,6 +487,10 @@ def main():
                 display_statistics(stats)
                 print(f"Log saved to: {log_path}")
 
+    except (ValueError, AuthenticationError) as e:
+        # Configuration errors (missing or invalid API key): no traceback
+        print(f"Error: {e}")
+        sys.exit(1)
     except Exception as e:
         print(f"Error: {e}")
         import traceback

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/coder/src/agentic_python_coder/kernel.py

@@ -11,6 +11,7 @@ import os
 import re
 import shutil
 import signal
+import sys
 import threading
 import time
 import uuid
@@ -101,6 +102,10 @@ class UVKernelManager(KernelManager):
         # Don't load project dependencies (create clean environment)
         uv_prefix.append("--no-project")
 
+        # Match the host's Python; uv's default may be a newer version for
+        # which packages (e.g. ortools) have no wheels yet
+        uv_prefix.extend(["--python", f"{sys.version_info[0]}.{sys.version_info[1]}"])
+
         # Add user-specified packages
         for pkg in self.with_packages:
             uv_prefix.extend(["--with", pkg])
@@ -175,9 +180,8 @@ class PythonKernel:
 
             # Wait for kernel to be ready
             logger.debug("Waiting for kernel to be ready...")
-            self.kc.wait_for_ready(
-                timeout=30
-            )  # Increased timeout for UV package installation
+            # Generous timeout: a cold uv cache or a source build can take minutes
+            self.kc.wait_for_ready(timeout=180)
             logger.info("Kernel started successfully")
 
             # Configure IPython and inject startup code

{agentic_python_coder-3.4.1 → agentic_python_coder-3.4.3}/coder/src/agentic_python_coder/llm.py

@@ -11,7 +11,7 @@ from dotenv import load_dotenv
 from openai import OpenAI
 
 # Default model name (without .json)
-DEFAULT_MODEL = "sonnet45"
+DEFAULT_MODEL = "sonnet55"
 
 
 @dataclass

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/gemini4argon.json

@@ -0,0 +1,6 @@
+{
+  "path": "google/gemini-4-argon",
+  "temperature": 0.0,
+  "max_tokens": 16384,
+  "request_timeout": 120
+}

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/gpt6astra.json

@@ -0,0 +1,5 @@
+{
+  "path": "openai/gpt-6-astra",
+  "max_tokens": 16384,
+  "no_sampling_params": true
+}

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/gpt6sol.json

@@ -0,0 +1,5 @@
+{
+  "path": "openai/gpt-6-sol",
+  "max_tokens": 16384,
+  "no_sampling_params": true
+}

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/grok47.json

@@ -0,0 +1,5 @@
+{
+  "path": "x-ai/grok-4.7",
+  "temperature": 0.0,
+  "max_tokens": 16384
+}

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/opus55.json

@@ -0,0 +1,5 @@
+{
+  "path": "anthropic/claude-opus-5.5",
+  "max_tokens": 32768,
+  "no_sampling_params": true
+}

agentic_python_coder-3.4.3/coder/src/agentic_python_coder/models/sonnet55.json

@@ -0,0 +1,5 @@
+{
+  "path": "anthropic/claude-sonnet-5.5",
+  "max_tokens": 32768,
+  "no_sampling_params": true
+}

agentic_python_coder-3.4.3/coder/tests/test_kernel.py

@@ -0,0 +1,212 @@
+"""Tests for the IPython kernel module."""
+
+import pytest
+from agentic_python_coder.kernel import (
+    PythonKernel,
+    get_kernel,
+    shutdown_kernel,
+    format_output,
+)
+
+
+@pytest.fixture
+def kernel():
+    """Create a fresh kernel for each test."""
+    # Shutdown any existing global kernel
+    shutdown_kernel()
+    # Create a new kernel
+    k = PythonKernel()
+    yield k
+    # Cleanup
+    k.shutdown()
+    shutdown_kernel()
+
+
+class TestKernelBasics:
+    """Test basic kernel execution."""
+
+    def test_simple_expression(self, kernel):
+        """Basic expression returns result."""
+        output = kernel.execute("2 + 2")
+        assert output["error"] is None
+        assert output["result"] == "4"
+
+    def test_print_output(self, kernel):
+        """Print statements captured in stdout."""
+        output = kernel.execute("print('hello world')")
+        assert output["error"] is None
+        assert "hello world" in output["stdout"]
+
+    def test_multiline_code(self, kernel):
+        """Multiline code executes correctly."""
+        code = """
+x = 10
+y = 20
+x + y
+"""
+        output = kernel.execute(code)
+        assert output["error"] is None
+        assert output["result"] == "30"
+
+    def test_syntax_error(self, kernel):
+        """Syntax errors are captured."""
+        output = kernel.execute("def (")
+        assert output["error"] is not None
+        assert "SyntaxError" in output["error"]
+
+    def test_runtime_error(self, kernel):
+        """Runtime errors are captured."""
+        output = kernel.execute("1 / 0")
+        assert output["error"] is not None
+        assert "ZeroDivisionError" in output["error"]
+
+    def test_stderr_capture(self, kernel):
+        """Stderr is captured."""
+        output = kernel.execute("import sys; print('error', file=sys.stderr)")
+        assert "error" in output["stderr"]
+
+
+class TestKernelStatePersistence:
+    """Test that state persists across executions."""
+
+    def test_variable_persistence(self, kernel):
+        """Variables persist between calls."""
+        kernel.execute("x = 42")
+        output = kernel.execute("x * 2")
+        assert output["result"] == "84"
+
+    def test_function_persistence(self, kernel):
+        """Functions persist between calls."""
+        kernel.execute("def double(n): return n * 2")
+        output = kernel.execute("double(21)")
+        assert output["result"] == "42"
+
+    def test_import_persistence(self, kernel):
+        """Imports persist between calls."""
+        kernel.execute("import math")
+        output = kernel.execute("math.pi")
+        assert "3.14" in output["result"]
+
+    def test_class_persistence(self, kernel):
+        """Classes persist between calls."""
+        kernel.execute("class Counter:\n    def __init__(self):\n        self.n = 0")
+        kernel.execute("c = Counter()")
+        kernel.execute("c.n = 5")
+        output = kernel.execute("c.n")
+        assert output["result"] == "5"
+
+
+class TestMessageLoopFix:
+    """Test that message loop waits for idle status."""
+
+    def test_slow_print_captured(self, kernel):
+        """Slow output is fully captured (waits for idle, not Empty)."""
+        # This used to break early on Empty before the fix
+        code = """
+import time
+for i in range(5):
+    print(f"line {i}")
+    time.sleep(0.1)
+print("done")
+"""
+        output = kernel.execute(code)
+        assert output["error"] is None
+        assert "line 0" in output["stdout"]
+        assert "line 4" in output["stdout"]
+        assert "done" in output["stdout"]
+
+    def test_delayed_result(self, kernel):
+        """Result from delayed computation is captured."""
+        code = """
+import time
+time.sleep(0.5)
+42 * 2
+"""
+        output = kernel.execute(code)
+        assert output["error"] is None
+        assert output["result"] == "84"
+
+
+class TestFormatOutput:
+    """Test output formatting."""
+
+    def test_success_format(self):
+        """Successful output is formatted correctly."""
+        output = {"stdout": "hello", "stderr": "", "result": "42", "error": None}
+        result = format_output(output)
+        assert '"success": true' in result
+        assert '"stdout": "hello"' in result
+        assert '"result": "42"' in result
+
+    def test_error_format(self):
+        """Error output is formatted correctly."""
+        output = {
+            "stdout": "",
+            "stderr": "",
+            "result": None,
+            "error": "ValueError: bad",
+        }
+        result = format_output(output)
+        assert '"success": false' in result
+        assert "ValueError" in result
+
+    def test_none_values_removed(self):
+        """None values are removed from output."""
+        output = {"stdout": "", "stderr": None, "result": None, "error": None}
+        result = format_output(output)
+        assert "stderr" not in result
+        assert "result" not in result
+
+
+class TestGlobalKernel:
+    """Test global kernel management."""
+
+    def test_get_kernel_creates_kernel(self):
+        """get_kernel creates a new kernel if none exists."""
+        shutdown_kernel()  # Ensure clean state
+        k = get_kernel()
+        assert k is not None
+        output = k.execute("1 + 1")
+        assert output["result"] == "2"
+        shutdown_kernel()
+
+    def test_get_kernel_reuses_kernel(self):
+        """get_kernel reuses existing kernel with same config."""
+        shutdown_kernel()
+        k1 = get_kernel()
+        k1.execute("x = 123")
+        k2 = get_kernel()
+        output = k2.execute("x")
+        assert output["result"] == "123"
+        shutdown_kernel()
+
+    def test_shutdown_clears_state(self):
+        """shutdown_kernel clears the global kernel."""
+        shutdown_kernel()
+        k = get_kernel()
+        k.execute("y = 456")
+        shutdown_kernel()
+        k2 = get_kernel()
+        output = k2.execute("y")
+        assert output["error"] is not None
+        assert "NameError" in output["error"]
+        shutdown_kernel()
+
+
+class TestDeadlineTimeout:
+    """Deadline overrun must report a timeout error, not silent success."""
+
+    def test_timeout_reports_error_and_interrupts(self, kernel):
+        output = kernel.execute("import time; time.sleep(30)", deadline_timeout=2)
+        assert output["error"] is not None
+        assert "timed out" in output["error"]
+
+        # The interrupt (plus grace period) must leave the kernel usable
+        output2 = kernel.execute("21 * 2")
+        assert output2["error"] is None
+        assert output2["result"] == "42"
+
+    def test_fast_code_unaffected_by_deadline(self, kernel):
+        output = kernel.execute("1 + 1", deadline_timeout=30)
+        assert output["error"] is None
+        assert output["result"] == "2"