stochast 0.1.0__tar.gz

stochast-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,20 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v3
+        with:
+          version: "latest"
+      - run: uv sync --python 3.11
+      - run: uv run ruff check .
+      - run: uv run ruff format --check .
+      - run: uv run mypy stochast
+      - run: uv run pytest

stochast-0.1.0/.gitignore

@@ -0,0 +1,218 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#   Usually these files are written by a python script from a template
+#   before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+# Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+# uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+# poetry.lock
+# poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+# pdm.lock
+# pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+# pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# Redis
+*.rdb
+*.aof
+*.pid
+
+# RabbitMQ
+mnesia/
+rabbitmq/
+rabbitmq-data/
+
+# ActiveMQ
+activemq-data/
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#   JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#   be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#   and can be added to the global gitignore or merged into this file.  For a more nuclear
+#   option (not recommended) you can uncomment the following to ignore the entire idea folder.
+# .idea/
+
+# Abstra
+#   Abstra is an AI-powered process automation framework.
+#   Ignore directories containing user credentials, local state, and settings.
+#   Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#   Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#   that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#   and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#   you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+# Temporary file for partial code execution
+tempCodeRunnerFile.py
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# Streamlit
+.streamlit/secrets.toml

stochast-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Victor Hugo Braz
+
+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.

stochast-0.1.0/PKG-INFO

@@ -0,0 +1,89 @@
+Metadata-Version: 2.5
+Name: stochast
+Version: 0.1.0
+Summary: A testing tool for LLM agents that reports statistics instead of verdicts.
+License-Expression: MIT
+License-File: LICENSE
+Requires-Python: >=3.11
+Requires-Dist: httpx>=0.27
+Requires-Dist: pydantic>=2.0
+Requires-Dist: rich>=13.0
+Requires-Dist: typer>=0.12
+Description-Content-Type: text/markdown
+
+# Stochast
+
+A testing tool for LLM agents that reports pass rates instead of pass/fail verdicts. Agents are
+non-deterministic: the same prompt can take a different tool-call path on every run, so a single
+pass or fail tells you almost nothing. Stochast runs a scenario N times and reports how often it
+passed.
+
+```python
+# scenarios.py
+from stochast import scenario, expect
+from stochast.adapters.openai import OpenAIAdapter, ToolSpec
+
+
+def lookup_order(order_id: int) -> dict:
+    return {"order_id": order_id, "status": "shipped"}
+
+
+def build_adapter():
+    return OpenAIAdapter(
+        model="gpt-4o-mini",
+        api_key="sk-...",
+        tools=[
+            ToolSpec(
+                name="lookup_order",
+                description="Look up an order by id",
+                parameters={
+                    "type": "object",
+                    "properties": {"order_id": {"type": "integer"}},
+                    "required": ["order_id"],
+                },
+                handler=lookup_order,
+            )
+        ],
+    )
+
+
+@scenario(runs=20)
+def refund_status_lookup(agent):
+    result = agent.run("What's the status of order 4471?")
+    expect.tool_called(result, "lookup_order")
+    expect.output_contains(result, "4471")
+```
+
+```
+stochast run scenarios.py --adapter scenarios:build_adapter
+```
+
+```
+refund_status_lookup (20 runs)
+  pass rate: 18/20 (90%)
+```
+
+Every run is persisted as JSON under `stochast-results/` for later inspection.
+
+## Retry policy
+
+Stochast retries transport errors (timeouts, connection failures, 429s, 5xxs) with backoff,
+because those are infrastructure problems. It never retries anything else: a model calling the
+wrong tool or giving a bad answer is data, and retrying it would silently destroy the measurement
+you're trying to take.
+
+## Status
+
+Early and incomplete. Currently implemented: the `@scenario` decorator, an OpenAI-compatible
+tool-calling adapter, a concurrent runner with the retry policy above and Ctrl-C-safe partial
+results, and two assertions (`tool_called`, `output_contains`). Confidence intervals, the full
+assertion vocabulary, tool-path frequency tables, cost/latency percentiles, and A/B comparison are
+planned but not yet built.
+
+## Install
+
+```
+pip install stochast
+```
+
+Requires Python 3.11+.

stochast-0.1.0/README.md

@@ -0,0 +1,76 @@
+# Stochast
+
+A testing tool for LLM agents that reports pass rates instead of pass/fail verdicts. Agents are
+non-deterministic: the same prompt can take a different tool-call path on every run, so a single
+pass or fail tells you almost nothing. Stochast runs a scenario N times and reports how often it
+passed.
+
+```python
+# scenarios.py
+from stochast import scenario, expect
+from stochast.adapters.openai import OpenAIAdapter, ToolSpec
+
+
+def lookup_order(order_id: int) -> dict:
+    return {"order_id": order_id, "status": "shipped"}
+
+
+def build_adapter():
+    return OpenAIAdapter(
+        model="gpt-4o-mini",
+        api_key="sk-...",
+        tools=[
+            ToolSpec(
+                name="lookup_order",
+                description="Look up an order by id",
+                parameters={
+                    "type": "object",
+                    "properties": {"order_id": {"type": "integer"}},
+                    "required": ["order_id"],
+                },
+                handler=lookup_order,
+            )
+        ],
+    )
+
+
+@scenario(runs=20)
+def refund_status_lookup(agent):
+    result = agent.run("What's the status of order 4471?")
+    expect.tool_called(result, "lookup_order")
+    expect.output_contains(result, "4471")
+```
+
+```
+stochast run scenarios.py --adapter scenarios:build_adapter
+```
+
+```
+refund_status_lookup (20 runs)
+  pass rate: 18/20 (90%)
+```
+
+Every run is persisted as JSON under `stochast-results/` for later inspection.
+
+## Retry policy
+
+Stochast retries transport errors (timeouts, connection failures, 429s, 5xxs) with backoff,
+because those are infrastructure problems. It never retries anything else: a model calling the
+wrong tool or giving a bad answer is data, and retrying it would silently destroy the measurement
+you're trying to take.
+
+## Status
+
+Early and incomplete. Currently implemented: the `@scenario` decorator, an OpenAI-compatible
+tool-calling adapter, a concurrent runner with the retry policy above and Ctrl-C-safe partial
+results, and two assertions (`tool_called`, `output_contains`). Confidence intervals, the full
+assertion vocabulary, tool-path frequency tables, cost/latency percentiles, and A/B comparison are
+planned but not yet built.
+
+## Install
+
+```
+pip install stochast
+```
+
+Requires Python 3.11+.

stochast-0.1.0/pyproject.toml

@@ -0,0 +1,47 @@
+[project]
+name = "stochast"
+version = "0.1.0"
+description = "A testing tool for LLM agents that reports statistics instead of verdicts."
+readme = "README.md"
+license = "MIT"
+requires-python = ">=3.11"
+dependencies = [
+    "pydantic>=2.0",
+    "typer>=0.12",
+    "rich>=13.0",
+    "httpx>=0.27",
+]
+
+[project.scripts]
+stochast = "stochast.cli:app"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["stochast"]
+
+[dependency-groups]
+dev = [
+    "pytest>=8.0",
+    "ruff>=0.6",
+    "mypy>=1.10",
+]
+
+[tool.ruff]
+line-length = 100
+target-version = "py311"
+
+[tool.ruff.lint]
+select = ["E", "F", "I", "UP", "B"]
+
+[tool.ruff.lint.flake8-bugbear]
+extend-immutable-calls = ["typer.Argument", "typer.Option"]
+
+[tool.mypy]
+python_version = "3.11"
+strict = true
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]

stochast-0.1.0/stochast/__init__.py

@@ -0,0 +1,4 @@
+from stochast import expect
+from stochast.scenario import scenario
+
+__all__ = ["scenario", "expect"]

stochast-0.1.0/stochast/adapters/__init__.py

@@ -0,0 +1,27 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any, Protocol, runtime_checkable
+
+from stochast.records import ToolCall
+
+
+@dataclass
+class AgentResult:
+    output: str
+    tool_calls: list[ToolCall] = field(default_factory=list)
+    prompt_tokens: int = 0
+    completion_tokens: int = 0
+    raw_messages: list[dict[str, Any]] = field(default_factory=list)
+
+
+@runtime_checkable
+class AgentAdapter(Protocol):
+    def run(self, prompt: str) -> AgentResult: ...
+
+
+# Raised by an adapter for infrastructure failures (timeouts, connection
+# errors, 429/5xx responses). The runner retries these; every other
+# exception is treated as model behaviour and is never retried.
+class TransportError(Exception):
+    pass

stochast-0.1.0/stochast/adapters/openai.py

@@ -0,0 +1,153 @@
+from __future__ import annotations
+
+import json
+import time
+from collections.abc import Callable
+from dataclasses import dataclass
+from typing import Any, cast
+
+import httpx
+
+from stochast.adapters import AgentResult, TransportError
+from stochast.records import ToolCall
+
+
+@dataclass
+class ToolSpec:
+    name: str
+    description: str
+    parameters: dict[str, Any]
+    handler: Callable[..., Any]
+
+    # Renders this tool in the OpenAI chat-completions `tools` wire format.
+    def to_openai_format(self) -> dict[str, Any]:
+        return {
+            "type": "function",
+            "function": {
+                "name": self.name,
+                "description": self.description,
+                "parameters": self.parameters,
+            },
+        }
+
+
+class OpenAIAdapter:
+    def __init__(
+        self,
+        model: str,
+        tools: list[ToolSpec] | None = None,
+        system_prompt: str | None = None,
+        api_key: str | None = None,
+        base_url: str = "https://api.openai.com/v1",
+        max_tool_iterations: int = 10,
+        seed: int | None = None,
+        timeout: float = 30.0,
+        client: httpx.Client | None = None,
+    ) -> None:
+        self.model = model
+        self.tools = tools or []
+        self.system_prompt = system_prompt
+        self.api_key = api_key
+        self.base_url = base_url.rstrip("/")
+        self.max_tool_iterations = max_tool_iterations
+        self.seed = seed
+        self._handlers = {tool.name: tool.handler for tool in self.tools}
+        self._client = client or httpx.Client(timeout=timeout)
+
+    def close(self) -> None:
+        self._client.close()
+
+    # Drives the chat-completions tool-calling loop for a single prompt,
+    # executing any requested tools locally, until the model answers
+    # without requesting further tool calls or max_tool_iterations is hit.
+    def run(self, prompt: str) -> AgentResult:
+        messages: list[dict[str, Any]] = []
+        if self.system_prompt:
+            messages.append({"role": "system", "content": self.system_prompt})
+        messages.append({"role": "user", "content": prompt})
+
+        tool_calls: list[ToolCall] = []
+        prompt_tokens = 0
+        completion_tokens = 0
+
+        for _ in range(self.max_tool_iterations):
+            response = self._complete(messages)
+            usage = response.get("usage") or {}
+            prompt_tokens += usage.get("prompt_tokens", 0)
+            completion_tokens += usage.get("completion_tokens", 0)
+
+            message = response["choices"][0]["message"]
+            messages.append(message)
+
+            requested_calls = message.get("tool_calls") or []
+            if not requested_calls:
+                return AgentResult(
+                    output=message.get("content") or "",
+                    tool_calls=tool_calls,
+                    prompt_tokens=prompt_tokens,
+                    completion_tokens=completion_tokens,
+                    raw_messages=messages,
+                )
+
+            for call in requested_calls:
+                tool_calls.append(self._execute_tool_call(call, messages))
+
+        return AgentResult(
+            output=messages[-1].get("content") or "",
+            tool_calls=tool_calls,
+            prompt_tokens=prompt_tokens,
+            completion_tokens=completion_tokens,
+            raw_messages=messages,
+        )
+
+    # Sends one chat-completions request, translating network-level and
+    # 429/5xx failures into TransportError so the runner knows to retry.
+    def _complete(self, messages: list[dict[str, Any]]) -> dict[str, Any]:
+        payload: dict[str, Any] = {"model": self.model, "messages": messages}
+        if self.tools:
+            payload["tools"] = [tool.to_openai_format() for tool in self.tools]
+        if self.seed is not None:
+            payload["seed"] = self.seed
+
+        try:
+            response = self._client.post(
+                f"{self.base_url}/chat/completions",
+                json=payload,
+                headers={"Authorization": f"Bearer {self.api_key}"},
+            )
+        except httpx.TransportError as exc:
+            raise TransportError(str(exc)) from exc
+
+        if response.status_code == 429 or response.status_code >= 500:
+            raise TransportError(f"{response.status_code}: {response.text}")
+        response.raise_for_status()
+        return cast(dict[str, Any], response.json())
+
+    # Runs one model-requested tool call locally and records its outcome,
+    # appending the tool's response (or its error) back into the transcript.
+    def _execute_tool_call(self, call: dict[str, Any], messages: list[dict[str, Any]]) -> ToolCall:
+        name = call["function"]["name"]
+        arguments = json.loads(call["function"]["arguments"] or "{}")
+        handler = self._handlers.get(name)
+
+        start = time.monotonic()
+        error: str | None = None
+        result: Any = None
+        try:
+            if handler is None:
+                raise LookupError(f"no handler registered for tool {name!r}")
+            result = handler(**arguments)
+        except Exception as exc:
+            error = str(exc)
+        duration_ms = (time.monotonic() - start) * 1000
+
+        messages.append(
+            {
+                "role": "tool",
+                "tool_call_id": call["id"],
+                "content": f"Error: {error}" if error else json.dumps(result),
+            }
+        )
+        return ToolCall(
+            name=name, arguments=arguments, result=result, error=error, duration_ms=duration_ms
+        )