prompture 0.0.47.dev1__tar.gz → 0.0.47.dev2__tar.gz

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: prompture
-Version: 0.0.47.dev1
+Version: 0.0.47.dev2
 Summary: Ask LLMs to return structured JSON and run cross-model tests. API-first.
 Author-email: Juan Denis <juan@vene.co>
 License-Expression: MIT
@@ -83,7 +83,7 @@ print(person.name)  # Maria
 - **Stepwise extraction** — Per-field prompts with smart type coercion (shorthand numbers, multilingual booleans, dates)
 - **Field registry** — 50+ predefined extraction fields with template variables and Pydantic integration
 - **Conversations** — Stateful multi-turn sessions with sync and async support
-- **Tool use** — Function calling and streaming across supported providers
+- **Tool use** — Function calling and streaming across supported providers, with automatic prompt-based simulation for models without native tool support
 - **Caching** — Built-in response cache with memory, SQLite, and Redis backends
 - **Plugin system** — Register custom drivers via entry points
 - **Usage tracking** — Token counts and cost calculation on every call
@@ -296,6 +296,39 @@ response = conv.send("What is the capital of France?")
 follow_up = conv.send("What about Germany?")  # retains context
 ```
 
+### Tool Use
+
+Register Python functions as tools the LLM can call during a conversation:
+
+```python
+from prompture import Conversation, ToolRegistry
+
+registry = ToolRegistry()
+
+@registry.tool
+def get_weather(city: str, units: str = "celsius") -> str:
+    """Get the current weather for a city."""
+    return f"Weather in {city}: 22 {units}"
+
+conv = Conversation("openai/gpt-4", tools=registry)
+result = conv.ask("What's the weather in London?")
+```
+
+For models without native function calling (Ollama, LM Studio, etc.), Prompture automatically simulates tool use by describing tools in the prompt and parsing structured JSON responses:
+
+```python
+# Auto-detect: uses native tool calling if available, simulation otherwise
+conv = Conversation("ollama/llama3.1:8b", tools=registry, simulated_tools="auto")
+
+# Force simulation even on capable models
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=True)
+
+# Disable tool use entirely
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=False)
+```
+
+The simulation loop describes tools in the system prompt, asks the model to respond with JSON (`tool_call` or `final_answer`), executes tools, and feeds results back — all transparent to the caller.
+
 ### Model Discovery
 
 Auto-detect available models from configured providers:

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/README.md

@@ -36,7 +36,7 @@ print(person.name)  # Maria
 - **Stepwise extraction** — Per-field prompts with smart type coercion (shorthand numbers, multilingual booleans, dates)
 - **Field registry** — 50+ predefined extraction fields with template variables and Pydantic integration
 - **Conversations** — Stateful multi-turn sessions with sync and async support
-- **Tool use** — Function calling and streaming across supported providers
+- **Tool use** — Function calling and streaming across supported providers, with automatic prompt-based simulation for models without native tool support
 - **Caching** — Built-in response cache with memory, SQLite, and Redis backends
 - **Plugin system** — Register custom drivers via entry points
 - **Usage tracking** — Token counts and cost calculation on every call
@@ -249,6 +249,39 @@ response = conv.send("What is the capital of France?")
 follow_up = conv.send("What about Germany?")  # retains context
 ```
 
+### Tool Use
+
+Register Python functions as tools the LLM can call during a conversation:
+
+```python
+from prompture import Conversation, ToolRegistry
+
+registry = ToolRegistry()
+
+@registry.tool
+def get_weather(city: str, units: str = "celsius") -> str:
+    """Get the current weather for a city."""
+    return f"Weather in {city}: 22 {units}"
+
+conv = Conversation("openai/gpt-4", tools=registry)
+result = conv.ask("What's the weather in London?")
+```
+
+For models without native function calling (Ollama, LM Studio, etc.), Prompture automatically simulates tool use by describing tools in the prompt and parsing structured JSON responses:
+
+```python
+# Auto-detect: uses native tool calling if available, simulation otherwise
+conv = Conversation("ollama/llama3.1:8b", tools=registry, simulated_tools="auto")
+
+# Force simulation even on capable models
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=True)
+
+# Disable tool use entirely
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=False)
+```
+
+The simulation loop describes tools in the system prompt, asks the model to respond with JSON (`tool_call` or `final_answer`), executes tools, and feeds results back — all transparent to the caller.
+
 ### Model Discovery
 
 Auto-detect available models from configured providers:

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.0.47.dev1'
-__version_tuple__ = version_tuple = (0, 0, 47, 'dev1')
+__version__ = version = '0.0.47.dev2'
+__version_tuple__ = version_tuple = (0, 0, 47, 'dev2')
 
 __commit_id__ = commit_id = None

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/async_conversation.py

@@ -55,6 +55,7 @@ class AsyncConversation:
         callbacks: DriverCallbacks | None = None,
         tools: ToolRegistry | None = None,
         max_tool_rounds: int = 10,
+        simulated_tools: bool | Literal["auto"] = "auto",
         conversation_id: str | None = None,
         auto_save: str | Path | None = None,
         tags: list[str] | None = None,
@@ -106,6 +107,7 @@ class AsyncConversation:
         }
         self._tools = tools or ToolRegistry()
         self._max_tool_rounds = max_tool_rounds
+        self._simulated_tools = simulated_tools
 
         # Persistence
         self._conversation_id = conversation_id or str(uuid.uuid4())
@@ -324,8 +326,13 @@ class AsyncConversation:
         If tools are registered and the driver supports tool use,
         dispatches to the async tool execution loop.
         """
-        if self._tools and getattr(self._driver, "supports_tool_use", False):
-            return await self._ask_with_tools(content, options, images=images)
+        # Route to appropriate tool handling
+        if self._tools:
+            use_native = getattr(self._driver, "supports_tool_use", False)
+            if self._simulated_tools is True or (self._simulated_tools == "auto" and not use_native):
+                return await self._ask_with_simulated_tools(content, options, images=images)
+            elif use_native and self._simulated_tools is not True:
+                return await self._ask_with_tools(content, options, images=images)
 
         merged = {**self._options, **(options or {})}
         messages = self._build_messages(content, images=images)
@@ -377,6 +384,11 @@ class AsyncConversation:
                 }
                 for tc in tool_calls
             ]
+            # Preserve reasoning_content for providers that require it
+            # on subsequent requests (e.g. Moonshot reasoning models).
+            if resp.get("reasoning_content") is not None:
+                assistant_msg["reasoning_content"] = resp["reasoning_content"]
+
             self._messages.append(assistant_msg)
             msgs.append(assistant_msg)
 
@@ -397,6 +409,63 @@ class AsyncConversation:
 
         raise RuntimeError(f"Tool execution loop exceeded {self._max_tool_rounds} rounds")
 
+    async def _ask_with_simulated_tools(
+        self,
+        content: str,
+        options: dict[str, Any] | None = None,
+        images: list[ImageInput] | None = None,
+    ) -> str:
+        """Async prompt-based tool calling for drivers without native tool use."""
+        from .simulated_tools import build_tool_prompt, format_tool_result, parse_simulated_response
+
+        merged = {**self._options, **(options or {})}
+        tool_prompt = build_tool_prompt(self._tools)
+
+        # Augment system prompt with tool descriptions
+        augmented_system = tool_prompt
+        if self._system_prompt:
+            augmented_system = f"{self._system_prompt}\n\n{tool_prompt}"
+
+        # Record user message in history
+        user_content = self._build_content_with_images(content, images)
+        self._messages.append({"role": "user", "content": user_content})
+
+        for _round in range(self._max_tool_rounds):
+            # Build messages with the augmented system prompt
+            msgs: list[dict[str, Any]] = []
+            msgs.append({"role": "system", "content": augmented_system})
+            msgs.extend(self._messages)
+
+            resp = await self._driver.generate_messages_with_hooks(msgs, merged)
+            text = resp.get("text", "")
+            meta = resp.get("meta", {})
+            self._accumulate_usage(meta)
+
+            parsed = parse_simulated_response(text, self._tools)
+
+            if parsed["type"] == "final_answer":
+                answer = parsed["content"]
+                self._messages.append({"role": "assistant", "content": answer})
+                return answer
+
+            # Tool call
+            tool_name = parsed["name"]
+            tool_args = parsed["arguments"]
+
+            # Record assistant's tool call as an assistant message
+            self._messages.append({"role": "assistant", "content": text})
+
+            try:
+                result = self._tools.execute(tool_name, tool_args)
+                result_msg = format_tool_result(tool_name, result)
+            except Exception as exc:
+                result_msg = format_tool_result(tool_name, f"Error: {exc}")
+
+            # Record tool result as a user message
+            self._messages.append({"role": "user", "content": result_msg})
+
+        raise RuntimeError(f"Simulated tool execution loop exceeded {self._max_tool_rounds} rounds")
+
     def _build_messages_raw(self) -> list[dict[str, Any]]:
         """Build messages array from system prompt + full history (including tool messages)."""
         msgs: list[dict[str, Any]] = []

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/conversation.py

@@ -56,6 +56,7 @@ class Conversation:
         callbacks: DriverCallbacks | None = None,
         tools: ToolRegistry | None = None,
         max_tool_rounds: int = 10,
+        simulated_tools: bool | Literal["auto"] = "auto",
         conversation_id: str | None = None,
         auto_save: str | Path | None = None,
         tags: list[str] | None = None,
@@ -109,6 +110,7 @@ class Conversation:
         }
         self._tools = tools or ToolRegistry()
         self._max_tool_rounds = max_tool_rounds
+        self._simulated_tools = simulated_tools
 
         # Persistence
         self._conversation_id = conversation_id or str(uuid.uuid4())
@@ -338,8 +340,13 @@ class Conversation:
             images: Optional list of images to include (bytes, path, URL,
                 base64 string, or :class:`ImageContent`).
         """
-        if self._tools and getattr(self._driver, "supports_tool_use", False):
-            return self._ask_with_tools(content, options, images=images)
+        # Route to appropriate tool handling
+        if self._tools:
+            use_native = getattr(self._driver, "supports_tool_use", False)
+            if self._simulated_tools is True or (self._simulated_tools == "auto" and not use_native):
+                return self._ask_with_simulated_tools(content, options, images=images)
+            elif use_native and self._simulated_tools is not True:
+                return self._ask_with_tools(content, options, images=images)
 
         merged = {**self._options, **(options or {})}
         messages = self._build_messages(content, images=images)
@@ -395,6 +402,11 @@ class Conversation:
                 }
                 for tc in tool_calls
             ]
+            # Preserve reasoning_content for providers that require it
+            # on subsequent requests (e.g. Moonshot reasoning models).
+            if resp.get("reasoning_content") is not None:
+                assistant_msg["reasoning_content"] = resp["reasoning_content"]
+
             self._messages.append(assistant_msg)
             msgs.append(assistant_msg)
 
@@ -416,6 +428,63 @@ class Conversation:
 
         raise RuntimeError(f"Tool execution loop exceeded {self._max_tool_rounds} rounds")
 
+    def _ask_with_simulated_tools(
+        self,
+        content: str,
+        options: dict[str, Any] | None = None,
+        images: list[ImageInput] | None = None,
+    ) -> str:
+        """Prompt-based tool calling for drivers without native tool use."""
+        from .simulated_tools import build_tool_prompt, format_tool_result, parse_simulated_response
+
+        merged = {**self._options, **(options or {})}
+        tool_prompt = build_tool_prompt(self._tools)
+
+        # Augment system prompt with tool descriptions
+        augmented_system = tool_prompt
+        if self._system_prompt:
+            augmented_system = f"{self._system_prompt}\n\n{tool_prompt}"
+
+        # Record user message in history
+        user_content = self._build_content_with_images(content, images)
+        self._messages.append({"role": "user", "content": user_content})
+
+        for _round in range(self._max_tool_rounds):
+            # Build messages with the augmented system prompt
+            msgs: list[dict[str, Any]] = []
+            msgs.append({"role": "system", "content": augmented_system})
+            msgs.extend(self._messages)
+
+            resp = self._driver.generate_messages_with_hooks(msgs, merged)
+            text = resp.get("text", "")
+            meta = resp.get("meta", {})
+            self._accumulate_usage(meta)
+
+            parsed = parse_simulated_response(text, self._tools)
+
+            if parsed["type"] == "final_answer":
+                answer = parsed["content"]
+                self._messages.append({"role": "assistant", "content": answer})
+                return answer
+
+            # Tool call
+            tool_name = parsed["name"]
+            tool_args = parsed["arguments"]
+
+            # Record assistant's tool call as an assistant message
+            self._messages.append({"role": "assistant", "content": text})
+
+            try:
+                result = self._tools.execute(tool_name, tool_args)
+                result_msg = format_tool_result(tool_name, result)
+            except Exception as exc:
+                result_msg = format_tool_result(tool_name, f"Error: {exc}")
+
+            # Record tool result as a user message (all drivers understand user/assistant)
+            self._messages.append({"role": "user", "content": result_msg})
+
+        raise RuntimeError(f"Simulated tool execution loop exceeded {self._max_tool_rounds} rounds")
+
     def _build_messages_raw(self) -> list[dict[str, Any]]:
         """Build messages array from system prompt + full history (including tool messages)."""
         msgs: list[dict[str, Any]] = []

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/drivers/async_moonshot_driver.py

@@ -271,11 +271,12 @@ class AsyncMoonshotDriver(CostMixin, AsyncDriver):
         }
 
         choice = resp["choices"][0]
-        text = choice["message"].get("content") or ""
+        message = choice["message"]
+        text = message.get("content") or ""
         stop_reason = choice.get("finish_reason")
 
         tool_calls_out: list[dict[str, Any]] = []
-        for tc in choice["message"].get("tool_calls", []):
+        for tc in message.get("tool_calls", []):
             try:
                 args = json.loads(tc["function"]["arguments"])
             except (json.JSONDecodeError, TypeError):
@@ -288,13 +289,21 @@ class AsyncMoonshotDriver(CostMixin, AsyncDriver):
                 }
             )
 
-        return {
+        result: dict[str, Any] = {
             "text": text,
             "meta": meta,
             "tool_calls": tool_calls_out,
             "stop_reason": stop_reason,
         }
 
+        # Preserve reasoning_content for reasoning models so the
+        # conversation loop can include it when sending the assistant
+        # message back (Moonshot requires it on subsequent requests).
+        if message.get("reasoning_content") is not None:
+            result["reasoning_content"] = message["reasoning_content"]
+
+        return result
+
     # ------------------------------------------------------------------
     # Streaming
     # ------------------------------------------------------------------

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/drivers/moonshot_driver.py

@@ -364,11 +364,12 @@ class MoonshotDriver(CostMixin, Driver):
         }
 
         choice = resp["choices"][0]
-        text = choice["message"].get("content") or ""
+        message = choice["message"]
+        text = message.get("content") or ""
         stop_reason = choice.get("finish_reason")
 
         tool_calls_out: list[dict[str, Any]] = []
-        for tc in choice["message"].get("tool_calls", []):
+        for tc in message.get("tool_calls", []):
             try:
                 args = json.loads(tc["function"]["arguments"])
             except (json.JSONDecodeError, TypeError):
@@ -381,13 +382,21 @@ class MoonshotDriver(CostMixin, Driver):
                 }
             )
 
-        return {
+        result: dict[str, Any] = {
             "text": text,
             "meta": meta,
             "tool_calls": tool_calls_out,
             "stop_reason": stop_reason,
         }
 
+        # Preserve reasoning_content for reasoning models so the
+        # conversation loop can include it when sending the assistant
+        # message back (Moonshot requires it on subsequent requests).
+        if message.get("reasoning_content") is not None:
+            result["reasoning_content"] = message["reasoning_content"]
+
+        return result
+
     # ------------------------------------------------------------------
     # Streaming
     # ------------------------------------------------------------------

prompture-0.0.47.dev2/prompture/simulated_tools.py

@@ -0,0 +1,115 @@
+"""Prompt-based tool calling for drivers without native tool use support.
+
+When a driver lacks ``supports_tool_use`` the conversation classes can
+fall back to *simulated* tool calling: the available tools are described
+in the system prompt, the model is asked to respond with a structured
+JSON object (either a tool call or a final answer), and Prompture
+parses + dispatches accordingly.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+from typing import Any
+
+from .tools import clean_json_text
+from .tools_schema import ToolRegistry
+
+logger = logging.getLogger("prompture.simulated_tools")
+
+
+def build_tool_prompt(tools: ToolRegistry) -> str:
+    """Build a plain-text prompt section describing all registered tools.
+
+    The returned string should be appended to the system prompt so the
+    model knows which tools are available and how to call them.
+    """
+    lines = [
+        "You have access to the following tools:",
+        "",
+        tools.to_prompt_format(),
+        "",
+        "To use a tool, respond with ONLY a JSON object in this exact format:",
+        '{"type": "tool_call", "name": "<tool_name>", "arguments": {<args>}}',
+        "",
+        "When you have the final answer (after using tools or if no tool is needed), "
+        "respond with ONLY a JSON object in this format:",
+        '{"type": "final_answer", "content": "<your answer>"}',
+        "",
+        "IMPORTANT: Your entire response must be a single JSON object. "
+        "Do not include any other text, markdown, or explanation outside the JSON.",
+    ]
+    return "\n".join(lines)
+
+
+def parse_simulated_response(text: str, tools: ToolRegistry) -> dict[str, Any]:
+    """Parse the model's response into a tool call or final answer dict.
+
+    Returns one of:
+    - ``{"type": "tool_call", "name": str, "arguments": dict}``
+    - ``{"type": "final_answer", "content": str}``
+    """
+    cleaned = clean_json_text(text).strip()
+
+    # Try JSON parse
+    try:
+        obj = json.loads(cleaned)
+    except (json.JSONDecodeError, ValueError):
+        # Non-JSON text → treat as final answer
+        logger.debug("Response is not valid JSON, treating as final answer")
+        return {"type": "final_answer", "content": text.strip()}
+
+    if not isinstance(obj, dict):
+        return {"type": "final_answer", "content": text.strip()}
+
+    # Explicit type discriminator
+    resp_type = obj.get("type")
+
+    if resp_type == "tool_call":
+        return {
+            "type": "tool_call",
+            "name": obj.get("name", ""),
+            "arguments": obj.get("arguments", {}),
+        }
+
+    if resp_type == "final_answer":
+        return {
+            "type": "final_answer",
+            "content": obj.get("content", ""),
+        }
+
+    # Infer type from keys when "type" is missing
+    if "name" in obj and "arguments" in obj:
+        logger.debug("Inferred tool_call from keys (no 'type' field)")
+        return {
+            "type": "tool_call",
+            "name": obj["name"],
+            "arguments": obj.get("arguments", {}),
+        }
+
+    if "content" in obj:
+        logger.debug("Inferred final_answer from keys (no 'type' field)")
+        return {
+            "type": "final_answer",
+            "content": obj["content"],
+        }
+
+    # Unrecognised JSON structure → final answer with the raw text
+    return {"type": "final_answer", "content": text.strip()}
+
+
+def format_tool_result(tool_name: str, result: Any) -> str:
+    """Format a tool execution result as a user message for the next round."""
+    if isinstance(result, str):
+        result_str = result
+    else:
+        try:
+            result_str = json.dumps(result)
+        except (TypeError, ValueError):
+            result_str = str(result)
+
+    return (
+        f"Tool '{tool_name}' returned:\n{result_str}\n\n"
+        "Continue using the JSON format. Either call another tool or provide your final answer."
+    )

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture/tools_schema.py

@@ -109,6 +109,24 @@ class ToolDefinition:
             "input_schema": self.parameters,
         }
 
+    def to_prompt_format(self) -> str:
+        """Plain-text description suitable for prompt-based tool calling."""
+        lines = [f"Tool: {self.name}", f"  Description: {self.description}", "  Parameters:"]
+        props = self.parameters.get("properties", {})
+        required = set(self.parameters.get("required", []))
+        if not props:
+            lines.append("    (none)")
+        else:
+            for pname, pschema in props.items():
+                ptype = pschema.get("type", "string")
+                req_label = "required" if pname in required else "optional"
+                desc = pschema.get("description", "")
+                line = f"    - {pname} ({ptype}, {req_label})"
+                if desc:
+                    line += f": {desc}"
+                lines.append(line)
+        return "\n".join(lines)
+
 
 def tool_from_function(
     fn: Callable[..., Any], *, name: str | None = None, description: str | None = None
@@ -244,6 +262,10 @@ class ToolRegistry:
     def to_anthropic_format(self) -> list[dict[str, Any]]:
         return [td.to_anthropic_format() for td in self._tools.values()]
 
+    def to_prompt_format(self) -> str:
+        """Join all tool descriptions into a single plain-text block."""
+        return "\n\n".join(td.to_prompt_format() for td in self._tools.values())
+
     # ------------------------------------------------------------------
     # Execution
     # ------------------------------------------------------------------

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: prompture
-Version: 0.0.47.dev1
+Version: 0.0.47.dev2
 Summary: Ask LLMs to return structured JSON and run cross-model tests. API-first.
 Author-email: Juan Denis <juan@vene.co>
 License-Expression: MIT
@@ -83,7 +83,7 @@ print(person.name)  # Maria
 - **Stepwise extraction** — Per-field prompts with smart type coercion (shorthand numbers, multilingual booleans, dates)
 - **Field registry** — 50+ predefined extraction fields with template variables and Pydantic integration
 - **Conversations** — Stateful multi-turn sessions with sync and async support
-- **Tool use** — Function calling and streaming across supported providers
+- **Tool use** — Function calling and streaming across supported providers, with automatic prompt-based simulation for models without native tool support
 - **Caching** — Built-in response cache with memory, SQLite, and Redis backends
 - **Plugin system** — Register custom drivers via entry points
 - **Usage tracking** — Token counts and cost calculation on every call
@@ -296,6 +296,39 @@ response = conv.send("What is the capital of France?")
 follow_up = conv.send("What about Germany?")  # retains context
 ```
 
+### Tool Use
+
+Register Python functions as tools the LLM can call during a conversation:
+
+```python
+from prompture import Conversation, ToolRegistry
+
+registry = ToolRegistry()
+
+@registry.tool
+def get_weather(city: str, units: str = "celsius") -> str:
+    """Get the current weather for a city."""
+    return f"Weather in {city}: 22 {units}"
+
+conv = Conversation("openai/gpt-4", tools=registry)
+result = conv.ask("What's the weather in London?")
+```
+
+For models without native function calling (Ollama, LM Studio, etc.), Prompture automatically simulates tool use by describing tools in the prompt and parsing structured JSON responses:
+
+```python
+# Auto-detect: uses native tool calling if available, simulation otherwise
+conv = Conversation("ollama/llama3.1:8b", tools=registry, simulated_tools="auto")
+
+# Force simulation even on capable models
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=True)
+
+# Disable tool use entirely
+conv = Conversation("openai/gpt-4", tools=registry, simulated_tools=False)
+```
+
+The simulation loop describes tools in the system prompt, asks the model to respond with JSON (`tool_call` or `final_answer`), executes tools, and feeds results back — all transparent to the caller.
+
 ### Model Discovery
 
 Auto-detect available models from configured providers:

{prompture-0.0.47.dev1 → prompture-0.0.47.dev2}/prompture.egg-info/SOURCES.txt

@@ -81,6 +81,7 @@ prompture/serialization.py
 prompture/server.py
 prompture/session.py
 prompture/settings.py
+prompture/simulated_tools.py
 prompture/tools.py
 prompture/tools_schema.py
 prompture/validator.py