pydantic-ai-slim 0.0.47__tar.gz → 0.0.49__tar.gz

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-ai-slim
-Version: 0.0.47
+Version: 0.0.49
 Summary: Agent Framework / shim to use Pydantic with LLMs, slim package
 Author-email: Samuel Colvin <samuel@pydantic.dev>
 License-Expression: MIT
@@ -29,7 +29,7 @@ Requires-Dist: exceptiongroup; python_version < '3.11'
 Requires-Dist: griffe>=1.3.2
 Requires-Dist: httpx>=0.27
 Requires-Dist: opentelemetry-api>=1.28.0
-Requires-Dist: pydantic-graph==0.0.47
+Requires-Dist: pydantic-graph==0.0.49
 Requires-Dist: pydantic>=2.10
 Requires-Dist: typing-inspection>=0.4.0
 Provides-Extra: anthropic
@@ -45,7 +45,7 @@ Requires-Dist: cohere>=5.13.11; (platform_system != 'Emscripten') and extra == '
 Provides-Extra: duckduckgo
 Requires-Dist: duckduckgo-search>=7.0.0; extra == 'duckduckgo'
 Provides-Extra: evals
-Requires-Dist: pydantic-evals==0.0.41; extra == 'evals'
+Requires-Dist: pydantic-evals==0.0.49; extra == 'evals'
 Provides-Extra: groq
 Requires-Dist: groq>=0.15.0; extra == 'groq'
 Provides-Extra: logfire

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/_cli.py

@@ -3,19 +3,20 @@ from __future__ import annotations as _annotations
 import argparse
 import asyncio
 import sys
+from asyncio import CancelledError
 from collections.abc import Sequence
 from contextlib import ExitStack
 from datetime import datetime, timezone
 from importlib.metadata import version
 from pathlib import Path
-from typing import cast
+from typing import Any, cast
 
 from typing_inspection.introspection import get_literal_values
 
 from pydantic_ai.agent import Agent
 from pydantic_ai.exceptions import UserError
 from pydantic_ai.messages import ModelMessage, PartDeltaEvent, TextPartDelta
-from pydantic_ai.models import KnownModelName
+from pydantic_ai.models import KnownModelName, infer_model
 
 try:
     import argcomplete
@@ -47,7 +48,7 @@ class SimpleCodeBlock(CodeBlock):
     This avoids a background color which messes up copy-pasting and sets the language name as dim prefix and suffix.
     """
 
-    def __rich_console__(self, console: Console, options: ConsoleOptions) -> RenderResult:  # pragma: no cover
+    def __rich_console__(self, console: Console, options: ConsoleOptions) -> RenderResult:
         code = str(self.text).rstrip()
         yield Text(self.lexer_name, style='dim')
         yield Syntax(code, self.lexer_name, theme=self.theme, background_color='default', word_wrap=True)
@@ -57,7 +58,7 @@ class SimpleCodeBlock(CodeBlock):
 class LeftHeading(Heading):
     """Customised headings in markdown to stop centering and prepend markdown style hashes."""
 
-    def __rich_console__(self, console: Console, options: ConsoleOptions) -> RenderResult:  # pragma: no cover
+    def __rich_console__(self, console: Console, options: ConsoleOptions) -> RenderResult:
         # note we use `Style(bold=True)` not `self.style_name` here to disable underlining which is ugly IMHO
         yield Text(f'{"#" * int(self.tag[1:])} {self.text.plain}', style=Style(bold=True))
 
@@ -68,7 +69,21 @@ Markdown.elements.update(
 )
 
 
-def cli(args_list: Sequence[str] | None = None) -> int:  # noqa: C901  # pragma: no cover
+cli_agent = Agent()
+
+
+@cli_agent.system_prompt
+def cli_system_prompt() -> str:
+    now_utc = datetime.now(timezone.utc)
+    tzinfo = now_utc.astimezone().tzinfo
+    tzname = tzinfo.tzname(now_utc) if tzinfo else ''
+    return f"""\
+Help the user by responding to their request, the output should be concise and always written in markdown.
+The current date and time is {datetime.now()} {tzname}.
+The user is running {sys.platform}."""
+
+
+def cli(args_list: Sequence[str] | None = None) -> int:
     parser = argparse.ArgumentParser(
         prog='pai',
         description=f"""\
@@ -124,18 +139,10 @@ Special prompt:
             console.print(f'  {model}', highlight=False)
         return 0
 
-    now_utc = datetime.now(timezone.utc)
-    tzname = now_utc.astimezone().tzinfo.tzname(now_utc)  # type: ignore
     try:
-        agent = Agent(
-            model=args.model,
-            system_prompt=f"""\
-    Help the user by responding to their request, the output should be concise and always written in markdown.
-    The current date and time is {datetime.now()} {tzname}.
-    The user is running {sys.platform}.""",
-        )
-    except UserError:
-        console.print(f'[red]Invalid model "{args.model}"[/red]')
+        cli_agent.model = infer_model(args.model)
+    except UserError as e:
+        console.print(f'Error initializing [magenta]{args.model}[/magenta]:\n[red]{e}[/red]')
         return 1
 
     stream = not args.no_stream
@@ -148,67 +155,44 @@ Special prompt:
 
     if prompt := cast(str, args.prompt):
         try:
-            asyncio.run(ask_agent(agent, prompt, stream, console, code_theme))
+            asyncio.run(ask_agent(cli_agent, prompt, stream, console, code_theme))
         except KeyboardInterrupt:
             pass
         return 0
 
     history = Path.home() / '.pai-prompt-history.txt'
-    session = PromptSession(history=FileHistory(str(history)))  # type: ignore
+    # doing this instead of `PromptSession[Any](history=` allows mocking of PromptSession in tests
+    session: PromptSession[Any] = PromptSession(history=FileHistory(str(history)))
+    try:
+        return asyncio.run(run_chat(session, stream, cli_agent, console, code_theme))
+    except KeyboardInterrupt:  # pragma: no cover
+        return 0
+
+
+async def run_chat(session: PromptSession[Any], stream: bool, agent: Agent, console: Console, code_theme: str) -> int:
     multiline = False
     messages: list[ModelMessage] = []
 
     while True:
         try:
             auto_suggest = CustomAutoSuggest(['/markdown', '/multiline', '/exit'])
-            text = cast(str, session.prompt('pai ➤ ', auto_suggest=auto_suggest, multiline=multiline))
-        except (KeyboardInterrupt, EOFError):
+            text = await session.prompt_async('pai ➤ ', auto_suggest=auto_suggest, multiline=multiline)
+        except (KeyboardInterrupt, EOFError):  # pragma: no cover
             return 0
 
         if not text.strip():
             continue
 
-        ident_prompt = text.lower().strip(' ').replace(' ', '-').lstrip(' ')
+        ident_prompt = text.lower().strip().replace(' ', '-')
         if ident_prompt.startswith('/'):
-            if ident_prompt == '/markdown':
-                try:
-                    parts = messages[-1].parts
-                except IndexError:
-                    console.print('[dim]No markdown output available.[/dim]')
-                    continue
-                console.print('[dim]Markdown output of last question:[/dim]\n')
-                for part in parts:
-                    if part.part_kind == 'text':
-                        console.print(
-                            Syntax(
-                                part.content,
-                                lexer='markdown',
-                                theme=code_theme,
-                                word_wrap=True,
-                                background_color='default',
-                            )
-                        )
-
-            elif ident_prompt == '/multiline':
-                multiline = not multiline
-                if multiline:
-                    console.print(
-                        'Enabling multiline mode. '
-                        '[dim]Press [Meta+Enter] or [Esc] followed by [Enter] to accept input.[/dim]'
-                    )
-                else:
-                    console.print('Disabling multiline mode.')
-            elif ident_prompt == '/exit':
-                console.print('[dim]Exiting…[/dim]')
-                return 0
-            else:
-                console.print(f'[red]Unknown command[/red] [magenta]`{ident_prompt}`[/magenta]')
+            exit_value, multiline = handle_slash_command(ident_prompt, messages, multiline, console, code_theme)
+            if exit_value is not None:
+                return exit_value
         else:
             try:
-                messages = asyncio.run(ask_agent(agent, text, stream, console, code_theme, messages))
-            except KeyboardInterrupt:
+                messages = await ask_agent(agent, text, stream, console, code_theme, messages)
+            except CancelledError:  # pragma: no cover
                 console.print('[dim]Interrupted[/dim]')
-                messages = []
 
 
 async def ask_agent(
@@ -218,7 +202,7 @@ async def ask_agent(
     console: Console,
     code_theme: str,
     messages: list[ModelMessage] | None = None,
-) -> list[ModelMessage]:  # pragma: no cover
+) -> list[ModelMessage]:
     status = Status('[dim]Working on it…[/dim]', console=console)
 
     if not stream:
@@ -248,7 +232,7 @@ async def ask_agent(
 
 
 class CustomAutoSuggest(AutoSuggestFromHistory):
-    def __init__(self, special_suggestions: list[str] | None = None):  # pragma: no cover
+    def __init__(self, special_suggestions: list[str] | None = None):
         super().__init__()
         self.special_suggestions = special_suggestions or []
 
@@ -264,5 +248,44 @@ class CustomAutoSuggest(AutoSuggestFromHistory):
         return suggestion
 
 
+def handle_slash_command(
+    ident_prompt: str, messages: list[ModelMessage], multiline: bool, console: Console, code_theme: str
+) -> tuple[int | None, bool]:
+    if ident_prompt == '/markdown':
+        try:
+            parts = messages[-1].parts
+        except IndexError:
+            console.print('[dim]No markdown output available.[/dim]')
+        else:
+            console.print('[dim]Markdown output of last question:[/dim]\n')
+            for part in parts:
+                if part.part_kind == 'text':
+                    console.print(
+                        Syntax(
+                            part.content,
+                            lexer='markdown',
+                            theme=code_theme,
+                            word_wrap=True,
+                            background_color='default',
+                        )
+                    )
+
+    elif ident_prompt == '/multiline':
+        multiline = not multiline
+        if multiline:
+            console.print(
+                'Enabling multiline mode. [dim]Press [Meta+Enter] or [Esc] followed by [Enter] to accept input.[/dim]'
+            )
+        else:
+            console.print('Disabling multiline mode.')
+        return None, multiline
+    elif ident_prompt == '/exit':
+        console.print('[dim]Exiting…[/dim]')
+        return 0, multiline
+    else:
+        console.print(f'[red]Unknown command[/red] [magenta]`{ident_prompt}`[/magenta]')
+    return None, multiline
+
+
 def app():  # pragma: no cover
     sys.exit(cli())

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/_result.py

@@ -13,7 +13,7 @@ from typing_inspection.introspection import is_union_origin
 from . import _utils, messages as _messages
 from .exceptions import ModelRetry
 from .result import ResultDataT, ResultDataT_inv, ResultValidatorFunc
-from .tools import AgentDepsT, RunContext, ToolDefinition
+from .tools import AgentDepsT, GenerateToolJsonSchema, RunContext, ToolDefinition
 
 T = TypeVar('T')
 """An invariant TypeVar."""
@@ -159,7 +159,9 @@ class ResultTool(Generic[ResultDataT]):
             self.type_adapter = TypeAdapter(response_type)
             outer_typed_dict_key: str | None = None
             # noinspection PyArgumentList
-            parameters_json_schema = _utils.check_object_json_schema(self.type_adapter.json_schema())
+            parameters_json_schema = _utils.check_object_json_schema(
+                self.type_adapter.json_schema(schema_generator=GenerateToolJsonSchema)
+            )
         else:
             response_data_typed_dict = TypedDict(  # noqa: UP013
                 'response_data_typed_dict',
@@ -168,7 +170,9 @@ class ResultTool(Generic[ResultDataT]):
             self.type_adapter = TypeAdapter(response_data_typed_dict)
             outer_typed_dict_key = 'response'
             # noinspection PyArgumentList
-            parameters_json_schema = _utils.check_object_json_schema(self.type_adapter.json_schema())
+            parameters_json_schema = _utils.check_object_json_schema(
+                self.type_adapter.json_schema(schema_generator=GenerateToolJsonSchema)
+            )
             # including `response_data_typed_dict` as a title here doesn't add anything and could confuse the LLM
             parameters_json_schema.pop('title')
 

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/models/__init__.py

@@ -106,6 +106,7 @@ KnownModelName = TypeAliasType(
         'google-gla:gemini-2.0-flash',
         'google-gla:gemini-2.0-flash-lite-preview-02-05',
         'google-gla:gemini-2.0-pro-exp-02-05',
+        'google-gla:gemini-2.5-pro-exp-03-25',
         'google-vertex:gemini-1.0-pro',
         'google-vertex:gemini-1.5-flash',
         'google-vertex:gemini-1.5-flash-8b',
@@ -116,6 +117,7 @@ KnownModelName = TypeAliasType(
         'google-vertex:gemini-2.0-flash',
         'google-vertex:gemini-2.0-flash-lite-preview-02-05',
         'google-vertex:gemini-2.0-pro-exp-02-05',
+        'google-vertex:gemini-2.5-pro-exp-03-25',
         'gpt-3.5-turbo',
         'gpt-3.5-turbo-0125',
         'gpt-3.5-turbo-0301',

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/models/anthropic.py

@@ -1,6 +1,5 @@
 from __future__ import annotations as _annotations
 
-import base64
 import io
 from collections.abc import AsyncGenerator, AsyncIterable, AsyncIterator
 from contextlib import asynccontextmanager
@@ -9,7 +8,6 @@ from datetime import datetime, timezone
 from json import JSONDecodeError, loads as json_loads
 from typing import Any, Literal, Union, cast, overload
 
-from anthropic.types import DocumentBlockParam
 from typing_extensions import assert_never
 
 from .. import ModelHTTPError, UnexpectedModelBehavior, _utils, usage
@@ -40,6 +38,7 @@ try:
     from anthropic.types import (
         Base64PDFSourceParam,
         ContentBlock,
+        DocumentBlockParam,
         ImageBlockParam,
         Message as AnthropicMessage,
         MessageParam,
@@ -354,48 +353,13 @@ class AnthropicModel(Model):
                     else:
                         raise RuntimeError('Only images and PDFs are supported for binary content')
                 elif isinstance(item, ImageUrl):
-                    try:
-                        response = await cached_async_http_client().get(item.url)
-                        response.raise_for_status()
-                        yield ImageBlockParam(
-                            source={
-                                'data': io.BytesIO(response.content),
-                                'media_type': item.media_type,
-                                'type': 'base64',
-                            },
-                            type='image',
-                        )
-                    except ValueError:
-                        # Download the file if can't find the mime type.
-                        client = cached_async_http_client()
-                        response = await client.get(item.url, follow_redirects=True)
-                        response.raise_for_status()
-                        base64_encoded = base64.b64encode(response.content).decode('utf-8')
-                        if (mime_type := response.headers['Content-Type']) in (
-                            'image/jpeg',
-                            'image/png',
-                            'image/gif',
-                            'image/webp',
-                        ):
-                            yield ImageBlockParam(
-                                source={'data': base64_encoded, 'media_type': mime_type, 'type': 'base64'},
-                                type='image',
-                            )
-                        else:  # pragma: no cover
-                            raise RuntimeError(f'Unsupported image type: {mime_type}')
+                    yield ImageBlockParam(source={'type': 'url', 'url': item.url}, type='image')
                 elif isinstance(item, DocumentUrl):
-                    response = await cached_async_http_client().get(item.url)
-                    response.raise_for_status()
                     if item.media_type == 'application/pdf':
-                        yield DocumentBlockParam(
-                            source=Base64PDFSourceParam(
-                                data=io.BytesIO(response.content),
-                                media_type=item.media_type,
-                                type='base64',
-                            ),
-                            type='document',
-                        )
+                        yield DocumentBlockParam(source={'url': item.url, 'type': 'url'}, type='document')
                     elif item.media_type == 'text/plain':
+                        response = await cached_async_http_client().get(item.url)
+                        response.raise_for_status()
                         yield DocumentBlockParam(
                             source=PlainTextSourceParam(data=response.text, media_type=item.media_type, type='text'),
                             type='document',

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/models/cohere.py

@@ -5,7 +5,6 @@ from dataclasses import dataclass, field
 from itertools import chain
 from typing import Literal, Union, cast
 
-from cohere import TextAssistantMessageContentItem
 from typing_extensions import assert_never
 
 from .. import ModelHTTPError, result
@@ -38,6 +37,7 @@ try:
         ChatMessageV2,
         ChatResponse,
         SystemChatMessageV2,
+        TextAssistantMessageContentItem,
         ToolCallV2,
         ToolCallV2Function,
         ToolChatMessageV2,

{pydantic_ai_slim-0.0.47 → pydantic_ai_slim-0.0.49}/pydantic_ai/models/gemini.py

@@ -57,6 +57,7 @@ LatestGeminiModelNames = Literal[
     'gemini-2.0-flash',
     'gemini-2.0-flash-lite-preview-02-05',
     'gemini-2.0-pro-exp-02-05',
+    'gemini-2.5-pro-exp-03-25',
 ]
 """Latest Gemini models."""