Merge pull request #128 from PrefectHQ/python3dot9

.github/workflows/build-docs.yml

@@ -16,7 +16,7 @@ jobs:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.9"
       - uses: actions/cache@v2
         with:
           key: ${{ github.ref }}

.github/workflows/publish-docs.yml

@@ -18,7 +18,7 @@ jobs:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.9"
       - uses: actions/cache@v2
         with:
           key: ${{ github.ref }}

.github/workflows/run-llm-tests.yml

@@ -33,10 +33,10 @@ jobs:
 
     steps:
       - uses: actions/checkout@v3
-      - name: Set up Python 3.10
+      - name: Set up Python 3.9
         uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.9"
           cache: "pip" # caching pip dependencies
           cache-dependency-path: "**/pyproject.toml"
       - name: Install Marvin

.github/workflows/run-tests.yml

@@ -42,10 +42,10 @@ jobs:
 
     steps:
       - uses: actions/checkout@v3
-      - name: Set up Python 3.10
+      - name: Set up Python 3.9
         uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.9"
           cache: "pip" # caching pip dependencies
           cache-dependency-path: "**/pyproject.toml"
       - name: Install Marvin

.github/workflows/static-analysis.yml

@@ -26,6 +26,6 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.9"
       - name: Run pre-commit
         uses: pre-commit/[email protected]

.pre-commit-config.yaml

@@ -11,4 +11,4 @@ repos:
     hooks:
       - id: black
         args: ["--preview"]
-        language_version: python3.10
+        language_version: python3.9

docs/getting_started/installation.md

@@ -5,10 +5,12 @@
 
 ## Requirements
 
-Marvin requires Python 3.10. While nothing in Marvin should prevent use with Python 3.11, some dependencies may not be fully compatible yet.
+Marvin requires Python 3.9+.
 ## For normal use
 To install Marvin, run `pip install marvin`. As a matter of best practice, we recommend installing Marvin in a [virtual environment](https://realpython.com/python-virtual-environments-a-primer/).
 
+To use Marvin's [knowledge features](../guide/concepts/infra.md), please include the [Chroma](https://www.trychroma.com/) dependency: `pip install marvin[chromadb]`.
+
 Before using Marvin, you'll need to [configure your OpenAI API key](openai.md).
 
 ## For development

pyproject.toml

@@ -12,7 +12,7 @@ classifiers = [
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
 keywords = ["ai", "chatbot", "llm"]
-requires-python = ">=3.10"
+requires-python = ">=3.9"
 dependencies = [
     "aiofiles~=23.1.0",
     "aiosqlite~=0.18.0",

src/marvin/api/bots.py

@@ -54,7 +54,7 @@ async def create_bot_config(
 async def get_bot_config(
     name: str,
     session: AsyncSession = Depends(fastapi_session),
-) -> BotConfigRead | None:
+) -> BotConfigRead:
     result = await session.execute(
         sa.select(BotConfig).where(BotConfig.name == name).limit(1)
     )

src/marvin/api/threads.py

@@ -33,7 +33,7 @@ async def create_thread(
 async def get_thread_by_lookup_key(
     lookup_key: str,
     session: AsyncSession = Depends(fastapi_session),
-) -> Thread | None:
+) -> Thread:
     result = await session.execute(
         sa.select(Thread).where(Thread.lookup_key == lookup_key).limit(1)
     )
@@ -48,7 +48,7 @@ async def get_thread_by_lookup_key(
 async def get_thread(
     thread_id: ThreadID = Path(..., alias="id"),
     session: AsyncSession = Depends(fastapi_session),
-) -> Thread | None:
+) -> Thread:
     result = await session.execute(
         sa.select(Thread).where(Thread.id == thread_id).limit(1)
     )

src/marvin/bots/ai_functions.py

@@ -94,7 +94,6 @@ def rhyme(word: str) -> str:
         - bot_kwargs (dict):  kwargs to pass to the `Bot` constructor
 
     """
-
     # this allows the decorator to be used with or without calling it
     if fn is None:
         return partial(

src/marvin/bots/base.py

@@ -27,6 +27,8 @@ class BotResponse(BaseMessage):
     parsed_content: Any = None
 
 
+MAX_VALIDATION_ATTEMPTS = 3
+
 if TYPE_CHECKING:
     from marvin.models.bots import BotConfig
 DEFAULT_NAME = "Marvin"
@@ -312,31 +314,33 @@ async def say(self, *args, response_format=None, **kwargs) -> BotResponse:
         # validate response format
         parsed_response = response
         validated = False
-        validation_attempts = 0
 
-        while not validated and validation_attempts < 3:
-            validation_attempts += 1
+        for _ in range(MAX_VALIDATION_ATTEMPTS):
             try:
                 self.response_format.validate_response(response)
-                validated = True
+                break
             except Exception as exc:
-                match self.response_format.on_error:
-                    case "ignore":
-                        validated = True
-                    case "raise":
-                        raise exc
-                    case "reformat":
-                        self.logger.debug(
-                            "Response did not pass validation. Attempted to reformat:"
-                            f" {response}"
-                        )
-                        reformatted_response = _reformat_response(
-                            user_message=user_message.content,
-                            ai_response=response,
-                            error_message=repr(exc),
-                            target_return_type=self.response_format.format,
-                        )
-                        response = str(reformatted_response)
+                on_error = self.response_format.on_error
+                if on_error == "ignore":
+                    break
+                elif on_error == "raise":
+                    raise exc
+                elif on_error == "reformat":
+                    self.logger.debug(
+                        "Response did not pass validation. Attempted to reformat:"
+                        f" {response}"
+                    )
+                    reformatted_response = _reformat_response(
+                        user_message=user_message.content,
+                        ai_response=response,
+                        error_message=repr(exc),
+                        target_return_type=self.response_format.format,
+                    )
+                    response = str(reformatted_response)
+                else:
+                    raise ValueError(f"Unknown on_error value: {on_error}")
+        else:
+            raise RuntimeError("Failed to validate response after 3 attempts")
 
         if validated:
             parsed_response = self.response_format.parse_response(response)

src/marvin/bots/response_formatters.py

@@ -2,7 +2,7 @@
 import re
 import warnings
 from types import GenericAlias
-from typing import Any, Literal
+from typing import Any, Literal, Union
 
 import pydantic
 from pydantic import BaseModel, Field, PrivateAttr
@@ -56,7 +56,7 @@ def parse_response(self, response):
 
 
 class TypeFormatter(ResponseFormatter):
-    _cached_type: type | GenericAlias = PrivateAttr(SENTINEL)
+    _cached_type: Union[type, GenericAlias] = PrivateAttr(SENTINEL)
     type_schema: dict[str, Any] = Field(
         ..., description="The OpenAPI schema for the type"
     )
@@ -101,7 +101,7 @@ def __init__(self, type_: type = SENTINEL, **kwargs):
         if type_ is not SENTINEL:
             self._cached_type = type_
 
-    def get_type(self) -> type | GenericAlias:
+    def get_type(self) -> Union[type, GenericAlias]:
         if self._cached_type is not SENTINEL:
             return self._cached_type
 
@@ -132,7 +132,7 @@ class PydanticFormatter(ResponseFormatter):
         ..., description="The OpenAPI schema for the model"
     )
 
-    def __init__(self, model: BaseModel | GenericAlias = None, **kwargs):
+    def __init__(self, model: Union[BaseModel, GenericAlias] = None, **kwargs):
         if model is not None:
             for key in ["format", "type_schema"]:
                 if key in kwargs:
@@ -172,35 +172,20 @@ def parse_response(self, response):
 
 
 def load_formatter_from_shorthand(shorthand_response_format) -> ResponseFormatter:
-    match shorthand_response_format:
-        # shorthand for None
-        case None:
-            return ResponseFormatter()
-
-        # x is a ResponseFormatter - no shorthand
-        case x if isinstance(x, ResponseFormatter):
-            return x
-
-        # x is a boolean
-        case x if x is bool:
-            return BooleanFormatter()
-
-        # x is a string that contains the word "json"
-        case x if isinstance(x, str) and re.search(r"\bjson\b", x.lower()):
-            return JSONFormatter(format=x)
-
-        # x is a string
-        case x if isinstance(x, str):
-            return ResponseFormatter(format=x)
-
-        # x is a pydantic model
-        case x if genericalias_contains(x, pydantic.BaseModel):
-            return PydanticFormatter(model=x)
-
-        # x is a type or GenericAlias
-        case x if isinstance(x, (type, GenericAlias)):
-            return TypeFormatter(type_=x)
-
-        # unsupported values
-        case _:
-            raise ValueError("Invalid output format")
+    if shorthand_response_format is None:
+        return ResponseFormatter()
+    elif isinstance(shorthand_response_format, ResponseFormatter):
+        return shorthand_response_format
+    elif shorthand_response_format is bool:
+        return BooleanFormatter()
+    elif isinstance(shorthand_response_format, str):
+        if re.search(r"\bjson\b", shorthand_response_format.lower()):
+            return JSONFormatter(format=shorthand_response_format)
+        else:
+            return ResponseFormatter(format=shorthand_response_format)
+    elif genericalias_contains(shorthand_response_format, pydantic.BaseModel):
+        return PydanticFormatter(model=shorthand_response_format)
+    elif isinstance(shorthand_response_format, (type, GenericAlias)):
+        return TypeFormatter(type_=shorthand_response_format)
+    else:
+        raise ValueError("Invalid output format")

src/marvin/config.py

@@ -1,7 +1,7 @@
 import os
 from contextlib import contextmanager
 from pathlib import Path
-from typing import Literal
+from typing import Literal, Optional
 
 try:
     import chromadb
@@ -64,7 +64,7 @@ def export_to_env_file(self):
     # LOGGING
     verbose: bool = False
     log_level: Literal["DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"] = "INFO"
-    log_console_width: int | None = Field(
+    log_console_width: Optional[int] = Field(
         None,
         description=(
             "Marvin will auto-detect the console width when possible, but in deployed"

src/marvin/loaders/github.py

@@ -3,7 +3,7 @@
 import functools
 import re
 from pathlib import Path
-from typing import Dict, List, Tuple
+from typing import Dict, List, Optional, Tuple
 
 import aiofiles
 import chardet
@@ -55,7 +55,7 @@ class GitHubIssue(BaseModel):
     html_url: str = Field(...)
     number: int = Field(...)
     title: str = Field(default="")
-    body: str | None = Field(default="")
+    body: Optional[str] = Field(default="")
     labels: List[GitHubLabel] = Field(default_factory=GitHubLabel)
     user: GitHubUser = Field(default_factory=GitHubUser)
 

src/marvin/models/documents.py

@@ -1,5 +1,6 @@
 import asyncio
 import inspect
+from typing import Optional
 
 from jinja2 import Template
 from pydantic import Field, confloat, validator
@@ -95,15 +96,15 @@ class Document(MarvinBaseModel):
     text: str = Field(
         ..., description="Any text content that you want to keep / embed."
     )
-    embedding: list[float] | None = Field(default=None)
+    embedding: Optional[list[float]] = Field(default=None)
     metadata: Metadata = Field(default_factory=Metadata)
 
-    source: str | None = Field(default=None)
+    source: Optional[str] = Field(default=None)
     type: DocumentType = Field(default="original")
-    parent_document_id: DocumentID | None = Field(default=None)
+    parent_document_id: Optional[DocumentID] = Field(default=None)
     topic_name: str = Field(default=marvin.settings.default_topic)
-    tokens: int | None = Field(default=None)
-    order: int | None = Field(default=None)
+    tokens: Optional[int] = Field(default=None)
+    order: Optional[int] = Field(default=None)
     keywords: list[str] = Field(default_factory=list)
 
     @validator("tokens", pre=True, always=True)

src/marvin/plugins/chroma.py

@@ -1,4 +1,4 @@
-from typing import Any, Callable
+from typing import Any, Callable, Optional
 
 import pendulum
 from pydantic import Field
@@ -111,8 +111,8 @@ def iso_to_timestamp(filter: dict) -> float:
 @plugin
 async def chroma_search(
     query: str,
-    where: dict[str, Any] | None = None,
-    where_document: dict[str, Any] | None = None,
+    where: Optional[dict[str, Any]] = None,
+    where_document: Optional[dict[str, Any]] = None,
 ) -> str:
     """
     query (str): A verbose natural language query.

src/marvin/utilities/strings.py

@@ -91,7 +91,7 @@ def split_text(
     chunk_overlap: float = None,
     last_chunk_threshold: float = None,
     return_index: bool = False,
-) -> str | tuple[str, int]:
+) -> Union[str, tuple[str, int]]:
     """
     Split a text into a list of strings. Chunks are split by tokens.