Merge branch 'main' into fix-cerebras

docs/concepts/partial.md

@@ -5,6 +5,23 @@ description: Learn to utilize field-level streaming with Instructor and OpenAI f
 
 # Streaming Partial Responses
 
+!!! info "Literal"
+
+    If the data structure you're using has literal values, you need to make sure to import the `PartialLiteralMixin` mixin.
+
+    ```python
+    from instructor.dsl.partial import PartialLiteralMixin
+
+    class User(BaseModel, PartialLiteralMixin):
+        name: str
+        age: int
+        category: Literal["admin", "user", "guest"]
+
+    // The rest of your code below
+    ```
+
+    This is because `jiter` throws an error otherwise if it encounters a incomplete Literal value while it's being streamed in
+
 Field level streaming provides incremental snapshots of the current state of the response model that are immediately useable. This approach is particularly relevant in contexts like rendering UI components.
 
 Instructor supports this pattern by making use of `create_partial`. This lets us dynamically create a new class that treats all of the original model's fields as `Optional`.

instructor/dsl/partial.py

@@ -34,6 +34,10 @@ class MakeFieldsOptional:
     pass
 
 
+class PartialLiteralMixin:
+    pass
+
+
 def _make_field_optional(
     field: FieldInfo,
 ) -> tuple[Any, FieldInfo]:
@@ -127,10 +131,13 @@ def model_from_chunks(
     ) -> Generator[T_Model, None, None]:
         potential_object = ""
         partial_model = cls.get_partial_model()
+        partial_mode = (
+            "on" if issubclass(cls, PartialLiteralMixin) else "trailing-strings"
+        )
         for chunk in json_chunks:
             potential_object += chunk
             obj = from_json(
-                (potential_object.strip() or "{}").encode(), partial_mode="on"
+                (potential_object.strip() or "{}").encode(), partial_mode=partial_mode
             )
             obj = partial_model.model_validate(obj, strict=None, **kwargs)
             yield obj
@@ -141,10 +148,13 @@ async def model_from_chunks_async(
     ) -> AsyncGenerator[T_Model, None]:
         potential_object = ""
         partial_model = cls.get_partial_model()
+        partial_mode = (
+            "on" if issubclass(cls, PartialLiteralMixin) else "trailing-strings"
+        )
         async for chunk in json_chunks:
             potential_object += chunk
             obj = from_json(
-                (potential_object.strip() or "{}").encode(), partial_mode="on"
+                (potential_object.strip() or "{}").encode(), partial_mode=partial_mode
             )
             obj = partial_model.model_validate(obj, strict=None, **kwargs)
             yield obj

tests/dsl/test_partial.py

@@ -1,6 +1,6 @@
 # type: ignore[all]
 from pydantic import BaseModel, Field
-from instructor.dsl.partial import Partial
+from instructor.dsl.partial import Partial, PartialLiteralMixin
 import pytest
 import instructor
 from openai import OpenAI, AsyncOpenAI
@@ -116,7 +116,7 @@ async def async_generator():
 
 
 def test_summary_extraction():
-    class Summary(BaseModel):
+    class Summary(BaseModel, PartialLiteralMixin):
         summary: str = Field(description="A detailed summary")
 
     client = OpenAI()
@@ -143,7 +143,7 @@ class Summary(BaseModel):
 
 @pytest.mark.asyncio
 async def test_summary_extraction_async():
-    class Summary(BaseModel):
+    class Summary(BaseModel, PartialLiteralMixin):
         summary: str = Field(description="A detailed summary")
 
     client = AsyncOpenAI()

tests/llm/test_openai/test_stream.py

@@ -3,7 +3,7 @@
 from pydantic import BaseModel
 import pytest
 import instructor
-from instructor.dsl.partial import Partial
+from instructor.dsl.partial import Partial, PartialLiteralMixin
 
 from .util import models, modes
 
@@ -81,3 +81,124 @@ async def test_partial_model_async(model, mode, aclient):
     )
     async for m in model:
         assert isinstance(m, UserExtract)
+
+
+@pytest.mark.parametrize("model,mode", product(models, modes))
+def test_literal_partial_mixin(model, mode, client):
+    class UserWithMixin(BaseModel, PartialLiteralMixin):
+        name: str
+        age: int
+
+    client = instructor.patch(client, mode=mode)
+    resp = client.chat.completions.create(
+        model=model,
+        response_model=Partial[UserWithMixin],
+        max_retries=2,
+        stream=True,
+        messages=[
+            {"role": "user", "content": "Jason Liu is 12 years old"},
+        ],
+    )
+
+    changes = 0
+    last_name = None
+    last_age = None
+    for m in resp:
+        assert isinstance(m, UserWithMixin)
+        if m.name != last_name:
+            last_name = m.name
+            changes += 1
+        if m.age != last_age:
+            last_age = m.age
+            changes += 1
+
+    assert changes == 2  # Ensure we got at least one field update
+
+    class UserWithoutMixin(BaseModel):
+        name: str
+        age: int
+
+    resp = client.chat.completions.create(
+        model=model,
+        response_model=Partial[UserWithoutMixin],
+        max_retries=2,
+        stream=True,
+        messages=[
+            {"role": "user", "content": "Jason Liu is 12 years old"},
+        ],
+    )
+
+    changes = 0
+    last_name = None
+    last_age = None
+    for m in resp:
+        assert isinstance(m, UserWithoutMixin)
+        if m.name != last_name:
+            last_name = m.name
+            changes += 1
+        if m.age != last_age:
+            last_age = m.age
+            changes += 1
+
+    assert changes > 3
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("model,mode", product(models, modes))
+async def test_literal_partial_mixin_async(model, mode, aclient):
+    class UserWithMixin(BaseModel, PartialLiteralMixin):
+        name: str
+        age: int
+
+    client = instructor.patch(aclient, mode=mode)
+    resp = await client.chat.completions.create(
+        model=model,
+        response_model=Partial[UserWithMixin],
+        max_retries=2,
+        stream=True,
+        messages=[
+            {"role": "user", "content": "Jason Liu is 12 years old"},
+        ],
+    )
+
+    changes = 0
+    last_name = None
+    last_age = None
+    async for m in resp:
+        assert isinstance(m, UserWithMixin)
+        if m.name != last_name:
+            last_name = m.name
+            changes += 1
+        if m.age != last_age:
+            last_age = m.age
+            changes += 1
+
+    assert changes == 2  # Ensure we got at least one field update
+
+    class UserWithoutMixin(BaseModel):
+        name: str
+        age: int
+
+    resp = await client.chat.completions.create(
+        model=model,
+        response_model=Partial[UserWithoutMixin],
+        max_retries=2,
+        stream=True,
+        messages=[
+            {"role": "user", "content": "Jason Liu is 12 years old"},
+        ],
+    )
+
+    changes = 0
+    last_name = None
+    last_age = None
+    async for m in resp:
+        assert isinstance(m, UserWithoutMixin)
+        if m.name != last_name:
+            last_name = m.name
+            changes += 1
+        if m.age != last_age:
+            last_age = m.age
+            changes += 1
+
+    assert changes > 3