Add support for models.predictions.create endpoint (#207)

Signed-off-by: Mattt Zmuda <mattt@replicate.com>

Author: mattt

.github/workflows/ci.yaml

@@ -13,6 +13,8 @@ jobs:
 
     name: "Test Python ${{ matrix.python-version }}"
 
+    timeout-minutes: 10
+
     strategy:
       fail-fast: false
       matrix:

replicate/identifier.py

@@ -2,6 +2,29 @@
 from typing import NamedTuple
 
 
+class ModelIdentifier(NamedTuple):
+    """
+    A reference to a model in the format owner/name:version.
+    """
+
+    owner: str
+    name: str
+
+    @classmethod
+    def parse(cls, ref: str) -> "ModelIdentifier":
+        """
+        Split a reference in the format owner/name:version into its components.
+        """
+
+        match = re.match(r"^(?P<owner>[^/]+)/(?P<name>[^:]+)$", ref)
+        if not match:
+            raise ValueError(
+                f"Invalid reference to model version: {ref}. Expected format: owner/name"
+            )
+
+        return cls(match.group("owner"), match.group("name"))
+
+
 class ModelVersionIdentifier(NamedTuple):
     """
     A reference to a model version in the format owner/name:version.

replicate/model.py

@@ -1,10 +1,15 @@
-from typing import TYPE_CHECKING, Any, Dict, Literal, Optional, Union
+from typing import TYPE_CHECKING, Any, Dict, Literal, Optional, Tuple, Union
 
 from typing_extensions import NotRequired, TypedDict, Unpack, deprecated
 
 from replicate.exceptions import ReplicateException
+from replicate.identifier import ModelIdentifier
 from replicate.pagination import Page
-from replicate.prediction import Prediction
+from replicate.prediction import (
+    Prediction,
+    _create_prediction_body,
+    _json_to_prediction,
+)
 from replicate.resource import Namespace, Resource
 from replicate.version import Version, Versions
 
@@ -16,6 +21,7 @@
 
 if TYPE_CHECKING:
     from replicate.client import Client
+    from replicate.prediction import Predictions
 
 
 class Model(Resource):
@@ -140,6 +146,14 @@ class Models(Namespace):
 
     model = Model
 
+    @property
+    def predictions(self) -> "ModelsPredictions":
+        """
+        Get a namespace for operations related to predictions on a model.
+        """
+
+        return ModelsPredictions(client=self._client)
+
     def list(self, cursor: Union[str, "ellipsis", None] = ...) -> Page[Model]:  # noqa: F821
         """
         List all public models.
@@ -275,6 +289,54 @@ async def async_create(
         return _json_to_model(self._client, resp.json())
 
 
+class ModelsPredictions(Namespace):
+    """
+    Namespace for operations related to predictions in a deployment.
+    """
+
+    def create(
+        self,
+        model: Optional[Union[str, Tuple[str, str], "Model"]],
+        input: Dict[str, Any],
+        **params: Unpack["Predictions.CreatePredictionParams"],
+    ) -> Prediction:
+        """
+        Create a new prediction with the deployment.
+        """
+
+        url = _create_prediction_url_from_model(model)
+        body = _create_prediction_body(version=None, input=input, **params)
+
+        resp = self._client._request(
+            "POST",
+            url,
+            json=body,
+        )
+
+        return _json_to_prediction(self._client, resp.json())
+
+    async def async_create(
+        self,
+        model: Optional[Union[str, Tuple[str, str], "Model"]],
+        input: Dict[str, Any],
+        **params: Unpack["Predictions.CreatePredictionParams"],
+    ) -> Prediction:
+        """
+        Create a new prediction with the deployment.
+        """
+
+        url = _create_prediction_url_from_model(model)
+        body = _create_prediction_body(version=None, input=input, **params)
+
+        resp = await self._client._async_request(
+            "POST",
+            url,
+            json=body,
+        )
+
+        return _json_to_prediction(self._client, resp.json())
+
+
 def _create_model_body(  # pylint: disable=too-many-arguments
     owner: str,
     name: str,
@@ -318,3 +380,22 @@ def _json_to_model(client: "Client", json: Dict[str, Any]) -> Model:
     if model.default_example is not None:
         model.default_example._client = client
     return model
+
+
+def _create_prediction_url_from_model(
+    model: Union[str, Tuple[str, str], "Model"]
+) -> str:
+    owner, name = None, None
+    if isinstance(model, Model):
+        owner, name = model.owner, model.name
+    elif isinstance(model, tuple):
+        owner, name = model[0], model[1]
+    elif isinstance(model, str):
+        owner, name = ModelIdentifier.parse(model)
+
+    if owner is None or name is None:
+        raise ValueError(
+            "model must be a Model, a tuple of (owner, name), or a string in the format 'owner/name'"
+        )
+
+    return f"/v1/models/{owner}/{name}/predictions"

replicate/training.py

@@ -13,7 +13,7 @@
 from typing_extensions import NotRequired, Unpack
 
 from replicate.files import upload_file
-from replicate.identifier import ModelVersionIdentifier
+from replicate.identifier import ModelIdentifier, ModelVersionIdentifier
 from replicate.json import encode_json
 from replicate.model import Model
 from replicate.pagination import Page
@@ -378,8 +378,12 @@ def _create_training_url_from_model_and_version(
         owner, name = model.owner, model.name
     elif isinstance(model, tuple):
         owner, name = model[0], model[1]
+    elif isinstance(model, str):
+        owner, name = ModelIdentifier.parse(model)
     else:
-        raise ValueError("model must be a Model or a tuple of (owner, name)")
+        raise ValueError(
+            "model must be a Model, a tuple of (owner, name), or a string in the format 'owner/name'"
+        )
 
     if isinstance(version, Version):
         version_id = version.id

replicate/version.py

@@ -35,7 +35,7 @@ class Versions(Namespace):
     model: Tuple[str, str]
 
     def __init__(
-        self, client: "Client", model: Union["Model", str, Tuple[str, str]]
+        self, client: "Client", model: Union[str, Tuple[str, str], "Model"]
     ) -> None:
         super().__init__(client=client)
 

tests/cassettes/models-predictions-create.yaml

@@ -0,0 +1,55 @@
+interactions:
+- request:
+    body: '{"input": {"prompt": "Please write a haiku about llamas."}}'
+    headers:
+      accept:
+      - '*/*'
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '59'
+      content-type:
+      - application/json
+      host:
+      - api.replicate.com
+      user-agent:
+      - replicate-python/0.21.0
+    method: POST
+    uri: https://api.replicate.com/v1/models/meta/llama-2-70b-chat/predictions
+  response:
+    content: '{"id":"heat2o3bzn3ahtr6bjfftvbaci","model":"replicate/lifeboat-70b","version":"d-c6559c5791b50af57b69f4a73f8e021c","input":{"prompt":"Please
+      write a haiku about llamas."},"logs":"","error":null,"status":"starting","created_at":"2023-11-27T13:35:45.99397566Z","urls":{"cancel":"https://api.replicate.com/v1/predictions/heat2o3bzn3ahtr6bjfftvbaci/cancel","get":"https://api.replicate.com/v1/predictions/heat2o3bzn3ahtr6bjfftvbaci"}}
+
+      '
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 82cac197efaec53d-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '431'
+      Content-Type:
+      - application/json
+      Date:
+      - Mon, 27 Nov 2023 13:35:46 GMT
+      NEL:
+      - '{"success_fraction":0,"report_to":"cf-nel","max_age":604800}'
+      Report-To:
+      - '{"endpoints":[{"url":"https:\/\/a.nel.cloudflare.com\/report\/v3?s=7R5RONMF6xaGRc39n0wnSe3jU1FbpX64Xz4U%2B%2F2nasvFaz0pKARxPhnzDgYkLaWgdK9zWrD2jxU04aKOy5HMPHAXboJ993L4zfsOyto56lBtdqSjNgkptzzxYEsKD%2FxIhe2F"}],"group":"cf-nel","max_age":604800}'
+      Server:
+      - cloudflare
+      Strict-Transport-Security:
+      - max-age=15552000
+      ratelimit-remaining:
+      - '599'
+      ratelimit-reset:
+      - '1'
+      via:
+      - 1.1 google
+    http_version: HTTP/1.1
+    status_code: 201
+version: 1

tests/test_model.py

@@ -107,3 +107,26 @@ async def test_models_create_with_positional_arguments(async_flag):
     assert model.owner == "test"
     assert model.name == "python-example"
     assert model.visibility == "private"
+
+
+@pytest.mark.vcr("models-predictions-create.yaml")
+@pytest.mark.asyncio
+@pytest.mark.parametrize("async_flag", [True, False])
+async def test_models_predictions_create(async_flag):
+    input = {
+        "prompt": "Please write a haiku about llamas.",
+    }
+
+    if async_flag:
+        prediction = await replicate.models.predictions.async_create(
+            "meta/llama-2-70b-chat", input=input
+        )
+    else:
+        prediction = replicate.models.predictions.create(
+            "meta/llama-2-70b-chat", input=input
+        )
+
+    assert prediction.id is not None
+    # assert prediction.model == "meta/llama-2-70b-chat"
+    assert prediction.model == "replicate/lifeboat-70b"  # FIXME: this is temporary
+    assert prediction.status == "starting"