exa-py 1.13.1__py3-none-any.whl → 1.13.2__py3-none-any.whl

exa_py/api.py

@@ -38,6 +38,7 @@ from exa_py.utils import (
 )
 from .websets import WebsetsClient
 from .websets.core.base import ExaJSONEncoder
+from .research.client import ResearchClient, AsyncResearchClient
 
 is_beta = os.getenv("IS_BETA") == "True"
 
@@ -837,37 +838,6 @@ def nest_fields(original_dict: Dict, fields_to_nest: List[str], new_key: str):
     return original_dict
 
 
-@dataclass
-class ResearchTaskResponse:
-    """A class representing the response for a research task.
-
-    Attributes:
-        id (str): The unique identifier for the research request.
-        status (str): Status of the research request.
-        output (Optional[Dict[str, Any]]): The answer structured as JSON, if available.
-        citations (Optional[Dict[str, List[_Result]]]): List of citations used to generate the answer, grouped by root field in the output schema.
-    """
-
-    id: str
-    status: str
-    output: Optional[Dict[str, Any]]
-    citations: Dict[str, List[_Result]]
-
-    def __str__(self):
-        output_repr = (
-            json.dumps(self.output, indent=2, ensure_ascii=False)
-            if self.output is not None
-            else "None"
-        )
-        citations_str = "\n\n".join(str(src) for src in self.citations)
-        return (
-            f"ID: {self.id}\n"
-            f"Status: {self.status}\n"
-            f"Output: {output_repr}\n\n"
-            f"Citations:\n{citations_str}"
-        )
-
-
 class Exa:
     """A client for interacting with Exa API."""
 
@@ -875,7 +845,7 @@ class Exa:
         self,
         api_key: Optional[str],
         base_url: str = "https://api.exa.ai",
-        user_agent: str = "exa-py 1.12.1",
+        user_agent: str = "exa-py 1.12.4",
     ):
         """Initialize the Exa client with the provided API key and optional base URL and user agent.
 
@@ -898,6 +868,8 @@ class Exa:
             "Content-Type": "application/json",
         }
         self.websets = WebsetsClient(self)
+        # Research tasks client (new, mirrors Websets design)
+        self.research = ResearchClient(self)
 
     def request(
         self,
@@ -1952,40 +1924,12 @@ class Exa:
         raw_response = self.request("/answer", options)
         return StreamAnswerResponse(raw_response)
 
-    def researchTask(
-        self,
-        *,
-        input_instructions: str,
-        output_schema: Dict[str, Any],
-    ) -> ResearchTaskResponse:
-        """Submit a research request to Exa.
-
-        Args:
-            input_instructions (str): The instructions for the research task.
-            output_schema (Dict[str, Any]): JSON schema describing the desired answer structure.
-        """
-        # Build the request payload expected by the Exa API
-        options = {
-            "input": {"instructions": input_instructions},
-            "output": {"schema": output_schema},
-        }
-
-        response = self.request("/research/tasks", options)
-
-        return ResearchTaskResponse(
-            id=response["id"],
-            status=response["status"],
-            output=response.get("output"),
-            citations={
-                key: [_Result(**to_snake_case(citation)) for citation in citations_list]
-                for key, citations_list in response.get("citations", {}).items()
-            },
-        )
-
 
 class AsyncExa(Exa):
     def __init__(self, api_key: str, api_base: str = "https://api.exa.ai"):
         super().__init__(api_key, api_base)
+        # Override the synchronous ResearchClient with its async counterpart.
+        self.research = AsyncResearchClient(self)
         self._client = None
 
     @property
@@ -2316,36 +2260,3 @@ class AsyncExa(Exa):
         options["stream"] = True
         raw_response = await self.async_request("/answer", options)
         return AsyncStreamAnswerResponse(raw_response)
-
-    async def researchTask(
-        self,
-        *,
-        input_instructions: str,
-        output_schema: Dict[str, Any],
-    ) -> ResearchTaskResponse:
-        """Asynchronously submit a research request to Exa.
-
-        Args:
-            input_instructions (str): The instructions for the research task.
-            output_schema (Dict[str, Any]): JSON schema describing the desired answer structure.
-
-        Returns:
-            ResearchTaskResponse: The parsed response from the Exa API.
-        """
-        # Build the request payload expected by the Exa API
-        options = {
-            "input": {"instructions": input_instructions},
-            "output": {"schema": output_schema},
-        }
-
-        response = await self.async_request("/research/tasks", options)
-
-        return ResearchTaskResponse(
-            id=response["id"],
-            status=response["status"],
-            output=response.get("output"),
-            citations={
-                key: [_Result(**to_snake_case(citation)) for citation in citations_list]
-                for key, citations_list in response.get("citations", {}).items()
-            },
-        )

exa_py/research/__init__.py

@@ -0,0 +1,9 @@
+from .client import ResearchClient, AsyncResearchClient
+from .models import ResearchTask
+
+__all__ = [
+    "ResearchClient",
+    "AsyncResearchClient",
+    "ResearchTaskId",
+    "ResearchTask",
+]

exa_py/research/client.py

@@ -0,0 +1,232 @@
+"""Lightweight research client wrappers for the Exa REST API.
+
+This module purposefully keeps its import surface minimal to avoid circular
+import problems with :pymod:`exa_py.api`.  Any heavy dependencies (including
+`exa_py.api` itself) are imported lazily **inside** functions.  This means
+that type-checkers still see the full, precise types via the ``TYPE_CHECKING``
+block, but at runtime we only pay the cost if/when a helper is actually used.
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any, Dict
+
+if TYPE_CHECKING:  # pragma: no cover – only for static analysers
+    # Import with full type info when static type-checking.  `_Result` still
+    # lives in ``exa_py.api`` but the response model moved to
+    # ``exa_py.research.models``.
+    from ..api import _Result  # noqa: F401
+    from .models import ResearchTask, ResearchTaskId  # noqa: F401
+
+# ---------------------------------------------------------------------------
+# Public, user-facing clients
+# ---------------------------------------------------------------------------
+
+
+class ResearchClient:
+    """Synchronous helper namespace accessed via :pyattr:`Exa.research`."""
+
+    def __init__(self, parent_client):
+        # A reference to the *already-constructed* ``Exa`` instance so that we
+        # can piggy-back on its HTTP plumbing (headers, base URL, retries, …).
+        self._client = parent_client
+
+    def create_task(
+        self,
+        *,
+        input_instructions: str,
+        output_schema: Dict[str, Any],
+    ) -> "ResearchTaskId":
+        """Submit a research request and return the *task identifier*."""
+        payload = {
+            "input": {"instructions": input_instructions},
+            "output": {"schema": output_schema},
+        }
+
+        raw_response: Dict[str, Any] = self._client.request("/research/tasks", payload)
+
+        # Defensive checks so that we fail loudly if the contract changes.
+        if not isinstance(raw_response, dict) or "id" not in raw_response:
+            raise RuntimeError(
+                f"Unexpected response while creating research task: {raw_response}"
+            )
+
+        # Lazily import to avoid circular deps at runtime.
+        from .models import ResearchTaskId  # noqa: WPS433 – runtime import
+
+        return ResearchTaskId(id=raw_response["id"])
+
+    def get_task(
+        self, id: str
+    ) -> "ResearchTask":  # noqa: D401 – imperative mood is fine
+        """Fetch the current status / result for a research task."""
+        endpoint = f"/research/tasks/{id}"
+
+        # The new endpoint is a simple GET.
+        raw_response: Dict[str, Any] = self._client.request(endpoint, method="GET")
+
+        return _build_research_task(raw_response)
+
+    # ------------------------------------------------------------------
+    # Convenience helpers
+    # ------------------------------------------------------------------
+
+    def poll_task(
+        self,
+        id: str,
+        *,
+        poll_interval: float = 1.0,
+        timeout_seconds: int = 15 * 60,
+    ) -> "ResearchTask":
+        """Blocking helper that polls until task completes or fails.
+
+        Parameters
+        ----------
+        id:
+            The ID of the research task to poll.
+        poll_interval:
+            Seconds to wait between successive polls (default 1s).
+        timeout_seconds:
+            Maximum time to wait before raising :class:`TimeoutError` (default 15 min).
+        """
+
+        import time
+
+        deadline = time.monotonic() + timeout_seconds
+
+        while True:
+            task = self.get_task(id)
+            status = task.status.lower() if isinstance(task.status, str) else ""
+
+            if status in {"completed", "failed", "complete", "finished", "done"}:
+                return task
+
+            if time.monotonic() > deadline:
+                raise TimeoutError(
+                    f"Research task {id} did not finish within {timeout_seconds} seconds"
+                )
+
+            time.sleep(poll_interval)
+
+
+class AsyncResearchClient:
+    """Async counterpart used via :pyattr:`AsyncExa.research`."""
+
+    def __init__(self, parent_client):
+        self._client = parent_client
+
+    async def create_task(
+        self,
+        *,
+        input_instructions: str,
+        output_schema: Dict[str, Any],
+    ) -> "ResearchTaskId":
+        """Submit a research request and return the *task identifier* (async)."""
+
+        payload = {
+            "input": {"instructions": input_instructions},
+            "output": {"schema": output_schema},
+        }
+
+        raw_response: Dict[str, Any] = await self._client.async_request(
+            "/research/tasks", payload
+        )
+
+        # Defensive checks so that we fail loudly if the contract changes.
+        if not isinstance(raw_response, dict) or "id" not in raw_response:
+            raise RuntimeError(
+                f"Unexpected response while creating research task: {raw_response}"
+            )
+
+        # Lazily import to avoid circular deps at runtime.
+        from .models import ResearchTaskId  # noqa: WPS433 – runtime import
+
+        return ResearchTaskId(id=raw_response["id"])
+
+    async def get_task(self, id: str) -> "ResearchTask":  # noqa: D401
+        """Fetch the current status / result for a research task (async)."""
+
+        endpoint = f"/research/tasks/{id}"
+
+        # Perform GET using the underlying HTTP client because `async_request`
+        # only supports POST semantics.
+        resp = await self._client.client.get(
+            self._client.base_url + endpoint, headers=self._client.headers
+        )
+
+        if resp.status_code >= 400:
+            raise RuntimeError(
+                f"Request failed with status code {resp.status_code}: {resp.text}"
+            )
+
+        raw_response: Dict[str, Any] = resp.json()
+
+        return _build_research_task(raw_response)
+
+    # ------------------------------------------------------------------
+    # Convenience helpers
+    # ------------------------------------------------------------------
+
+    async def poll_task(
+        self,
+        id: str,
+        *,
+        poll_interval: float = 1.0,
+        timeout_seconds: int = 15 * 60,
+    ) -> "ResearchTask":
+        """Async helper that polls until task completes or fails.
+
+        Mirrors :py:meth:`ResearchClient.poll_task` but uses ``await`` and
+        :pyfunc:`asyncio.sleep`.  Raises :class:`TimeoutError` on timeout.
+        """
+
+        import asyncio
+        import time
+
+        deadline = time.monotonic() + timeout_seconds
+
+        while True:
+            task = await self.get_task(id)
+            status = task.status.lower() if isinstance(task.status, str) else ""
+
+            if status in {"completed", "failed", "complete", "finished", "done"}:
+                return task
+
+            if time.monotonic() > deadline:
+                raise TimeoutError(
+                    f"Research task {id} did not finish within {timeout_seconds} seconds"
+                )
+
+            await asyncio.sleep(poll_interval)
+
+
+# ---------------------------------------------------------------------------
+# Internal helpers (lazy imports to avoid cycles)
+# ---------------------------------------------------------------------------
+
+
+def _build_research_task(raw: Dict[str, Any]):
+    """Convert raw API response into a :class:`ResearchTask` instance."""
+
+    # Defensive check – fail loudly if the API contract changes.
+    if not isinstance(raw, dict) or "id" not in raw:
+        raise RuntimeError(f"Unexpected response while fetching research task: {raw}")
+
+    # Lazily import heavy deps to avoid cycles and unnecessary startup cost.
+    from .models import ResearchTask  # noqa: WPS433 – runtime import
+    from ..api import _Result, to_snake_case  # noqa: WPS433 – runtime import
+
+    citations_raw = raw.get("citations", {}) or {}
+    citations_parsed = {
+        key: [_Result(**to_snake_case(c)) for c in cites]
+        for key, cites in citations_raw.items()
+    }
+
+    return ResearchTask(
+        id=raw["id"],
+        status=raw["status"],
+        instructions=raw.get("instructions", ""),
+        schema=raw.get("schema", {}),
+        data=raw.get("data"),
+        citations=citations_parsed,
+    )

exa_py/research/models.py

@@ -0,0 +1,98 @@
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from typing import Any, Dict, List, Optional
+
+# Local import placed inside TYPE_CHECKING block to avoid runtime cycles.
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:  # pragma: no cover – for static analysers only
+    from ..api import _Result  # noqa: F401
+
+
+@dataclass
+class ResearchTaskId:
+    """Structured research task ID.
+
+    Attributes
+    ----------
+    id:
+        Unique identifier for the research task.
+    """
+
+    id: str
+
+    # ---------------------------------------------------------------------
+    # Pretty representation helpers
+    # ---------------------------------------------------------------------
+    def __str__(self) -> str:  # pragma: no cover – convenience only
+        return f"ID: {self.id}\n"
+
+
+@dataclass
+class ResearchTask:
+    """Structured research task.
+
+    Attributes
+    ----------
+    id:
+        Unique identifier for the research task.
+    status:
+        Current task status
+    instructions:
+        Instructions for the task
+    schema:
+        Output schema defining the task
+    data:
+        JSON-serialisable answer generated by Exa (may be ``None`` until the task
+        completes).
+    citations:
+        Mapping from *root field* in the output schema to the list of search
+        results that were used to generate that part of the answer.
+    """
+
+    id: str
+    status: str
+    instructions: str
+    schema: Dict[str, Any]
+    data: Optional[Dict[str, Any]]
+    citations: Dict[str, List["_Result"]]
+
+    # ---------------------------------------------------------------------
+    # Pretty representation helpers
+    # ---------------------------------------------------------------------
+    def __str__(self) -> str:  # pragma: no cover – convenience only
+        """Human-readable representation including *all* relevant fields."""
+        schema_repr = json.dumps(self.schema, indent=2, ensure_ascii=False)
+        data_repr = (
+            json.dumps(self.data, indent=2, ensure_ascii=False)
+            if self.data is not None
+            else "None"
+        )
+
+        # Render citations grouped by the root field they belong to.
+        if self.citations:
+            # Each key is a root field, each value is a list of _Result objects.
+            citations_lines = []
+            for field, sources in self.citations.items():
+                rendered_sources = "\n    ".join(str(src) for src in sources)
+                citations_lines.append(f"{field}:\n    {rendered_sources}")
+            citations_str = "\n\n".join(citations_lines)
+        else:
+            citations_str = "None"
+
+        return (
+            f"ID: {self.id}\n"
+            f"Status: {self.status}\n"
+            f"Instructions: {self.instructions}\n"
+            f"Schema:\n{schema_repr}\n"
+            f"Data:\n{data_repr}\n\n"
+            f"Citations:\n{citations_str}"
+        )
+
+
+__all__ = [
+    "ResearchTaskId",
+    "ResearchTask",
+]

exa_py/websets/client.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 
 import time
-from datetime import datetime
 from typing import List, Optional, Literal, Dict, Any, Union
 
 from .types import (
@@ -17,6 +16,7 @@ from .items import WebsetItemsClient
 from .searches import WebsetSearchesClient
 from .enrichments import WebsetEnrichmentsClient
 from .webhooks import WebsetWebhooksClient
+from .streams import StreamsClient
 
 class WebsetsClient(WebsetsBaseClient):
     """Client for managing Websets."""
@@ -27,6 +27,7 @@ class WebsetsClient(WebsetsBaseClient):
         self.searches = WebsetSearchesClient(client)
         self.enrichments = WebsetEnrichmentsClient(client)
         self.webhooks = WebsetWebhooksClient(client)
+        self.streams = StreamsClient(client)
 
     def create(self, params: Union[Dict[str, Any], CreateWebsetParameters]) -> Webset:
         """Create a new Webset.

exa_py/websets/core/base.py

@@ -29,7 +29,7 @@ class ExaBaseModel(BaseModel):
         str_to_upper=False,  # Don't convert strings to uppercase
         from_attributes=True,  # Allow initialization from attributes
         validate_assignment=True,  # Validate on assignment
-        extra='forbid',  # Forbid extra fields
+        extra='allow',
         json_encoders={AnyUrl: str}  # Convert AnyUrl to string when serializing to JSON
     )
 
@@ -92,5 +92,9 @@ class WebsetsBaseClient:
             # If data is a model instance, convert it to a dict
             data = data.model_dump(by_alias=True, exclude_none=True)
             
+        # Ensure proper URL construction by removing leading slash from endpoint if present
+        if endpoint.startswith("/"):
+            endpoint = endpoint[1:]
+            
         return self._client.request("/websets/" + endpoint, data=data, method=method, params=params) 
-    
+

exa_py/websets/streams/__init__.py

@@ -0,0 +1,4 @@
+from .client import StreamsClient
+from .runs import StreamRunsClient
+
+__all__ = ["StreamsClient", "StreamRunsClient"] 

exa_py/websets/streams/client.py

@@ -0,0 +1,96 @@
+from __future__ import annotations
+
+from typing import Dict, Any, Union, Optional
+
+from ..types import (
+    Stream,
+    CreateStreamParameters,
+    UpdateStream,
+    ListStreamsResponse,
+)
+from ..core.base import WebsetsBaseClient
+from .runs import StreamRunsClient
+
+class StreamsClient(WebsetsBaseClient):
+    """Client for managing Streams."""
+    
+    def __init__(self, client):
+        super().__init__(client)
+        self.runs = StreamRunsClient(client)
+
+    def create(self, params: Union[Dict[str, Any], CreateStreamParameters]) -> Stream:
+        """Create a new Stream to continuously keep your Websets updated with fresh data.
+        
+        Streams automatically run on your defined schedule to ensure your Websets stay current without manual intervention:
+        - Find new content: Execute search operations to discover fresh items matching your criteria
+        - Update existing content: Run refresh operations to update items contents and enrichments
+        - Automated scheduling: Configure frequency, timezone, and execution times
+        
+        Args:
+            params (CreateStreamParameters): The parameters for creating a stream.
+        
+        Returns:
+            Stream: The created stream.
+        """
+        response = self.request("/v0/streams", data=params)
+        return Stream.model_validate(response)
+
+    def get(self, stream_id: str) -> Stream:
+        """Get a specific stream.
+        
+        Args:
+            stream_id (str): The id of the Stream.
+        
+        Returns:
+            Stream: The retrieved stream.
+        """
+        response = self.request(f"/v0/streams/{stream_id}", method="GET")
+        return Stream.model_validate(response)
+
+    def list(self, *, cursor: Optional[str] = None, limit: Optional[int] = None, webset_id: Optional[str] = None) -> ListStreamsResponse:
+        """List all streams.
+        
+        Args:
+            cursor (str, optional): The cursor to paginate through the results.
+            limit (int, optional): The number of results to return (1-200, default 25).
+            webset_id (str, optional): The id of the Webset to list streams for.
+        
+        Returns:
+            ListStreamsResponse: List of streams with pagination info.
+        """
+        params = {
+            k: v 
+            for k, v in {
+                "cursor": cursor, 
+                "limit": limit, 
+                "websetId": webset_id
+            }.items() 
+            if v is not None
+        }
+        response = self.request("/v0/streams", params=params, method="GET")
+        return ListStreamsResponse.model_validate(response)
+
+    def update(self, stream_id: str, params: Union[Dict[str, Any], UpdateStream]) -> Stream:
+        """Update a stream configuration.
+        
+        Args:
+            stream_id (str): The id of the Stream.
+            params (UpdateStream): The parameters for updating a stream.
+        
+        Returns:
+            Stream: The updated stream.
+        """
+        response = self.request(f"/v0/streams/{stream_id}", data=params, method="PATCH")
+        return Stream.model_validate(response)
+
+    def delete(self, stream_id: str) -> Stream:
+        """Delete a stream.
+        
+        Args:
+            stream_id (str): The id of the Stream.
+        
+        Returns:
+            Stream: The deleted stream.
+        """
+        response = self.request(f"/v0/streams/{stream_id}", method="DELETE")
+        return Stream.model_validate(response) 

exa_py/websets/streams/runs/__init__.py

@@ -0,0 +1,3 @@
+from .client import StreamRunsClient
+
+__all__ = ["StreamRunsClient"] 

exa_py/websets/streams/runs/client.py

@@ -0,0 +1,38 @@
+from __future__ import annotations
+
+from ...types import (
+    StreamRun,
+    ListStreamRunsResponse,
+)
+from ...core.base import WebsetsBaseClient
+
+class StreamRunsClient(WebsetsBaseClient):
+    """Client for managing Stream Runs."""
+    
+    def __init__(self, client):
+        super().__init__(client)
+
+    def list(self, stream_id: str) -> ListStreamRunsResponse:
+        """List all runs for the Stream.
+        
+        Args:
+            stream_id (str): The id of the Stream to list runs for.
+        
+        Returns:
+            ListStreamRunsResponse: List of stream runs.
+        """
+        response = self.request(f"/v0/streams/{stream_id}/runs", method="GET")
+        return ListStreamRunsResponse.model_validate(response)
+
+    def get(self, stream_id: str, run_id: str) -> StreamRun:
+        """Get a specific stream run.
+        
+        Args:
+            stream_id (str): The id of the Stream to get the run for.
+            run_id (str): The id of the stream run.
+        
+        Returns:
+            StreamRun: The stream run details.
+        """
+        response = self.request(f"/v0/streams/{stream_id}/runs/{run_id}", method="GET")
+        return StreamRun.model_validate(response)