futurehouse-client 0.3.17.dev56__py3-none-any.whl → 0.3.18__py3-none-any.whl

futurehouse_client/clients/rest_client.py

@@ -1,23 +1,27 @@
 import ast
+import asyncio
 import base64
+import contextlib
 import copy
 import importlib.metadata
 import inspect
 import json
 import logging
 import os
+import sys
 import tempfile
+import time
 import uuid
-from collections.abc import Mapping
-from datetime import datetime
+from collections.abc import Collection
 from pathlib import Path
 from types import ModuleType
-from typing import Any, ClassVar, assert_never, cast
+from typing import Any, ClassVar, cast
 from uuid import UUID
 
 import cloudpickle
 from aviary.functional import EnvironmentBuilder
 from httpx import (
+    AsyncClient,
     Client,
     CloseError,
     ConnectError,
@@ -29,7 +33,6 @@ from httpx import (
     RemoteProtocolError,
 )
 from ldp.agent import AgentConfig
-from pydantic import BaseModel, ConfigDict, model_validator
 from requests.exceptions import RequestException, Timeout
 from tenacity import (
     retry,
@@ -37,15 +40,22 @@ from tenacity import (
     stop_after_attempt,
     wait_exponential,
 )
+from tqdm import tqdm as sync_tqdm
+from tqdm.asyncio import tqdm
 
 from futurehouse_client.clients import JobNames
 from futurehouse_client.models.app import (
-    APIKeyPayload,
     AuthType,
     JobDeploymentConfig,
+    PQATaskResponse,
     Stage,
     TaskRequest,
+    TaskResponse,
+    TaskResponseVerbose,
 )
+from futurehouse_client.models.rest import ExecutionStatus
+from futurehouse_client.utils.auth import RefreshingJWT
+from futurehouse_client.utils.general import gather_with_concurrency
 from futurehouse_client.utils.module_utils import (
     OrganizationSelector,
     fetch_environment_function_docstring,
@@ -55,24 +65,14 @@ from futurehouse_client.utils.monitoring import (
 )
 
 logger = logging.getLogger(__name__)
-
+logging.basicConfig(
+    level=logging.WARNING,
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    stream=sys.stdout,
+)
+logging.getLogger("httpx").setLevel(logging.WARNING)
 TaskRequest.model_rebuild()
 
-retry_if_connection_error = retry_if_exception_type((
-    # From requests
-    Timeout,
-    ConnectionError,
-    RequestException,
-    # From httpx
-    ConnectError,
-    ConnectTimeout,
-    ReadTimeout,
-    ReadError,
-    NetworkError,
-    RemoteProtocolError,
-    CloseError,
-))
-
 FILE_UPLOAD_IGNORE_PARTS = {
     ".ruff_cache",
     "__pycache__",
@@ -103,114 +103,35 @@ class InvalidTaskDescriptionError(Exception):
     """Raised when the task description is invalid or empty."""
 
 
-class SimpleOrganization(BaseModel):
-    id: int
-    name: str
-    display_name: str
-
-
-# 5 minute default for JWTs
-JWT_TOKEN_CACHE_EXPIRY: int = 300  # seconds
-
-
-class TaskResponse(BaseModel):
-    """Base class for task responses. This holds attributes shared over all futurehouse jobs."""
-
-    model_config = ConfigDict(extra="ignore")
-
-    status: str
-    query: str
-    user: str | None = None
-    created_at: datetime
-    job_name: str
-    public: bool
-    shared_with: list[SimpleOrganization] | None = None
-    build_owner: str | None = None
-    environment_name: str | None = None
-    agent_name: str | None = None
-    task_id: UUID | None = None
-
-    @model_validator(mode="before")
-    @classmethod
-    def validate_fields(cls, data: Mapping[str, Any]) -> Mapping[str, Any]:
-        # Extract fields from environment frame state
-        if not isinstance(data, dict):
-            return data
-        # TODO: We probably want to remove these two once we define the final names.
-        data["job_name"] = data.get("crow")
-        data["query"] = data.get("task")
-        if not (env_frame := data.get("environment_frame", {})):
-            return data
-        state = env_frame.get("state", {}).get("state", {})
-        data["task_id"] = cast(UUID, state.get("id")) if state.get("id") else None
-        if not (metadata := data.get("metadata", {})):
-            return data
-        data["environment_name"] = metadata.get("environment_name")
-        data["agent_name"] = metadata.get("agent_name")
-        return data
-
-
-class PQATaskResponse(TaskResponse):
-    model_config = ConfigDict(extra="ignore")
-
-    answer: str | None = None
-    formatted_answer: str | None = None
-    answer_reasoning: str | None = None
-    has_successful_answer: bool | None = None
-    total_cost: float | None = None
-    total_queries: int | None = None
-
-    @model_validator(mode="before")
-    @classmethod
-    def validate_pqa_fields(cls, data: Mapping[str, Any]) -> Mapping[str, Any]:
-        # Extract fields from environment frame state
-        if not isinstance(data, dict):
-            return data
-        if not (env_frame := data.get("environment_frame", {})):
-            return data
-        state = env_frame.get("state", {}).get("state", {})
-        response = state.get("response", {})
-        answer = response.get("answer", {})
-        usage = state.get("info", {}).get("usage", {})
-
-        # Add additional PQA specific fields to data so that pydantic can validate the model
-        data["answer"] = answer.get("answer")
-        data["formatted_answer"] = answer.get("formatted_answer")
-        data["answer_reasoning"] = answer.get("answer_reasoning")
-        data["has_successful_answer"] = answer.get("has_successful_answer")
-        data["total_cost"] = cast(float, usage.get("total_cost"))
-        data["total_queries"] = cast(int, usage.get("total_queries"))
-
-        return data
-
-    def clean_verbose(self) -> "TaskResponse":
-        """Clean the verbose response from the server."""
-        self.request = None
-        self.response = None
-        return self
-
-
-class TaskResponseVerbose(TaskResponse):
-    """Class for responses to include all the fields of a task response."""
-
-    model_config = ConfigDict(extra="allow")
-
-    public: bool
-    agent_state: list[dict[str, Any]] | None = None
-    environment_frame: dict[str, Any] | None = None
-    metadata: dict[str, Any] | None = None
-    shared_with: list[SimpleOrganization] | None = None
-
-
 class FileUploadError(RestClientError):
     """Raised when there's an error uploading a file."""
 
 
+retry_if_connection_error = retry_if_exception_type((
+    # From requests
+    Timeout,
+    ConnectionError,
+    RequestException,
+    # From httpx
+    ConnectError,
+    ConnectTimeout,
+    ReadTimeout,
+    ReadError,
+    NetworkError,
+    RemoteProtocolError,
+    CloseError,
+    FileUploadError,
+))
+
+DEFAULT_AGENT_TIMEOUT: int = 2400  # seconds
+
+
 class RestClient:
     REQUEST_TIMEOUT: ClassVar[float] = 30.0  # sec
     MAX_RETRY_ATTEMPTS: ClassVar[int] = 3
     RETRY_MULTIPLIER: ClassVar[int] = 1
     MAX_RETRY_WAIT: ClassVar[int] = 10
+    DEFAULT_POLLING_TIME: ClassVar[int] = 5  # seconds
     CHUNK_SIZE: ClassVar[int] = 16 * 1024 * 1024  # 16MB chunks
 
     def __init__(
@@ -222,62 +143,116 @@ class RestClient:
         api_key: str | None = None,
         jwt: str | None = None,
         headers: dict[str, str] | None = None,
+        verbose_logging: bool = False,
     ):
+        if verbose_logging:
+            logger.setLevel(logging.INFO)
+        else:
+            logger.setLevel(logging.WARNING)
+
         self.base_url = service_uri or stage.value
         self.stage = stage
         self.auth_type = auth_type
         self.api_key = api_key
-        self._clients: dict[str, Client] = {}
+        self._clients: dict[str, Client | AsyncClient] = {}
         self.headers = headers or {}
-        self.auth_jwt = self._run_auth(jwt=jwt)
+        self.jwt = jwt
         self.organizations: list[str] = self._filter_orgs(organization)
 
     @property
     def client(self) -> Client:
-        """Lazily initialized and cached HTTP client with authentication."""
-        return self.get_client("application/json", with_auth=True)
+        """Authenticated HTTP client for regular API calls."""
+        return cast(Client, self.get_client("application/json", authenticated=True))
 
     @property
-    def auth_client(self) -> Client:
-        """Lazily initialized and cached HTTP client without authentication."""
-        return self.get_client("application/json", with_auth=False)
+    def async_client(self) -> AsyncClient:
+        """Authenticated async HTTP client for regular API calls."""
+        return cast(
+            AsyncClient,
+            self.get_client("application/json", authenticated=True, async_client=True),
+        )
+
+    @property
+    def unauthenticated_client(self) -> Client:
+        """Unauthenticated HTTP client for auth operations."""
+        return cast(Client, self.get_client("application/json", authenticated=False))
 
     @property
     def multipart_client(self) -> Client:
-        """Lazily initialized and cached HTTP client for multipart uploads."""
-        return self.get_client(None, with_auth=True)
+        """Authenticated HTTP client for multipart uploads."""
+        return cast(Client, self.get_client(None, authenticated=True))
 
     def get_client(
-        self, content_type: str | None = "application/json", with_auth: bool = True
-    ) -> Client:
+        self,
+        content_type: str | None = "application/json",
+        authenticated: bool = True,
+        async_client: bool = False,
+    ) -> Client | AsyncClient:
         """Return a cached HTTP client or create one if needed.
 
         Args:
             content_type: The desired content type header. Use None for multipart uploads.
-            with_auth: Whether the client should include an Authorization header.
+            authenticated: Whether the client should include authentication.
+            async_client: Whether to use an async client.
 
         Returns:
             An HTTP client configured with the appropriate headers.
         """
-        # Create a composite key based on content type and auth flag.
-        key = f"{content_type or 'multipart'}_{with_auth}"
+        # Create a composite key based on content type and auth flag
+        key = f"{content_type or 'multipart'}_{authenticated}_{async_client}"
+
         if key not in self._clients:
             headers = copy.deepcopy(self.headers)
-            if with_auth:
-                headers["Authorization"] = f"Bearer {self.auth_jwt}"
+            auth = None
+
+            if authenticated:
+                auth = RefreshingJWT(
+                    # authenticated=False will always return a synchronous client
+                    auth_client=cast(
+                        Client, self.get_client("application/json", authenticated=False)
+                    ),
+                    auth_type=self.auth_type,
+                    api_key=self.api_key,
+                    jwt=self.jwt,
+                )
+
             if content_type:
                 headers["Content-Type"] = content_type
-            self._clients[key] = Client(
-                base_url=self.base_url,
-                headers=headers,
-                timeout=self.REQUEST_TIMEOUT,
+
+            self._clients[key] = (
+                AsyncClient(
+                    base_url=self.base_url,
+                    headers=headers,
+                    timeout=self.REQUEST_TIMEOUT,
+                    auth=auth,
+                )
+                if async_client
+                else Client(
+                    base_url=self.base_url,
+                    headers=headers,
+                    timeout=self.REQUEST_TIMEOUT,
+                    auth=auth,
+                )
             )
+
         return self._clients[key]
 
-    def __del__(self):
-        """Ensure all cached clients are properly closed when the instance is destroyed."""
+    def close(self):
+        """Explicitly close all cached clients."""
         for client in self._clients.values():
-            client.close()
+            if isinstance(client, Client):
+                with contextlib.suppress(RuntimeError, CloseError):
+                    client.close()
+
+    async def aclose(self):
+        """Asynchronously close all cached clients."""
+        for client in self._clients.values():
+            if isinstance(client, AsyncClient):
+                with contextlib.suppress(RuntimeError, CloseError):
+                    await client.aclose()
+
+    def __del__(self):
+        self.close()
 
     def _filter_orgs(self, organization: str | None = None) -> list[str]:
         filtered_orgs = [
@@ -289,31 +264,6 @@ class RestClient:
             raise ValueError(f"Organization '{organization}' not found.")
         return filtered_orgs
 
-    def _run_auth(self, jwt: str | None = None) -> str:
-        auth_payload: APIKeyPayload | None
-        if self.auth_type == AuthType.API_KEY:
-            auth_payload = APIKeyPayload(api_key=self.api_key)
-        elif self.auth_type == AuthType.JWT:
-            auth_payload = None
-        else:
-            assert_never(self.auth_type)
-        try:
-            # Use the unauthenticated client for login
-            if auth_payload:
-                response = self.auth_client.post(
-                    "/auth/login", json=auth_payload.model_dump()
-                )
-                response.raise_for_status()
-                token_data = response.json()
-            elif jwt:
-                token_data = {"access_token": jwt, "expires_in": JWT_TOKEN_CACHE_EXPIRY}
-            else:
-                raise ValueError("JWT token required for JWT authentication.")
-
-            return token_data["access_token"]
-        except Exception as e:
-            raise RestClientError(f"Error authenticating: {e!s}") from e
-
     def _check_job(self, name: str, organization: str) -> dict[str, Any]:
         try:
             response = self.client.get(
@@ -407,8 +357,11 @@ class RestClient:
                 ),
                 self.client.stream("GET", url, params={"history": history}) as response,
             ):
+                response.raise_for_status()
                 json_data = "".join(response.iter_text(chunk_size=1024))
                 data = json.loads(json_data)
+                if "id" not in data:
+                    data["id"] = task_id
                 verbose_response = TaskResponseVerbose(**data)
 
             if verbose:
@@ -419,8 +372,52 @@ class RestClient:
             ):
                 return PQATaskResponse(**data)
             return TaskResponse(**data)
-        except ValueError as e:
-            raise ValueError("Invalid task ID format. Must be a valid UUID.") from e
+        except Exception as e:
+            raise TaskFetchError(f"Error getting task: {e!s}") from e
+
+    @retry(
+        stop=stop_after_attempt(MAX_RETRY_ATTEMPTS),
+        wait=wait_exponential(multiplier=RETRY_MULTIPLIER, max=MAX_RETRY_WAIT),
+        retry=retry_if_connection_error,
+    )
+    async def aget_task(
+        self, task_id: str | None = None, history: bool = False, verbose: bool = False
+    ) -> "TaskResponse":
+        """Get details for a specific task asynchronously."""
+        try:
+            task_id = task_id or self.trajectory_id
+            url = f"/v0.1/trajectories/{task_id}"
+            full_url = f"{self.base_url}{url}"
+
+            with external_trace(
+                url=full_url,
+                method="GET",
+                library="httpx",
+                custom_params={
+                    "operation": "get_job",
+                    "job_id": task_id,
+                },
+            ):
+                async with self.async_client.stream(
+                    "GET", url, params={"history": history}
+                ) as response:
+                    response.raise_for_status()
+                    json_data = "".join([
+                        chunk async for chunk in response.aiter_text()
+                    ])
+                    data = json.loads(json_data)
+                    if "id" not in data:
+                        data["id"] = task_id
+                    verbose_response = TaskResponseVerbose(**data)
+
+            if verbose:
+                return verbose_response
+            if any(
+                JobNames.from_string(job_name) in verbose_response.job_name
+                for job_name in ["crow", "falcon", "owl", "dummy"]
+            ):
+                return PQATaskResponse(**data)
+            return TaskResponse(**data)
         except Exception as e:
             raise TaskFetchError(f"Error getting task: {e!s}") from e
 
@@ -445,10 +442,179 @@ class RestClient:
                 "/v0.1/crows", json=task_data.model_dump(mode="json")
             )
             response.raise_for_status()
-            self.trajectory_id = response.json()["trajectory_id"]
+            trajectory_id = response.json()["trajectory_id"]
+            self.trajectory_id = trajectory_id
         except Exception as e:
             raise TaskFetchError(f"Error creating task: {e!s}") from e
-        return self.trajectory_id
+        return trajectory_id
+
+    @retry(
+        stop=stop_after_attempt(MAX_RETRY_ATTEMPTS),
+        wait=wait_exponential(multiplier=RETRY_MULTIPLIER, max=MAX_RETRY_WAIT),
+        retry=retry_if_connection_error,
+    )
+    async def acreate_task(self, task_data: TaskRequest | dict[str, Any]):
+        """Create a new futurehouse task."""
+        if isinstance(task_data, dict):
+            task_data = TaskRequest.model_validate(task_data)
+
+        if isinstance(task_data.name, JobNames):
+            task_data.name = task_data.name.from_stage(
+                task_data.name.name,
+                self.stage,
+            )
+
+        try:
+            response = await self.async_client.post(
+                "/v0.1/crows", json=task_data.model_dump(mode="json")
+            )
+            response.raise_for_status()
+            trajectory_id = response.json()["trajectory_id"]
+            self.trajectory_id = trajectory_id
+        except Exception as e:
+            raise TaskFetchError(f"Error creating task: {e!s}") from e
+        return trajectory_id
+
+    async def arun_tasks_until_done(
+        self,
+        task_data: TaskRequest
+        | dict[str, Any]
+        | Collection[TaskRequest]
+        | Collection[dict[str, Any]],
+        verbose: bool = False,
+        progress_bar: bool = False,
+        concurrency: int = 10,
+        timeout: int = DEFAULT_AGENT_TIMEOUT,
+    ) -> list[TaskResponse]:
+        all_tasks: Collection[TaskRequest | dict[str, Any]] = (
+            cast(Collection[TaskRequest | dict[str, Any]], [task_data])
+            if (isinstance(task_data, dict) or not isinstance(task_data, Collection))
+            else cast(Collection[TaskRequest | dict[str, Any]], task_data)
+        )
+
+        trajectory_ids = await gather_with_concurrency(
+            concurrency,
+            [self.acreate_task(task) for task in all_tasks],
+            progress=progress_bar,
+        )
+
+        start_time = time.monotonic()
+        completed_tasks: dict[str, TaskResponse] = {}
+
+        if progress_bar:
+            progress = tqdm(
+                total=len(trajectory_ids), desc="Waiting for tasks to finish", ncols=0
+            )
+
+        while (time.monotonic() - start_time) < timeout:
+            task_results = await gather_with_concurrency(
+                concurrency,
+                [
+                    self.aget_task(task_id, verbose=verbose)
+                    for task_id in trajectory_ids
+                    if task_id not in completed_tasks
+                ],
+            )
+
+            for task in task_results:
+                task_id = str(task.task_id)
+                if (
+                    task_id not in completed_tasks
+                    and ExecutionStatus(task.status).is_terminal_state()
+                ):
+                    completed_tasks[task_id] = task
+                    if progress_bar:
+                        progress.update(1)
+
+            all_done = len(completed_tasks) == len(trajectory_ids)
+
+            if all_done:
+                break
+            await asyncio.sleep(self.DEFAULT_POLLING_TIME)
+
+        else:
+            logger.warning(
+                f"Timed out waiting for tasks to finish after {timeout} seconds. Returning with {len(completed_tasks)} completed tasks and {len(trajectory_ids)} total tasks."
+            )
+
+        if progress_bar:
+            progress.close()
+
+        return [
+            completed_tasks.get(task_id)
+            or (await self.aget_task(task_id, verbose=verbose))
+            for task_id in trajectory_ids
+        ]
+
+    def run_tasks_until_done(
+        self,
+        task_data: TaskRequest
+        | dict[str, Any]
+        | Collection[TaskRequest]
+        | Collection[dict[str, Any]],
+        verbose: bool = False,
+        progress_bar: bool = False,
+        timeout: int = DEFAULT_AGENT_TIMEOUT,
+    ) -> list[TaskResponse]:
+        """Run multiple tasks and wait for them to complete.
+
+        Args:
+            task_data: A single task or collection of tasks to run
+            verbose: Whether to return verbose task responses
+            progress_bar: Whether to display a progress bar
+            timeout: Maximum time to wait for task completion in seconds
+
+        Returns:
+            A list of completed task responses
+        """
+        all_tasks: Collection[TaskRequest | dict[str, Any]] = (
+            cast(Collection[TaskRequest | dict[str, Any]], [task_data])
+            if (isinstance(task_data, dict) or not isinstance(task_data, Collection))
+            else cast(Collection[TaskRequest | dict[str, Any]], task_data)
+        )
+
+        trajectory_ids = [self.create_task(task) for task in all_tasks]
+
+        start_time = time.monotonic()
+        completed_tasks: dict[str, TaskResponse] = {}
+
+        if progress_bar:
+            progress = sync_tqdm(
+                total=len(trajectory_ids), desc="Waiting for tasks to finish", ncols=0
+            )
+
+        while (time.monotonic() - start_time) < timeout:
+            all_done = True
+
+            for task_id in trajectory_ids:
+                if task_id in completed_tasks:
+                    continue
+
+                task = self.get_task(task_id, verbose=verbose)
+
+                if not ExecutionStatus(task.status).is_terminal_state():
+                    all_done = False
+                elif task_id not in completed_tasks:
+                    completed_tasks[task_id] = task
+                    if progress_bar:
+                        progress.update(1)
+
+            if all_done:
+                break
+            time.sleep(self.DEFAULT_POLLING_TIME)
+
+        else:
+            logger.warning(
+                f"Timed out waiting for tasks to finish after {timeout} seconds. Returning with {len(completed_tasks)} completed tasks and {len(trajectory_ids)} total tasks."
+            )
+
+        if progress_bar:
+            progress.close()
+
+        return [
+            completed_tasks.get(task_id) or self.get_task(task_id, verbose=verbose)
+            for task_id in trajectory_ids
+        ]
 
     @retry(
         stop=stop_after_attempt(MAX_RETRY_ATTEMPTS),
@@ -457,9 +623,12 @@ class RestClient:
     )
     def get_build_status(self, build_id: UUID | None = None) -> dict[str, Any]:
         """Get the status of a build."""
-        build_id = build_id or self.build_id
-        response = self.client.get(f"/v0.1/builds/{build_id}")
-        response.raise_for_status()
+        try:
+            build_id = build_id or self.build_id
+            response = self.client.get(f"/v0.1/builds/{build_id}")
+            response.raise_for_status()
+        except Exception as e:
+            raise JobFetchError(f"Error getting build status: {e!s}") from e
         return response.json()
 
     # TODO: Refactor later so we don't have to ignore PLR0915
@@ -660,14 +829,14 @@ class RestClient:
         self,
         job_name: str,
         file_path: str | os.PathLike,
-        folder_name: str | None = None,
+        upload_id: str | None = None,
     ) -> str:
         """Upload a file or directory to a futurehouse job bucket.
 
         Args:
             job_name: The name of the futurehouse job to upload to.
             file_path: The local path to the file or directory to upload.
-            folder_name: Optional folder name to use for the upload. If not provided, a random UUID will be used.
+            upload_id: Optional folder name to use for the upload. If not provided, a random UUID will be used.
 
         Returns:
             The upload ID used for the upload.
@@ -679,7 +848,7 @@ class RestClient:
         if not file_path.exists():
             raise FileNotFoundError(f"File or directory not found: {file_path}")
 
-        upload_id = folder_name or str(uuid.uuid4())
+        upload_id = upload_id or str(uuid.uuid4())
 
         if file_path.is_dir():
             # Process directory recursively
@@ -742,6 +911,12 @@ class RestClient:
         """
         file_name = file_name or file_path.name
         file_size = file_path.stat().st_size
+
+        # Skip empty files
+        if file_size == 0:
+            logger.warning(f"Skipping upload of empty file: {file_path}")
+            return
+
         total_chunks = (file_size + self.CHUNK_SIZE - 1) // self.CHUNK_SIZE
 
         logger.info(f"Uploading {file_path} as {file_name} ({total_chunks} chunks)")
@@ -789,7 +964,6 @@ class RestClient:
                         )
 
             logger.info(f"Successfully uploaded {file_name}")
-
         except Exception as e:
             logger.exception(f"Error uploading file {file_path}")
             raise FileUploadError(f"Error uploading file {file_path}: {e}") from e
@@ -799,12 +973,18 @@ class RestClient:
         wait=wait_exponential(multiplier=RETRY_MULTIPLIER, max=MAX_RETRY_WAIT),
         retry=retry_if_connection_error,
     )
-    def list_files(self, job_name: str, folder_name: str) -> dict[str, list[str]]:
+    def list_files(
+        self,
+        job_name: str,
+        trajectory_id: str | None = None,
+        upload_id: str | None = None,
+    ) -> dict[str, list[str]]:
         """List files and directories in a GCS location for a given job_name and upload_id.
 
         Args:
             job_name: The name of the futurehouse job.
-            folder_name: The specific folder name (upload_id) to list files from.
+            trajectory_id: The specific trajectory id to list files from.
+            upload_id: The specific upload id to list files from.
 
         Returns:
             A list of files in the GCS folder.
@@ -812,22 +992,27 @@ class RestClient:
         Raises:
             RestClientError: If there is an error listing the files.
         """
+        if not bool(trajectory_id) ^ bool(upload_id):
+            raise RestClientError(
+                "Must at least specify one of trajectory_id or upload_id, but not both"
+            )
         try:
             url = f"/v0.1/crows/{job_name}/list-files"
-            params = {"upload_id": folder_name}
+            params = {"trajectory_id": trajectory_id, "upload_id": upload_id}
+            params = {k: v for k, v in params.items() if v is not None}
             response = self.client.get(url, params=params)
             response.raise_for_status()
             return response.json()
         except HTTPStatusError as e:
             logger.exception(
-                f"Error listing files for job {job_name}, folder {folder_name}: {e.response.text}"
+                f"Error listing files for job {job_name}, trajectory {trajectory_id}, upload_id {upload_id}: {e.response.text}"
             )
             raise RestClientError(
                 f"Error listing files: {e.response.status_code} - {e.response.text}"
             ) from e
         except Exception as e:
             logger.exception(
-                f"Error listing files for job {job_name}, folder {folder_name}"
+                f"Error listing files for job {job_name}, trajectory {trajectory_id}, upload_id {upload_id}"
             )
             raise RestClientError(f"Error listing files: {e!s}") from e
 
@@ -839,7 +1024,7 @@ class RestClient:
     def download_file(
         self,
         job_name: str,
-        folder_name: str,
+        trajectory_id: str,
         file_path: str,
         destination_path: str | os.PathLike,
     ) -> None:
@@ -847,14 +1032,14 @@ class RestClient:
 
         Args:
             job_name: The name of the futurehouse job.
-            folder_name: The specific folder name (upload_id) the file belongs to.
+            trajectory_id: The specific trajectory id the file belongs to.
             file_path: The relative path of the file to download
                        (e.g., 'data/my_file.csv' or 'my_image.png').
             destination_path: The local path where the file should be saved.
 
         Raises:
             RestClientError: If there is an error downloading the file.
-            FileNotFoundError: If the destination directory does not exist.
+            FileNotFoundError: If the destination directory does not exist or if the file is not found.
         """
         destination_path = Path(destination_path)
         # Ensure the destination directory exists
@@ -862,17 +1047,24 @@ class RestClient:
 
         try:
             url = f"/v0.1/crows/{job_name}/download-file"
-            params = {"upload_id": folder_name, "file_path": file_path}
+            params = {"trajectory_id": trajectory_id, "file_path": file_path}
 
             with self.client.stream("GET", url, params=params) as response:
                 response.raise_for_status()  # Check for HTTP errors before streaming
                 with open(destination_path, "wb") as f:
                     for chunk in response.iter_bytes(chunk_size=8192):
                         f.write(chunk)
+
+            # Check if the downloaded file is empty
+            if destination_path.stat().st_size == 0:
+                # Remove the empty file
+                destination_path.unlink()
+                raise FileNotFoundError(f"File not found or is empty: {file_path}")
+
             logger.info(f"File {file_path} downloaded to {destination_path}")
         except HTTPStatusError as e:
             logger.exception(
-                f"Error downloading file {file_path} for job {job_name}, folder {folder_name}: {e.response.text}"
+                f"Error downloading file {file_path} for job {job_name}, trajectory_id {trajectory_id}: {e.response.text}"
             )
             # Clean up partially downloaded file if an error occurs
             if destination_path.exists():
@@ -880,9 +1072,20 @@ class RestClient:
             raise RestClientError(
                 f"Error downloading file: {e.response.status_code} - {e.response.text}"
             ) from e
+        except RemoteProtocolError as e:
+            logger.error(
+                f"Connection error while downloading file {file_path} for job {job_name}, trajectory_id {trajectory_id}"
+            )
+            # Clean up partially downloaded file
+            if destination_path.exists():
+                destination_path.unlink()
+
+            # Often RemoteProtocolError during download means the file wasn't found
+            # or was empty/corrupted on the server side
+            raise FileNotFoundError(f"File not found or corrupted: {file_path}") from e
         except Exception as e:
             logger.exception(
-                f"Error downloading file {file_path} for job {job_name}, folder {folder_name}"
+                f"Error downloading file {file_path} for job {job_name}, trajectory_id {trajectory_id}"
             )
             if destination_path.exists():
                 destination_path.unlink()  # Clean up partial file