arize-phoenix 11.23.1__py3-none-any.whl → 12.28.1__py3-none-any.whl

phoenix/server/api/routers/v1/experiment_runs.py

@@ -1,33 +1,34 @@
+import json
 from datetime import datetime
 from typing import Any, Optional
 
-from fastapi import APIRouter, Depends, HTTPException
+from fastapi import APIRouter, Depends, HTTPException, Query
 from pydantic import Field
 from sqlalchemy import select
-from sqlalchemy.exc import IntegrityError as PostgreSQLIntegrityError
-from sqlean.dbapi2 import IntegrityError as SQLiteIntegrityError  # type: ignore[import-untyped]
 from starlette.requests import Request
-from starlette.status import HTTP_404_NOT_FOUND, HTTP_409_CONFLICT
 from strawberry.relay import GlobalID
 
 from phoenix.db import models
+from phoenix.db.helpers import get_runs_with_incomplete_evaluations_query
+from phoenix.db.insertion.helpers import OnConflict, insert_on_conflict
 from phoenix.db.models import ExperimentRunOutput
+from phoenix.server.api.routers.v1.datasets import DatasetExample
 from phoenix.server.api.types.node import from_global_id_with_expected_type
 from phoenix.server.authorization import is_not_locked
 from phoenix.server.dml_event import ExperimentRunInsertEvent
 
 from .models import V1RoutesBaseModel
-from .utils import ResponseBody, add_errors_to_responses
+from .utils import PaginatedResponseBody, ResponseBody, add_errors_to_responses
 
 router = APIRouter(tags=["experiments"], include_in_schema=True)
 
 
-class ExperimentRun(V1RoutesBaseModel):
+class ExperimentRunData(V1RoutesBaseModel):
     dataset_example_id: str = Field(
         description="The ID of the dataset example used in the experiment run"
     )
     output: Any = Field(description="The output of the experiment task")
-    repetition_number: int = Field(description="The repetition number of the experiment run")
+    repetition_number: int = Field(description="The repetition number of the experiment run", gt=0)
     start_time: datetime = Field(description="The start time of the experiment run")
     end_time: datetime = Field(description="The end time of the experiment run")
     trace_id: Optional[str] = Field(
@@ -39,7 +40,7 @@ class ExperimentRun(V1RoutesBaseModel):
     )
 
 
-class CreateExperimentRunRequestBody(ExperimentRun):
+class CreateExperimentRunRequestBody(ExperimentRunData):
     pass
 
 
@@ -60,12 +61,14 @@ class CreateExperimentRunResponseBody(ResponseBody[CreateExperimentRunResponseBo
     responses=add_errors_to_responses(
         [
             {
-                "status_code": HTTP_404_NOT_FOUND,
+                "status_code": 404,
                 "description": "Experiment or dataset example not found",
             },
             {
-                "status_code": HTTP_409_CONFLICT,
-                "description": "This experiment run has already been submitted",
+                "status_code": 409,
+                "description": (
+                    "Experiment run already exists with a successful result and cannot be updated"
+                ),
             },
         ]
     ),
@@ -79,7 +82,7 @@ async def create_experiment_run(
     except ValueError:
         raise HTTPException(
             detail=f"Experiment with ID {experiment_gid} does not exist",
-            status_code=HTTP_404_NOT_FOUND,
+            status_code=404,
         )
 
     example_gid = GlobalID.from_id(request_body.dataset_example_id)
@@ -88,7 +91,7 @@ async def create_experiment_run(
     except ValueError:
         raise HTTPException(
             detail=f"DatasetExample with ID {example_gid} does not exist",
-            status_code=HTTP_404_NOT_FOUND,
+            status_code=404,
         )
 
     trace_id = request_body.trace_id
@@ -99,37 +102,72 @@ async def create_experiment_run(
     error = request_body.error
 
     async with request.app.state.db() as session:
-        exp_run = models.ExperimentRun(
-            experiment_id=experiment_rowid,
-            dataset_example_id=dataset_example_id,
-            trace_id=trace_id,
-            output=ExperimentRunOutput(task_output=task_output),
-            repetition_number=repetition_number,
-            start_time=start_time,
-            end_time=end_time,
-            error=error,
+        # Check if a record already exists
+        existing_run = await session.scalar(
+            select(models.ExperimentRun)
+            .where(models.ExperimentRun.experiment_id == experiment_rowid)
+            .where(models.ExperimentRun.dataset_example_id == dataset_example_id)
+            .where(models.ExperimentRun.repetition_number == repetition_number)
         )
-        try:
-            session.add(exp_run)
-            await session.flush()
-        except (PostgreSQLIntegrityError, SQLiteIntegrityError):
+
+        if existing_run is not None and existing_run.error is None:
+            # Record exists and has no error - reject the update
+            run_gid = GlobalID("ExperimentRun", str(existing_run.id))
             raise HTTPException(
-                detail="This experiment run has already been submitted",
-                status_code=HTTP_409_CONFLICT,
+                status_code=409,
+                detail=(
+                    f"Experiment run {run_gid} already exists with a successful result "
+                    "and cannot be updated"
+                ),
             )
-    request.state.event_queue.put(ExperimentRunInsertEvent((exp_run.id,)))
-    run_gid = GlobalID("ExperimentRun", str(exp_run.id))
+        # Either no record exists, or existing record has an error - proceed with upsert
+        stmt = insert_on_conflict(
+            {
+                "experiment_id": experiment_rowid,
+                "dataset_example_id": dataset_example_id,
+                "trace_id": trace_id,
+                "output": ExperimentRunOutput(task_output=task_output),
+                "repetition_number": repetition_number,
+                "start_time": start_time,
+                "end_time": end_time,
+                "error": error,
+            },
+            table=models.ExperimentRun,
+            dialect=request.app.state.db.dialect,
+            unique_by=["experiment_id", "dataset_example_id", "repetition_number"],
+            on_conflict=OnConflict.DO_UPDATE,
+        ).returning(models.ExperimentRun.id)
+        id_ = await session.scalar(stmt)
+
+    request.state.event_queue.put(ExperimentRunInsertEvent((id_,)))
+    run_gid = GlobalID("ExperimentRun", str(id_))
     return CreateExperimentRunResponseBody(
         data=CreateExperimentRunResponseBodyData(id=str(run_gid))
     )
 
 
-class ExperimentRunResponse(ExperimentRun):
+class ExperimentRun(ExperimentRunData):
     id: str = Field(description="The ID of the experiment run")
     experiment_id: str = Field(description="The ID of the experiment")
 
 
-class ListExperimentRunsResponseBody(ResponseBody[list[ExperimentRunResponse]]):
+class ListExperimentRunsResponseBody(PaginatedResponseBody[ExperimentRun]):
+    pass
+
+
+class IncompleteExperimentEvaluation(V1RoutesBaseModel):
+    """
+    Information about an experiment run with incomplete evaluations
+    """
+
+    experiment_run: ExperimentRun = Field(description="The experiment run")
+    dataset_example: DatasetExample = Field(description="The dataset example")
+    evaluation_names: list[str] = Field(
+        description="List of evaluation names that are incomplete (either missing or failed)"
+    )
+
+
+class GetIncompleteEvaluationsResponseBody(PaginatedResponseBody[IncompleteExperimentEvaluation]):
     pass
 
 
@@ -137,47 +175,285 @@ class ListExperimentRunsResponseBody(ResponseBody[list[ExperimentRunResponse]]):
     "/experiments/{experiment_id}/runs",
     operation_id="listExperimentRuns",
     summary="List runs for an experiment",
+    description="Retrieve a paginated list of runs for an experiment",
     response_description="Experiment runs retrieved successfully",
     responses=add_errors_to_responses(
-        [{"status_code": HTTP_404_NOT_FOUND, "description": "Experiment not found"}]
+        [
+            {"status_code": 404, "description": "Experiment not found"},
+            {"status_code": 422, "description": "Invalid cursor format"},
+        ]
     ),
 )
 async def list_experiment_runs(
-    request: Request, experiment_id: str
+    request: Request,
+    experiment_id: str,
+    cursor: Optional[str] = Query(
+        default=None,
+        description="Cursor for pagination (base64-encoded experiment run ID)",
+    ),
+    limit: Optional[int] = Query(
+        default=None,
+        description="The max number of experiment runs to return at a time. "
+        "If not specified, returns all results.",
+        gt=0,
+    ),
 ) -> ListExperimentRunsResponseBody:
-    experiment_gid = GlobalID.from_id(experiment_id)
+    try:
+        experiment_gid = GlobalID.from_id(experiment_id)
+    except Exception as e:
+        raise HTTPException(
+            detail=f"Invalid experiment ID format: {experiment_id}",
+            status_code=422,
+        ) from e
     try:
         experiment_rowid = from_global_id_with_expected_type(experiment_gid, "Experiment")
     except ValueError:
         raise HTTPException(
             detail=f"Experiment with ID {experiment_gid} does not exist",
-            status_code=HTTP_404_NOT_FOUND,
+            status_code=404,
         )
 
+    stmt = (
+        select(models.ExperimentRun)
+        .filter_by(experiment_id=experiment_rowid)
+        .order_by(models.ExperimentRun.id.desc())
+    )
+
+    if cursor:
+        try:
+            cursor_id = GlobalID.from_id(cursor).node_id
+            stmt = stmt.where(models.ExperimentRun.id <= int(cursor_id))
+        except ValueError:
+            raise HTTPException(
+                detail=f"Invalid cursor format: {cursor}",
+                status_code=422,
+            )
+
+    # Apply limit only if specified for pagination
+    if limit is not None:
+        stmt = stmt.limit(limit + 1)
+
     async with request.app.state.db() as session:
-        experiment_runs = await session.execute(
-            select(models.ExperimentRun)
-            .where(models.ExperimentRun.experiment_id == experiment_rowid)
-            # order by dataset_example_id to be consistent with `list_dataset_examples`
-            .order_by(models.ExperimentRun.dataset_example_id.asc())
+        experiment_runs = (await session.scalars(stmt)).all()
+
+    if not experiment_runs:
+        return ListExperimentRunsResponseBody(next_cursor=None, data=[])
+
+    next_cursor = None
+    # Only check for next cursor if limit was specified
+    if limit is not None and len(experiment_runs) == limit + 1:
+        last_run = experiment_runs[-1]
+        next_cursor = str(GlobalID("ExperimentRun", str(last_run.id)))
+        experiment_runs = experiment_runs[:-1]
+
+    runs = []
+    for exp_run in experiment_runs:
+        run_gid = GlobalID("ExperimentRun", str(exp_run.id))
+        experiment_gid = GlobalID("Experiment", str(exp_run.experiment_id))
+        example_gid = GlobalID("DatasetExample", str(exp_run.dataset_example_id))
+        runs.append(
+            ExperimentRun(
+                start_time=exp_run.start_time,
+                end_time=exp_run.end_time,
+                experiment_id=str(experiment_gid),
+                dataset_example_id=str(example_gid),
+                repetition_number=exp_run.repetition_number,
+                output=exp_run.output.get("task_output"),
+                error=exp_run.error,
+                id=str(run_gid),
+                trace_id=exp_run.trace_id,
+            )
+        )
+    return ListExperimentRunsResponseBody(data=runs, next_cursor=next_cursor)
+
+
+@router.get(
+    "/experiments/{experiment_id}/incomplete-evaluations",
+    operation_id="getIncompleteExperimentEvaluations",
+    summary="Get incomplete evaluations for an experiment",
+    responses=add_errors_to_responses(
+        [
+            {"status_code": 400, "description": "No evaluator names provided"},
+            {"status_code": 404, "description": "Experiment not found"},
+            {"status_code": 422, "description": "Invalid cursor format"},
+        ]
+    ),
+    response_description="Incomplete evaluations retrieved successfully",
+)
+async def get_incomplete_evaluations(
+    request: Request,
+    experiment_id: str,
+    evaluation_name: list[str] = Query(default=[], description="Evaluation names to check"),
+    cursor: Optional[str] = Query(default=None, description="Cursor for pagination"),
+    limit: int = Query(
+        default=50, description="Maximum number of runs with incomplete evaluations to return", gt=0
+    ),
+) -> GetIncompleteEvaluationsResponseBody:
+    """
+    Get experiment runs that have incomplete evaluations.
+
+    Returns runs with:
+    - Missing evaluations (evaluator has not been run)
+    - Failed evaluations (evaluator ran but has errors)
+
+    Args:
+        experiment_id: The ID of the experiment
+        evaluation_name: List of evaluation names to check (required, at least one)
+        cursor: Cursor for pagination
+        limit: Maximum number of results to return
+
+    Returns:
+        Paginated list of runs with incomplete evaluations
+    """
+    try:
+        experiment_globalid = GlobalID.from_id(experiment_id)
+    except Exception as e:
+        raise HTTPException(
+            detail=f"Invalid experiment ID format: {experiment_id}",
+            status_code=422,
+        ) from e
+    try:
+        experiment_rowid = from_global_id_with_expected_type(experiment_globalid, "Experiment")
+    except ValueError:
+        raise HTTPException(
+            detail=f"Experiment with ID {experiment_globalid} does not exist",
+            status_code=404,
+        )
+
+    # Parse cursor if provided
+    cursor_run_rowid: Optional[int] = None
+    if cursor:
+        try:
+            cursor_gid = GlobalID.from_id(cursor)
+            cursor_run_rowid = from_global_id_with_expected_type(cursor_gid, "ExperimentRun")
+        except (ValueError, AttributeError):
+            raise HTTPException(
+                detail=f"Invalid cursor format: {cursor}",
+                status_code=422,
+            )
+
+    # Deduplicate evaluation names
+    evaluation_name = list(set(name.strip() for name in evaluation_name if name.strip()))
+
+    # Require at least one evaluation name
+    if not evaluation_name:
+        raise HTTPException(
+            detail="At least one evaluation_name must be provided",
+            status_code=400,
+        )
+
+    # Validate evaluation names - reject null bytes which are invalid in PostgreSQL
+    for name in evaluation_name:
+        if "\x00" in name:
+            raise HTTPException(
+                detail="Invalid evaluation name: null bytes are not allowed",
+                status_code=400,
+            )
+
+    async with request.app.state.db() as session:
+        # Verify experiment exists
+        experiment_result = await session.execute(
+            select(models.Experiment).filter_by(id=experiment_rowid)
         )
-        experiment_runs = experiment_runs.scalars().all()
-        runs = []
-        for exp_run in experiment_runs:
-            run_gid = GlobalID("ExperimentRun", str(exp_run.id))
-            experiment_gid = GlobalID("Experiment", str(exp_run.experiment_id))
-            example_gid = GlobalID("DatasetExample", str(exp_run.dataset_example_id))
-            runs.append(
-                ExperimentRunResponse(
-                    start_time=exp_run.start_time,
-                    end_time=exp_run.end_time,
-                    experiment_id=str(experiment_gid),
-                    dataset_example_id=str(example_gid),
-                    repetition_number=exp_run.repetition_number,
-                    output=exp_run.output.get("task_output"),
-                    error=exp_run.error,
-                    id=str(run_gid),
-                    trace_id=exp_run.trace_id,
+        experiment = experiment_result.scalar()
+        if not experiment:
+            raise HTTPException(
+                detail=f"Experiment with ID {experiment_globalid} does not exist",
+                status_code=404,
+            )
+
+        # Query for runs with incomplete evaluations in a single query
+        # This fetches runs, revisions, and annotations together to minimize round-trips
+        # A run has incomplete evaluations if:
+        # 1. It's missing an annotation for any of the requested evaluators
+        # 2. It has a failed annotation (error IS NOT NULL) for any evaluator
+
+        # Get dialect for SQL generation
+        dialect = request.app.state.db.dialect
+
+        # Single query: Get runs with incomplete evaluations + their revisions + annotations
+        combined_query = get_runs_with_incomplete_evaluations_query(
+            experiment_rowid,
+            evaluation_name,
+            dialect,
+            cursor_run_rowid=cursor_run_rowid,
+            limit=limit,
+            include_annotations_and_revisions=True,
+        )
+
+        combined_result = await session.execute(combined_query)
+        all_rows = combined_result.all()
+
+        if not all_rows:
+            return GetIncompleteEvaluationsResponseBody(data=[], next_cursor=None)
+
+        # Parse rows - now each row is a single run with successful annotations as JSON array
+        # Each row: (ExperimentRun, revision_id, DatasetExampleRevision, annotations_json)
+        runs_data: list[tuple[models.ExperimentRun, models.DatasetExampleRevision, set[str]]] = []
+
+        for row in all_rows:
+            run = row[0]  # ExperimentRun
+            revision = row[2]  # DatasetExampleRevision
+            annotations_json = row[3]  # JSON string or None
+
+            # Parse successful annotation names (just a list of strings now)
+            successful_eval_names: set[str] = set()
+            if annotations_json:
+                successful_eval_names = set(json.loads(annotations_json))
+
+            runs_data.append((run, revision, successful_eval_names))
+
+        # Apply pagination limit
+        has_more = len(runs_data) > limit
+        if has_more:
+            runs_to_process = runs_data[:limit]
+        else:
+            runs_to_process = runs_data
+
+        # Build response
+        incomplete_evaluations_list: list[IncompleteExperimentEvaluation] = []
+        for run, revision, successful_eval_names in runs_to_process:
+            # Determine incomplete evaluation names for this run
+            # Any evaluation not in the successful set is incomplete (either missing or failed)
+            incomplete_evaluation_names = sorted(
+                name for name in evaluation_name if name not in successful_eval_names
+            )
+
+            run_globalid = GlobalID("ExperimentRun", str(run.id))
+            example_globalid = GlobalID("DatasetExample", str(run.dataset_example_id))
+
+            incomplete_evaluations_list.append(
+                IncompleteExperimentEvaluation(
+                    experiment_run=ExperimentRun(
+                        id=str(run_globalid),
+                        experiment_id=str(experiment_globalid),
+                        dataset_example_id=str(example_globalid),
+                        output=run.output.get("task_output"),
+                        repetition_number=run.repetition_number,
+                        start_time=run.start_time,
+                        end_time=run.end_time,
+                        trace_id=run.trace_id,
+                        error=run.error,
+                    ),
+                    dataset_example=DatasetExample(
+                        id=str(example_globalid),
+                        input=revision.input,
+                        output=revision.output,
+                        metadata=revision.metadata_,
+                        updated_at=revision.created_at,
+                    ),
+                    evaluation_names=incomplete_evaluation_names,
                 )
             )
-    return ListExperimentRunsResponseBody(data=runs)
+
+        # Set next cursor if we have more results
+        next_cursor = None
+        if has_more:
+            # Cursor is the ID of the next item to fetch
+            # (the extra item we fetched but didn't process)
+            next_run, _, _ = runs_data[limit]  # First item after our limit
+            next_cursor = str(GlobalID("ExperimentRun", str(next_run.id)))
+
+        return GetIncompleteEvaluationsResponseBody(
+            data=incomplete_evaluations_list, next_cursor=next_cursor
+        )