arize-phoenix 4.22.0__py3-none-any.whl → 4.23.0__py3-none-any.whl

phoenix/server/api/dataloaders/evaluation_summaries.py

@@ -1,149 +0,0 @@
-from collections import defaultdict
-from datetime import datetime
-from typing import (
-    Any,
-    DefaultDict,
-    List,
-    Literal,
-    Optional,
-    Tuple,
-)
-
-import pandas as pd
-from aioitertools.itertools import groupby
-from cachetools import LFUCache, TTLCache
-from sqlalchemy import Select, func, or_, select
-from strawberry.dataloader import AbstractCache, DataLoader
-from typing_extensions import TypeAlias, assert_never
-
-from phoenix.db import models
-from phoenix.server.api.dataloaders.cache import TwoTierCache
-from phoenix.server.api.input_types.TimeRange import TimeRange
-from phoenix.server.api.types.EvaluationSummary import EvaluationSummary
-from phoenix.server.types import DbSessionFactory
-from phoenix.trace.dsl import SpanFilter
-
-Kind: TypeAlias = Literal["span", "trace"]
-ProjectRowId: TypeAlias = int
-TimeInterval: TypeAlias = Tuple[Optional[datetime], Optional[datetime]]
-FilterCondition: TypeAlias = Optional[str]
-EvalName: TypeAlias = str
-
-Segment: TypeAlias = Tuple[Kind, ProjectRowId, TimeInterval, FilterCondition]
-Param: TypeAlias = EvalName
-
-Key: TypeAlias = Tuple[Kind, ProjectRowId, Optional[TimeRange], FilterCondition, EvalName]
-Result: TypeAlias = Optional[EvaluationSummary]
-ResultPosition: TypeAlias = int
-DEFAULT_VALUE: Result = None
-
-
-def _cache_key_fn(key: Key) -> Tuple[Segment, Param]:
-    kind, project_rowid, time_range, filter_condition, eval_name = key
-    interval = (
-        (time_range.start, time_range.end) if isinstance(time_range, TimeRange) else (None, None)
-    )
-    return (kind, project_rowid, interval, filter_condition), eval_name
-
-
-_Section: TypeAlias = Tuple[ProjectRowId, EvalName, Kind]
-_SubKey: TypeAlias = Tuple[TimeInterval, FilterCondition]
-
-
-class EvaluationSummaryCache(
-    TwoTierCache[Key, Result, _Section, _SubKey],
-):
-    def __init__(self) -> None:
-        super().__init__(
-            # TTL=3600 (1-hour) because time intervals are always moving forward, but
-            # interval endpoints are rounded down to the hour by the UI, so anything
-            # older than an hour most likely won't be a cache-hit anyway.
-            main_cache=TTLCache(maxsize=64 * 32 * 2, ttl=3600),
-            sub_cache_factory=lambda: LFUCache(maxsize=2 * 2),
-        )
-
-    def invalidate_project(self, project_rowid: ProjectRowId) -> None:
-        for section in self._cache.keys():
-            if section[0] == project_rowid:
-                del self._cache[section]
-
-    def _cache_key(self, key: Key) -> Tuple[_Section, _SubKey]:
-        (kind, project_rowid, interval, filter_condition), eval_name = _cache_key_fn(key)
-        return (project_rowid, eval_name, kind), (interval, filter_condition)
-
-
-class EvaluationSummaryDataLoader(DataLoader[Key, Result]):
-    def __init__(
-        self,
-        db: DbSessionFactory,
-        cache_map: Optional[AbstractCache[Key, Result]] = None,
-    ) -> None:
-        super().__init__(
-            load_fn=self._load_fn,
-            cache_key_fn=_cache_key_fn,
-            cache_map=cache_map,
-        )
-        self._db = db
-
-    async def _load_fn(self, keys: List[Key]) -> List[Result]:
-        results: List[Result] = [DEFAULT_VALUE] * len(keys)
-        arguments: DefaultDict[
-            Segment,
-            DefaultDict[Param, List[ResultPosition]],
-        ] = defaultdict(lambda: defaultdict(list))
-        for position, key in enumerate(keys):
-            segment, param = _cache_key_fn(key)
-            arguments[segment][param].append(position)
-        for segment, params in arguments.items():
-            stmt = _get_stmt(segment, *params.keys())
-            async with self._db() as session:
-                data = await session.stream(stmt)
-                async for eval_name, group in groupby(data, lambda row: row.name):
-                    summary = EvaluationSummary(pd.DataFrame(group))
-                    for position in params[eval_name]:
-                        results[position] = summary
-        return results
-
-
-def _get_stmt(
-    segment: Segment,
-    *eval_names: Param,
-) -> Select[Any]:
-    kind, project_rowid, (start_time, end_time), filter_condition = segment
-    stmt = select()
-    if kind == "span":
-        msa = models.SpanAnnotation
-        name_column, label_column, score_column = msa.name, msa.label, msa.score
-        annotator_kind_column = msa.annotator_kind
-        time_column = models.Span.start_time
-        stmt = stmt.join(models.Span).join_from(models.Span, models.Trace)
-        if filter_condition:
-            sf = SpanFilter(filter_condition)
-            stmt = sf(stmt)
-    elif kind == "trace":
-        mta = models.TraceAnnotation
-        name_column, label_column, score_column = mta.name, mta.label, mta.score
-        annotator_kind_column = mta.annotator_kind
-        time_column = models.Trace.start_time
-        stmt = stmt.join(models.Trace)
-    else:
-        assert_never(kind)
-    stmt = stmt.add_columns(
-        name_column,
-        label_column,
-        func.count().label("record_count"),
-        func.count(label_column).label("label_count"),
-        func.count(score_column).label("score_count"),
-        func.sum(score_column).label("score_sum"),
-    )
-    stmt = stmt.group_by(name_column, label_column)
-    stmt = stmt.order_by(name_column, label_column)
-    stmt = stmt.where(models.Trace.project_rowid == project_rowid)
-    stmt = stmt.where(annotator_kind_column == "LLM")
-    stmt = stmt.where(or_(score_column.is_not(None), label_column.is_not(None)))
-    stmt = stmt.where(name_column.in_(eval_names))
-    if start_time:
-        stmt = stmt.where(start_time <= time_column)
-    if end_time:
-        stmt = stmt.where(time_column < end_time)
-    return stmt

phoenix/server/api/dataloaders/span_evaluations.py

@@ -1,35 +0,0 @@
-from collections import defaultdict
-from typing import (
-    DefaultDict,
-    List,
-)
-
-from sqlalchemy import select
-from strawberry.dataloader import DataLoader
-from typing_extensions import TypeAlias
-
-from phoenix.db import models
-from phoenix.server.api.types.Evaluation import SpanEvaluation
-from phoenix.server.types import DbSessionFactory
-
-Key: TypeAlias = int
-Result: TypeAlias = List[SpanEvaluation]
-
-
-class SpanEvaluationsDataLoader(DataLoader[Key, Result]):
-    def __init__(self, db: DbSessionFactory) -> None:
-        super().__init__(load_fn=self._load_fn)
-        self._db = db
-
-    async def _load_fn(self, keys: List[Key]) -> List[Result]:
-        span_evaluations_by_id: DefaultDict[Key, Result] = defaultdict(list)
-        msa = models.SpanAnnotation
-        async with self._db() as session:
-            data = await session.stream_scalars(
-                select(msa).where(msa.span_rowid.in_(keys)).where(msa.annotator_kind == "LLM")
-            )
-            async for span_evaluation in data:
-                span_evaluations_by_id[span_evaluation.span_rowid].append(
-                    SpanEvaluation.from_sql_span_annotation(span_evaluation)
-                )
-        return [span_evaluations_by_id[key] for key in keys]

phoenix/server/api/dataloaders/trace_evaluations.py

@@ -1,35 +0,0 @@
-from collections import defaultdict
-from typing import (
-    DefaultDict,
-    List,
-)
-
-from sqlalchemy import select
-from strawberry.dataloader import DataLoader
-from typing_extensions import TypeAlias
-
-from phoenix.db import models
-from phoenix.server.api.types.Evaluation import TraceEvaluation
-from phoenix.server.types import DbSessionFactory
-
-Key: TypeAlias = int
-Result: TypeAlias = List[TraceEvaluation]
-
-
-class TraceEvaluationsDataLoader(DataLoader[Key, Result]):
-    def __init__(self, db: DbSessionFactory) -> None:
-        super().__init__(load_fn=self._load_fn)
-        self._db = db
-
-    async def _load_fn(self, keys: List[Key]) -> List[Result]:
-        trace_evaluations_by_id: DefaultDict[Key, Result] = defaultdict(list)
-        mta = models.TraceAnnotation
-        async with self._db() as session:
-            data = await session.stream_scalars(
-                select(mta).where(mta.trace_rowid.in_(keys)).where(mta.annotator_kind == "LLM")
-            )
-            async for trace_evaluation in data:
-                trace_evaluations_by_id[trace_evaluation.trace_rowid].append(
-                    TraceEvaluation.from_sql_trace_annotation(trace_evaluation)
-                )
-        return [trace_evaluations_by_id[key] for key in keys]

phoenix/server/static/assets/index-CZg-95kd.js

@@ -1,100 +0,0 @@
-import{r as d,j as e,d2 as F,v as s,F as P,R as v,w as E,aN as L,d3 as R,d4 as S,d5 as a,d6 as w,d7 as z,b as A,d8 as j}from"./vendor-BMWfu6zp.js";import{S as C,j as k,Z as $,U as _,t as I,a4 as O}from"./vendor-arizeai-Sj74jm5V.js";import{b2 as T,d as D,R as N,b3 as G,b4 as M}from"./components-Bhx3QVW0.js";import{L as U,E as B,h as q,M as J,a as m,D as K,d as W,b as H,e as V,P as Y,c as Z,T as Q,p as X,f as u,g as ee,i as re,j as g,k as ae,l as h,m as x,n as oe,o as te,q as ne,r as se,s as le,A as ie}from"./pages-DG-5zgoV.js";import"./vendor-three-DwGkEfCM.js";import"./vendor-codemirror-DO3VqEcD.js";import"./vendor-recharts-BGN0SxgJ.js";(function(){const n=document.createElement("link").relList;if(n&&n.supports&&n.supports("modulepreload"))return;for(const o of document.querySelectorAll('link[rel="modulepreload"]'))c(o);new MutationObserver(o=>{for(const t of o)if(t.type==="childList")for(const l of t.addedNodes)l.tagName==="LINK"&&l.rel==="modulepreload"&&c(l)}).observe(document,{childList:!0,subtree:!0});function i(o){const t={};return o.integrity&&(t.integrity=o.integrity),o.referrerPolicy&&(t.referrerPolicy=o.referrerPolicy),o.crossOrigin==="use-credentials"?t.credentials="include":o.crossOrigin==="anonymous"?t.credentials="omit":t.credentials="same-origin",t}function c(o){if(o.ep)return;o.ep=!0;const t=i(o);fetch(o.href,t)}})();const b="arize-phoenix-feature-flags",p={__CLEAR__:!0};function ce(){const r=localStorage.getItem(b);if(!r)return p;try{const n=JSON.parse(r);return Object.assign({},p,n)}catch{return p}}const f=d.createContext(null);function de(){const r=v.useContext(f);if(r===null)throw new Error("useFeatureFlags must be used within a FeatureFlagsProvider");return r}function pe(r){const[n,i]=d.useState(ce()),c=o=>{localStorage.setItem(b,JSON.stringify(o)),i(o)};return e(f.Provider,{value:{featureFlags:n,setFeatureFlags:c},children:e(me,{children:r.children})})}function me(r){const{children:n}=r,{featureFlags:i,setFeatureFlags:c}=de(),[o,t]=d.useState(!1);return F("ctrl+shift+f",()=>t(!0)),s(P,{children:[n,e(_,{type:"modal",isDismissable:!0,onDismiss:()=>t(!1),children:o&&e(C,{title:"Feature Flags",children:e(k,{height:"size-1000",padding:"size-100",children:Object.keys(i).map(l=>e($,{isSelected:i[l],onChange:y=>c({...i,[l]:y}),children:l},l))})})})]})}function ue(){return e(L,{styles:r=>E`
-        body {
-          background-color: var(--ac-global-color-grey-75);
-          color: var(--ac-global-text-color-900);
-          font-family: "Roboto";
-          font-size: ${r.typography.sizes.medium.fontSize}px;
-          margin: 0;
-          #root,
-          #root > div[data-overlay-container="true"],
-          #root > div[data-overlay-container="true"] > .ac-theme {
-            height: 100vh;
-          }
-        }
-
-        /* Remove list styling */
-        ul {
-          display: block;
-          list-style-type: none;
-          margin-block-start: none;
-          margin-block-end: 0;
-          padding-inline-start: 0;
-          margin-block-start: 0;
-        }
-
-        /* A reset style for buttons */
-        .button--reset {
-          background: none;
-          border: none;
-          padding: 0;
-        }
-        /* this css class is added to html via modernizr @see modernizr.js */
-        .no-hiddenscroll {
-          /* Works on Firefox */
-          * {
-            scrollbar-width: thin;
-            scrollbar-color: var(--ac-global-color-grey-300)
-              var(--ac-global-color-grey-400);
-          }
-
-          /* Works on Chrome, Edge, and Safari */
-          *::-webkit-scrollbar {
-            width: 14px;
-          }
-
-          *::-webkit-scrollbar-track {
-            background: var(--ac-global-color-grey-100);
-          }
-
-          *::-webkit-scrollbar-thumb {
-            background-color: var(--ac-global-color-grey-75);
-            border-radius: 8px;
-            border: 1px solid var(--ac-global-color-grey-300);
-          }
-        }
-
-        :root {
-          --px-blue-color: ${r.colors.arizeBlue};
-
-          --px-flex-gap-sm: ${r.spacing.margin4}px;
-          --px-flex-gap-sm: ${r.spacing.margin8}px;
-
-          --px-section-background-color: ${r.colors.gray500};
-
-          /* An item is a typically something in a list */
-          --px-item-background-color: ${r.colors.gray800};
-          --px-item-border-color: ${r.colors.gray600};
-
-          --px-spacing-sm: ${r.spacing.padding4}px;
-          --px-spacing-med: ${r.spacing.padding8}px;
-          --px-spacing-lg: ${r.spacing.padding16}px;
-
-          --px-border-radius-med: ${r.borderRadius.medium}px;
-
-          --px-font-size-sm: ${r.typography.sizes.small.fontSize}px;
-          --px-font-size-med: ${r.typography.sizes.medium.fontSize}px;
-          --px-font-size-lg: ${r.typography.sizes.large.fontSize}px;
-
-          --px-gradient-bar-height: 8px;
-
-          --px-nav-collapsed-width: 45px;
-          --px-nav-expanded-width: 200px;
-        }
-
-        .ac-theme--dark {
-          --px-primary-color: #9efcfd;
-          --px-primary-color--transparent: rgb(158, 252, 253, 0.2);
-          --px-reference-color: #baa1f9;
-          --px-reference-color--transparent: #baa1f982;
-          --px-corpus-color: #92969c;
-          --px-corpus-color--transparent: #92969c63;
-        }
-        .ac-theme--light {
-          --px-primary-color: #00add0;
-          --px-primary-color--transparent: rgba(0, 173, 208, 0.2);
-          --px-reference-color: #4500d9;
-          --px-reference-color--transparent: rgba(69, 0, 217, 0.2);
-          --px-corpus-color: #92969c;
-          --px-corpus-color--transparent: #92969c63;
-        }
-      `})}const ge=R(S(s(a,{path:"/",element:e(U,{}),errorElement:e(B,{}),children:[e(a,{index:!0,loader:q}),s(a,{path:"/model",handle:{crumb:()=>"model"},element:e(J,{}),children:[e(a,{index:!0,element:e(m,{})}),e(a,{element:e(m,{}),children:e(a,{path:"dimensions",children:e(a,{path:":dimensionId",element:e(K,{}),loader:W})})}),e(a,{path:"embeddings",children:e(a,{path:":embeddingDimensionId",element:e(H,{}),loader:V,handle:{crumb:r=>r.embedding.name}})})]}),s(a,{path:"/projects",handle:{crumb:()=>"projects"},element:e(Y,{}),children:[e(a,{index:!0,element:e(Z,{})}),s(a,{path:":projectId",element:e(Q,{}),loader:X,handle:{crumb:r=>r.project.name},children:[e(a,{index:!0,element:e(u,{})}),e(a,{element:e(u,{}),children:e(a,{path:"traces/:traceId",element:e(ee,{})})})]})]}),s(a,{path:"/datasets",handle:{crumb:()=>"datasets"},children:[e(a,{index:!0,element:e(re,{})}),s(a,{path:":datasetId",loader:g,handle:{crumb:r=>r.dataset.name},children:[s(a,{element:e(ae,{}),loader:g,children:[e(a,{index:!0,element:e(h,{}),loader:x}),e(a,{path:"experiments",element:e(h,{}),loader:x}),e(a,{path:"examples",element:e(oe,{}),loader:te,children:e(a,{path:":exampleId",element:e(ne,{})})})]}),e(a,{path:"compare",handle:{crumb:()=>"compare"},loader:se,element:e(le,{})})]})]}),e(a,{path:"/apis",element:e(ie,{}),handle:{crumb:()=>"APIs"}})]})),{basename:window.Config.basename});function he(){return e(w,{router:ge})}function xe(){return e(T,{children:e(be,{})})}function be(){const{theme:r}=D();return e(O,{theme:r,children:e(z,{theme:I,children:s(A.RelayEnvironmentProvider,{environment:N,children:[e(ue,{}),e(pe,{children:e(G,{children:e(d.Suspense,{children:e(M,{children:e(he,{})})})})})]})})})}const fe=document.getElementById("root"),ye=j.createRoot(fe);ye.render(e(xe,{}));