Compare commits

..

1 Commits

Author SHA1 Message Date
nityanandagohain
5dae0b975a feat: add trace summary endpoint 2026-09-18 17:55:43 +05:30
25 changed files with 966 additions and 252 deletions

View File

@@ -67,6 +67,7 @@ jobs:
- semconvfamilies
- serviceaccount
- spanmapper
- tracedetail
- querier_json_body
- querier_skip_resource_fingerprint
- ttl

View File

@@ -9439,6 +9439,29 @@ components:
required:
- aggregations
type: object
SpantypesGettableTraceSummary:
properties:
ai:
$ref: '#/components/schemas/SpantypesTraceAISummary'
endTimestampMillis:
minimum: 0
type: integer
hasMissingSpans:
type: boolean
rootServiceEntryPoint:
type: string
rootServiceName:
type: string
startTimestampMillis:
minimum: 0
type: integer
totalErrorSpansCount:
minimum: 0
type: integer
totalSpansCount:
minimum: 0
type: integer
type: object
SpantypesGettableWaterfallTrace:
properties:
endTimestampMillis:
@@ -9722,6 +9745,32 @@ components:
nullable: true
type: object
type: object
SpantypesTraceAISummary:
properties:
tokens:
$ref: '#/components/schemas/SpantypesTraceAITokens'
totalCost:
nullable: true
type: number
type: object
SpantypesTraceAITokens:
properties:
cacheRead:
minimum: 0
type: integer
cacheWrite:
minimum: 0
type: integer
input:
minimum: 0
type: integer
output:
minimum: 0
type: integer
reasoning:
minimum: 0
type: integer
type: object
SpantypesUpdatableSpanMapper:
properties:
config:
@@ -15460,6 +15509,66 @@ paths:
summary: Get aggregations for a trace
tags:
- tracedetail
/api/v1/traces/{traceID}/summary:
get:
deprecated: false
description: Returns the trace-level fields of the waterfall (time range, root,
span counts, missing spans) and, when the trace has gen_ai spans, its token
and cost totals. Computed in one aggregate query.
operationId: GetTraceSummary
parameters:
- in: path
name: traceID
required: true
schema:
type: string
responses:
"200":
content:
application/json:
schema:
properties:
data:
$ref: '#/components/schemas/SpantypesGettableTraceSummary'
status:
type: string
required:
- status
- data
type: object
description: OK
"401":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Unauthorized
"403":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Forbidden
"404":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Not Found
"500":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Internal Server Error
security:
- api_key:
- VIEWER
- tokenizer:
- VIEWER
summary: Get summary for a trace
tags:
- tracedetail
/api/v1/user/me:
get:
deprecated: true

View File

@@ -10884,6 +10884,78 @@ export interface SpantypesGettableTraceAggregationsDTO {
aggregations: SpantypesSpanAggregationResultDTO[];
}
export interface SpantypesTraceAITokensDTO {
/**
* @type integer
* @minimum 0
*/
cacheRead?: number;
/**
* @type integer
* @minimum 0
*/
cacheWrite?: number;
/**
* @type integer
* @minimum 0
*/
input?: number;
/**
* @type integer
* @minimum 0
*/
output?: number;
/**
* @type integer
* @minimum 0
*/
reasoning?: number;
}
export interface SpantypesTraceAISummaryDTO {
tokens?: SpantypesTraceAITokensDTO;
/**
* @type number,null
*/
totalCost?: number | null;
}
export interface SpantypesGettableTraceSummaryDTO {
ai?: SpantypesTraceAISummaryDTO;
/**
* @type integer
* @minimum 0
*/
endTimestampMillis?: number;
/**
* @type boolean
*/
hasMissingSpans?: boolean;
/**
* @type string
*/
rootServiceEntryPoint?: string;
/**
* @type string
*/
rootServiceName?: string;
/**
* @type integer
* @minimum 0
*/
startTimestampMillis?: number;
/**
* @type integer
* @minimum 0
*/
totalErrorSpansCount?: number;
/**
* @type integer
* @minimum 0
*/
totalSpansCount?: number;
}
export interface SpantypesOtelSpanRefDTO {
/**
* @type string
@@ -12595,6 +12667,17 @@ export type GetTraceAggregations200 = {
status: string;
};
export type GetTraceSummaryPathParameters = {
traceID: string;
};
export type GetTraceSummary200 = {
data: SpantypesGettableTraceSummaryDTO;
/**
* @type string
*/
status: string;
};
export type ListUserPreferences200 = {
/**
* @type array

View File

@@ -4,11 +4,17 @@
* * regenerate with 'pnpm generate:api'
* SigNoz
*/
import { useMutation } from 'react-query';
import { useMutation, useQuery } from 'react-query';
import type {
InvalidateOptions,
MutationFunction,
QueryClient,
QueryFunction,
QueryKey,
UseMutationOptions,
UseMutationResult,
UseQueryOptions,
UseQueryResult,
} from 'react-query';
import type {
@@ -16,6 +22,8 @@ import type {
GetFlamegraphPathParameters,
GetTraceAggregations200,
GetTraceAggregationsPathParameters,
GetTraceSummary200,
GetTraceSummaryPathParameters,
GetWaterfallV4200,
GetWaterfallV4PathParameters,
RenderErrorResponseDTO,
@@ -27,6 +35,26 @@ import type {
import { GeneratedAPIInstance } from '../../../generatedAPIInstance';
import type { ErrorType, BodyType } from '../../../generatedAPIInstance';
const withQueryKey = <T extends object, K>(
query: T,
queryKey: K,
): T & { queryKey: K } => {
const result = { queryKey } as T & { queryKey: K };
for (const key of Object.keys(query)) {
// The explicit queryKey always wins, matching the previous
// `{ ...query, queryKey }` spread where it was set last.
if (key === 'queryKey') {
continue;
}
Object.defineProperty(result, key, {
enumerable: true,
configurable: true,
get: () => (query as Record<string, unknown>)[key],
});
}
return result;
};
/**
* Computes span aggregations grouped by requested field.
* @summary Get aggregations for a trace
@@ -127,6 +155,108 @@ export const useGetTraceAggregations = <
> => {
return useMutation(getGetTraceAggregationsMutationOptions(options));
};
/**
* Returns the trace-level fields of the waterfall (time range, root, span counts, missing spans) and, when the trace has gen_ai spans, its token and cost totals. Computed in one aggregate query.
* @summary Get summary for a trace
*/
export const getTraceSummary = (
{ traceID }: GetTraceSummaryPathParameters,
signal?: AbortSignal,
) => {
return GeneratedAPIInstance<GetTraceSummary200>({
url: `/api/v1/traces/${traceID}/summary`,
method: 'GET',
signal,
});
};
export const getGetTraceSummaryQueryKey = ({
traceID,
}: GetTraceSummaryPathParameters) => {
return [`/api/v1/traces/${traceID}/summary`] as const;
};
export const getGetTraceSummaryQueryOptions = <
TData = Awaited<ReturnType<typeof getTraceSummary>>,
TError = ErrorType<RenderErrorResponseDTO>,
>(
{ traceID }: GetTraceSummaryPathParameters,
options?: {
query?: UseQueryOptions<
Awaited<ReturnType<typeof getTraceSummary>>,
TError,
TData
>;
},
) => {
const { query: queryOptions } = options ?? {};
const queryKey =
queryOptions?.queryKey ?? getGetTraceSummaryQueryKey({ traceID });
const queryFn: QueryFunction<Awaited<ReturnType<typeof getTraceSummary>>> = ({
signal,
}) => getTraceSummary({ traceID }, signal);
return {
queryKey,
queryFn,
enabled: traceID !== null && traceID !== undefined,
...queryOptions,
} as UseQueryOptions<
Awaited<ReturnType<typeof getTraceSummary>>,
TError,
TData
> & { queryKey: QueryKey };
};
export type GetTraceSummaryQueryResult = NonNullable<
Awaited<ReturnType<typeof getTraceSummary>>
>;
export type GetTraceSummaryQueryError = ErrorType<RenderErrorResponseDTO>;
/**
* @summary Get summary for a trace
*/
export function useGetTraceSummary<
TData = Awaited<ReturnType<typeof getTraceSummary>>,
TError = ErrorType<RenderErrorResponseDTO>,
>(
{ traceID }: GetTraceSummaryPathParameters,
options?: {
query?: UseQueryOptions<
Awaited<ReturnType<typeof getTraceSummary>>,
TError,
TData
>;
},
): UseQueryResult<TData, TError> & { queryKey: QueryKey } {
const queryOptions = getGetTraceSummaryQueryOptions({ traceID }, options);
const query = useQuery(queryOptions) as UseQueryResult<TData, TError> & {
queryKey: QueryKey;
};
return withQueryKey(query, queryOptions.queryKey);
}
/**
* @summary Get summary for a trace
*/
export const invalidateGetTraceSummary = async (
queryClient: QueryClient,
{ traceID }: GetTraceSummaryPathParameters,
options?: InvalidateOptions,
): Promise<QueryClient> => {
await queryClient.invalidateQueries(
{ queryKey: getGetTraceSummaryQueryKey({ traceID }) },
options,
);
return queryClient;
};
/**
* Returns the flamegraph view of spans for a given trace ID.
* @summary Get flamegraph view for a trace

View File

@@ -25,7 +25,6 @@
.uplotTooltipItemContent {
width: 100%;
min-width: 0;
display: flex;
align-items: center;
gap: var(--spacing-2);
@@ -41,24 +40,16 @@
}
.uplotTooltipItemLabel {
min-width: 0;
// Not `anywhere`, which drops min-content to one character and lets a wide
// value squeeze the label into a mid-word break.
white-space: normal;
overflow-wrap: break-word;
display: -webkit-box;
-webkit-line-clamp: 3;
-webkit-box-orient: vertical;
overflow: hidden;
overflow-wrap: anywhere;
}
.uplotTooltipItemValue {
white-space: nowrap;
flex: 0 0 auto;
}
.uplotTooltipItemContentSeparator {
flex: 1 1 24px;
flex: 1;
border-width: 0.5px;
border-style: dashed;
min-width: 24px;

View File

@@ -34,9 +34,7 @@ export default function TooltipItem({
style={{ color: item.color }}
data-testid={contentTestId}
>
<span className={Styles.uplotTooltipItemLabel} title={item.label}>
{item.label}
</span>
<span className={Styles.uplotTooltipItemLabel}>{item.label}</span>
<span
className={Styles.uplotTooltipItemContentSeparator}
style={{ borderColor: item.color }}

View File

@@ -17,8 +17,7 @@ import { ChartWrapperProps } from 'lib/visualization/charts/types';
import { useChartStacking } from 'lib/visualization/charts/ChartWrapper/useChartStacking';
const TOOLTIP_WIDTH_PADDING = 120;
// Holds a tooltip row's value column next to a legend-length label.
const TOOLTIP_MIN_WIDTH = 360;
const TOOLTIP_MIN_WIDTH = 300;
export default function ChartWrapper({
legendConfig = { position: LegendPosition.BOTTOM },

View File

@@ -10,6 +10,23 @@ import (
)
func (provider *provider) addTraceDetailRoutes(router *mux.Router) error {
if err := router.Handle("/api/v1/traces/{traceID}/summary", handler.New(
provider.authzMiddleware.ViewAccess(provider.traceDetailHandler.GetTraceSummary),
handler.OpenAPIDef{
ID: "GetTraceSummary",
Tags: []string{"tracedetail"},
Summary: "Get summary for a trace",
Description: "Returns the trace-level fields of the waterfall (time range, root, span counts, missing spans) and, when the trace has gen_ai spans, its token and cost totals. Computed in one aggregate query.",
Response: new(spantypes.GettableTraceSummary),
ResponseContentType: "application/json",
SuccessStatusCode: http.StatusOK,
ErrorStatusCodes: []int{http.StatusNotFound},
SecuritySchemes: newSecuritySchemes(types.RoleViewer),
},
)).Methods(http.MethodGet).GetError(); err != nil {
return err
}
if err := router.Handle("/api/v4/traces/{traceID}/waterfall", handler.New(
provider.authzMiddleware.ViewAccess(provider.traceDetailHandler.GetWaterfallV4),
handler.OpenAPIDef{

View File

@@ -6,7 +6,9 @@ import (
"github.com/SigNoz/signoz/pkg/http/binding"
"github.com/SigNoz/signoz/pkg/http/render"
"github.com/SigNoz/signoz/pkg/modules/tracedetail"
"github.com/SigNoz/signoz/pkg/types/authtypes"
"github.com/SigNoz/signoz/pkg/types/spantypes"
"github.com/SigNoz/signoz/pkg/valuer"
"github.com/gorilla/mux"
)
@@ -18,6 +20,27 @@ func NewHandler(module tracedetail.Module) tracedetail.Handler {
return &handler{module: module}
}
func (h *handler) GetTraceSummary(rw http.ResponseWriter, r *http.Request) {
claims, err := authtypes.ClaimsFromContext(r.Context())
if err != nil {
render.Error(rw, err)
return
}
orgID, err := valuer.NewUUID(claims.OrgID)
if err != nil {
render.Error(rw, err)
return
}
stats, err := h.module.GetTraceStats(r.Context(), orgID, mux.Vars(r)["traceID"])
if err != nil {
render.Error(rw, err)
return
}
render.Success(rw, http.StatusOK, spantypes.NewGettableTraceSummary(stats))
}
func (h *handler) GetWaterfallV4(rw http.ResponseWriter, r *http.Request) {
req := new(spantypes.PostableWaterfall)
if err := binding.JSON.BindBody(r.Body, req); err != nil {

View File

@@ -8,6 +8,7 @@ import (
"github.com/SigNoz/signoz/pkg/modules/tracedetail"
"github.com/SigNoz/signoz/pkg/types/spantypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/valuer"
"go.opentelemetry.io/otel/metric"
)
@@ -39,6 +40,21 @@ func NewModule(traceStore spantypes.TraceStore, providerSettings factory.Provide
return m
}
func (m *module) GetTraceStats(ctx context.Context, orgID valuer.UUID, traceID string) (*spantypes.TraceStats, error) {
summary, err := m.store.GetTraceSummary(ctx, traceID)
if err != nil {
return nil, err
}
stats, err := m.store.GetTraceStats(ctx, orgID, traceID, summary)
if err != nil {
return nil, err
}
if stats.TotalSpans == 0 {
return nil, spantypes.ErrTraceNotFound
}
return stats, nil
}
// GetWaterfallV4 is the OOM-safe V4 waterfall.
// For large traces (NumSpans > effectiveLimit) it uses a two-step fetch:
// minimal fields for all spans to build the tree, then full fields for the

View File

@@ -10,9 +10,15 @@ import (
"github.com/SigNoz/signoz/pkg/clickhousesql"
"github.com/SigNoz/signoz/pkg/errors"
"github.com/SigNoz/signoz/pkg/flagger"
"github.com/SigNoz/signoz/pkg/querybuilder"
"github.com/SigNoz/signoz/pkg/telemetryschema/tracestelemetryschema"
"github.com/SigNoz/signoz/pkg/telemetrystore"
"github.com/SigNoz/signoz/pkg/types/aiobservabilitytypes"
qbtypes "github.com/SigNoz/signoz/pkg/types/querybuildertypes/querybuildertypesv5"
"github.com/SigNoz/signoz/pkg/types/spantypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/valuer"
)
const colServiceName = `resource_string_service$$$$name` // $ gets escaped so $$$$ converts to $$.
@@ -38,10 +44,18 @@ type spanDurationRow struct {
type traceStore struct {
telemetryStore telemetrystore.TelemetryStore
metadataStore telemetrytypes.MetadataStore
storage qbtypes.Storage
flagger flagger.Flagger
}
func NewTraceStore(ts telemetrystore.TelemetryStore) *traceStore {
return &traceStore{telemetryStore: ts}
func NewTraceStore(ts telemetrystore.TelemetryStore, metadataStore telemetrytypes.MetadataStore, fl flagger.Flagger) *traceStore {
return &traceStore{
telemetryStore: ts,
metadataStore: metadataStore,
storage: tracestelemetryschema.NewStorage(),
flagger: fl,
}
}
func (s *traceStore) GetTraceSummary(ctx context.Context, traceID string) (*spantypes.TraceSummary, error) {
@@ -65,6 +79,131 @@ func (s *traceStore) GetTraceSummary(ctx context.Context, traceID string) (*span
return &summary, nil
}
func (s *traceStore) GetTraceStats(ctx context.Context, orgID valuer.UUID, traceID string, summary *spantypes.TraceSummary) (*spantypes.TraceStats, error) {
table := fmt.Sprintf("%s.%s", spantypes.TraceDB, spantypes.TraceTable)
spans := sqlbuilder.NewSelectBuilder()
genAIColumns, err := s.genAISpanColumns(ctx, orgID, summary, spans)
if err != nil {
return nil, err
}
// A span whose parent was never recorded hangs off a synthetic "Missing Span" root in the waterfall.
ids := sqlbuilder.NewSelectBuilder()
ids.Select("span_id")
ids.From(table)
ids.Where(
ids.E("trace_id", traceID),
ids.GE("ts_bucket_start", summary.Start.Unix()-1800),
ids.LE("ts_bucket_start", summary.End.Unix()),
)
missingParent := fmt.Sprintf("parent_span_id <> '' AND parent_span_id GLOBAL NOT IN (%s)", spans.Var(ids))
spans.Select(
"toUnixTimestamp64Nano(timestamp) AS span_start_ns",
"span_start_ns + duration_nano AS span_end_ns",
"span_id",
"has_error",
"("+missingParent+") AS has_missing_parent",
"(parent_span_id = '' OR has_missing_parent) AS is_root",
"if(parent_span_id = '', name, 'Missing Span') AS root_name",
"if(parent_span_id = '', "+colServiceName+", '') AS root_service",
)
spans.SelectMore(genAIColumns...)
spans.From(table)
spans.Where(
spans.E("trace_id", traceID),
spans.GE("ts_bucket_start", summary.Start.Unix()-1800),
spans.LE("ts_bucket_start", summary.End.Unix()),
)
spans.SQL("LIMIT 1 BY span_id")
sb := sqlbuilder.NewSelectBuilder()
sb.Select(
"toUInt64(min(span_start_ns)) AS start_ns",
"toUInt64(max(span_end_ns)) AS end_ns",
"count() AS total_spans",
"countIf(has_error) AS total_error_spans",
"countIf(has_missing_parent) > 0 AS has_missing_spans",
"argMinIf(root_service, (span_start_ns, root_name), is_root) AS root_service_name",
"argMinIf(root_name, (span_start_ns, root_name), is_root) AS root_entry_point",
"countIf(is_gen_ai) AS gen_ai_span_count",
"toUInt64(coalesce(sum(input_tokens_value), 0)) AS input_tokens",
"toUInt64(coalesce(sum(output_tokens_value), 0)) AS output_tokens",
"toUInt64(coalesce(sum(cache_read_tokens_value), 0)) AS cache_read_tokens",
"toUInt64(coalesce(sum(cache_write_tokens_value), 0)) AS cache_write_tokens",
"toUInt64(coalesce(sum(reasoning_tokens_value), 0)) AS reasoning_tokens",
"sum(total_cost_value) AS total_cost",
)
sb.From(sb.BuilderAs(spans, "spans"))
query, args := sb.BuildWithFlavor(sqlbuilder.ClickHouse)
var stats spantypes.TraceStats
err = s.telemetryStore.ClickhouseDB().QueryRow(ctx, query, args...).Scan(
&stats.StartNs, &stats.EndNs, &stats.TotalSpans, &stats.TotalErrorSpans, &stats.HasMissingSpans,
&stats.RootServiceName, &stats.RootEntryPoint, &stats.GenAISpanCount,
&stats.Tokens.Input, &stats.Tokens.Output, &stats.Tokens.CacheRead, &stats.Tokens.CacheWrite, &stats.Tokens.Reasoning,
&stats.TotalCost,
)
if err != nil {
return nil, errors.WrapInternalf(err, errors.CodeInternal, "error querying trace stats")
}
return &stats, nil
}
// genAISpanColumns renders the per-span gen_ai gate and value reads through the shared
// traces storage, so each attribute is read from the column its evolutions place it in
// over the trace's own time window. Exists predicates bind their args into sb.
func (s *traceStore) genAISpanColumns(ctx context.Context, orgID valuer.UUID, summary *spantypes.TraceSummary, sb *sqlbuilder.SelectBuilder) ([]string, error) {
// no data type: metadata reports token counts as number, so a float64 request would
// miss them and fall back to a map read without evolutions
attributeKey := func(name string) *telemetrytypes.TelemetryFieldKey {
return &telemetrytypes.TelemetryFieldKey{Name: name, Signal: telemetrytypes.SignalTraces, FieldContext: telemetrytypes.FieldContextAttribute}
}
selectors := make([]*telemetrytypes.FieldKeySelector, 0, len(aiobservabilitytypes.GenAISpanGateKeys)+len(spantypes.TraceStatsGenAIColumns))
addSelector := func(name string) {
selectors = append(selectors, &telemetrytypes.FieldKeySelector{
Name: name,
Signal: telemetrytypes.SignalTraces,
FieldContext: telemetrytypes.FieldContextAttribute,
SelectorMatchType: telemetrytypes.FieldSelectorMatchTypeExact,
})
}
for _, name := range aiobservabilitytypes.GenAISpanGateKeys {
addSelector(name)
}
for _, col := range spantypes.TraceStatsGenAIColumns {
addSelector(col.Key)
}
keys, _, err := s.metadataStore.GetKeysMulti(ctx, orgID, querybuilder.ExpandKeySelectorsForFamilies(ctx, orgID, s.flagger, selectors))
if err != nil {
return nil, err
}
q := querybuilder.NewQueryInfo(ctx, orgID, s.flagger, telemetrytypes.SignalTraces, nil, uint64(summary.Start.UnixNano()), uint64(summary.End.UnixNano()))
gate := make([]string, 0, len(aiobservabilitytypes.GenAISpanGateKeys))
for _, name := range aiobservabilitytypes.GenAISpanGateKeys {
conds, _, err := querybuilder.Conditions(ctx, q, s.storage, attributeKey(name), qbtypes.FilterOperatorExists, nil, keys, false, sb)
if err != nil {
return nil, err
}
gate = append(gate, conds...)
}
columns := []string{sb.Or(gate...) + " AS is_gen_ai"}
for _, col := range spantypes.TraceStatsGenAIColumns {
expr, err := querybuilder.ResolveColumn(ctx, q, s.storage, attributeKey(col.Key), telemetrytypes.FieldDataTypeFloat64, keys)
if err != nil {
return nil, err
}
// a materialized column name carries `$$`, which Build would otherwise unescape
columns = append(columns, sqlbuilder.Escape(expr)+" AS "+col.Column+"_value")
}
return columns, nil
}
func (s *traceStore) GetTraceSpans(ctx context.Context, traceID string, summary *spantypes.TraceSummary) ([]spantypes.StorableSpan, error) {
// DISTINCT ON (span_id) is ClickHouse-specific syntax not supported by sqlbuilder
query := fmt.Sprintf(`

File diff suppressed because one or more lines are too long

View File

@@ -6,10 +6,12 @@ import (
"github.com/SigNoz/signoz/pkg/types/spantypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/valuer"
)
// Handler exposes HTTP handlers for trace detail APIs.
type Handler interface {
GetTraceSummary(http.ResponseWriter, *http.Request)
GetWaterfallV4(http.ResponseWriter, *http.Request)
GetTraceAggregations(http.ResponseWriter, *http.Request)
GetFlamegraph(http.ResponseWriter, *http.Request)
@@ -17,6 +19,7 @@ type Handler interface {
// Module defines the business logic for trace detail operations.
type Module interface {
GetTraceStats(ctx context.Context, orgID valuer.UUID, traceID string) (*spantypes.TraceStats, error)
GetWaterfallV4(ctx context.Context, traceID string, selectedSpanID string, uncollapsedSpans []string) (*spantypes.GettableWaterfallTrace, error)
GetTraceAggregations(ctx context.Context, traceID string, req *spantypes.PostableTraceAggregations) (*spantypes.GettableTraceAggregations, error)
GetFlamegraph(ctx context.Context, traceID string, selectedSpanID string, selectFields []telemetrytypes.TelemetryFieldKey) (*spantypes.GettableFlamegraphTrace, error)

View File

@@ -25,9 +25,8 @@ const (
// ResolveLogicalFields picks which logical fields a filter term builds conditions
// for. With 0 or 1 field it returns the input unchanged and no warning. When a
// name is ambiguous (several logical fields — a family is one field and never
// ambiguous with itself) it returns a warning; a resource + other-context mix
// (attribute, body, scope, …) defaults to the resource fields (the common
// intent), noted in the warning.
// ambiguous with itself) it returns a warning; a resource+attribute mix defaults
// to the resource fields (the common intent), noted in the warning.
func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields []*telemetrytypes.LogicalField) ([]*telemetrytypes.LogicalField, string) {
if len(logicalFields) <= 1 {
return logicalFields, ""
@@ -40,17 +39,18 @@ func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields
logicalFields,
)
hasResource, hasOther := false, false
hasResource, hasAttribute := false, false
for _, item := range logicalFields {
if item.FieldContext == telemetrytypes.FieldContextResource {
switch item.FieldContext {
case telemetrytypes.FieldContextResource:
hasResource = true
} else {
hasOther = true
case telemetrytypes.FieldContextAttribute:
hasAttribute = true
}
}
// with resource and any other context, default to resource only
if hasResource && hasOther {
// when there is both resource and attribute context, default to resource only
if hasResource && hasAttribute {
filtered := make([]*telemetrytypes.LogicalField, 0, len(logicalFields))
for _, item := range logicalFields {
if item.FieldContext == telemetrytypes.FieldContextResource {
@@ -58,8 +58,8 @@ func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields
}
}
logicalFields = filtered
warning += " " + "Using `resource` context by default. To query another context explicitly, " +
fmt.Sprintf("use the fully qualified name (e.g., 'attribute.%s' or 'body.%s')", field.Name, field.Name)
warning += " " + "Using `resource` context by default. To query attributes explicitly, " +
fmt.Sprintf("use the fully qualified name (e.g., 'attribute.%s')", field.Name)
}
return logicalFields, warning

View File

@@ -175,42 +175,6 @@ func TestResolveLogicalFieldsKeepsFamilyThroughAmbiguity(t *testing.T) {
assert.Equal(t, []string{"deployment.environment.name", "deployment.environment"}, memberNames(resolved[0]))
}
// Resource wins over every other context, not just attribute: a bare key that
// also lives in body or scope must collapse to resource alone, so the surviving
// candidate does not AND against the resource fingerprint CTE.
func TestResolveLogicalFieldsResourceWinsOverOtherContexts(t *testing.T) {
testCases := []struct {
name string
other telemetrytypes.FieldContext
}{
{name: "ResourceOverBody", other: telemetrytypes.FieldContextBody},
{name: "ResourceOverScope", other: telemetrytypes.FieldContextScope},
}
for _, testCase := range testCases {
t.Run(testCase.name, func(t *testing.T) {
requested := &telemetrytypes.TelemetryFieldKey{Name: "service.name"}
fields := []*telemetrytypes.LogicalField{
telemetrytypes.SingleLogicalField("service.name", &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
FieldContext: telemetrytypes.FieldContextResource,
FieldDataType: telemetrytypes.FieldDataTypeString,
}),
telemetrytypes.SingleLogicalField("service.name", &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
FieldContext: testCase.other,
FieldDataType: telemetrytypes.FieldDataTypeString,
}),
}
resolved, warning := ResolveLogicalFields(requested, fields)
assert.NotEmpty(t, warning)
require.Len(t, resolved, 1)
assert.Equal(t, telemetrytypes.FieldContextResource, resolved[0].FieldContext)
})
}
}
// Members of a family with different data types never merge: the identity
// (signal, context, data type) separates them into distinct logical fields.
func TestMatchingLogicalFieldsNeverMergesAcrossDataTypes(t *testing.T) {

View File

@@ -161,7 +161,7 @@ func NewModules(
LogsPipeline: impllogspipeline.NewModule(sqlstore),
RuleStateHistory: implrulestatehistory.NewModule(implrulestatehistory.NewStore(telemetryStore, telemetryMetadataStore, providerSettings.Logger), ruleStore),
CloudIntegration: cloudIntegrationModule,
TraceDetail: impltracedetail.NewModule(impltracedetail.NewTraceStore(telemetryStore), providerSettings, config.TraceDetail),
TraceDetail: impltracedetail.NewModule(impltracedetail.NewTraceStore(telemetryStore, telemetryMetadataStore, fl), providerSettings, config.TraceDetail),
SpanMapper: implspanmapper.NewModule(implspanmapper.NewStore(sqlstore), fl),
LLMPricingRule: impllmpricingrule.NewModule(impllmpricingrule.NewStore(sqlstore), fl, querier),
Tag: tagModule,

View File

@@ -1,90 +0,0 @@
package logsstatementbuilder
import (
"context"
"testing"
"github.com/SigNoz/signoz/pkg/flagger/flaggertest"
"github.com/SigNoz/signoz/pkg/instrumentation/instrumentationtest"
"github.com/SigNoz/signoz/pkg/querybuilder"
"github.com/SigNoz/signoz/pkg/statementbuilder"
"github.com/SigNoz/signoz/pkg/telemetryschema/logstelemetryschema"
qbtypes "github.com/SigNoz/signoz/pkg/types/querybuildertypes/querybuildertypesv5"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes/telemetrytypestest"
"github.com/SigNoz/signoz/pkg/valuer"
"github.com/stretchr/testify/require"
)
// A key present in both resource and body contexts must filter on resource only.
// The resource condition builds the fingerprint CTE, so a surviving body condition
// would AND against it and match almost nothing (engineering-pod#6086).
func TestStatementBuilderResourceBodyConflict(t *testing.T) {
store := telemetrytypestest.NewMockMetadataStore()
store.SetStaticFields(logstelemetryschema.IntrinsicFields)
store.SetKey(&telemetrytypes.TelemetryFieldKey{
Name: "service.name",
Signal: telemetrytypes.SignalLogs,
FieldContext: telemetrytypes.FieldContextResource,
FieldDataType: telemetrytypes.FieldDataTypeString,
})
bodyKey := &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
Signal: telemetrytypes.SignalLogs,
FieldContext: telemetrytypes.FieldContextBody,
FieldDataType: telemetrytypes.FieldDataTypeString,
}
require.NoError(t, bodyKey.SetJSONAccessPlan(telemetrytypes.JSONColumnMetadata{
BaseColumn: logstelemetryschema.LogsV2BodyV2Column,
PromotedColumn: logstelemetryschema.LogsV2BodyPromotedColumn,
}, map[string][]telemetrytypes.FieldDataType{"service.name": {telemetrytypes.FieldDataTypeString}}))
store.SetKey(bodyKey)
fl := flaggertest.WithUseJSONBody(t, true)
storage := logstelemetryschema.NewStorage()
aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, storage, fl, telemetrytypes.SignalLogs)
statementBuilder := NewLogQueryStatementBuilder(
instrumentationtest.New().ToProviderSettings(),
store,
storage,
aggExprRewriter,
logstelemetryschema.DefaultFullTextColumn,
fl,
nil,
statementbuilder.Config{SkipResourceFingerprint: statementbuilder.SkipResourceFingerprint{Enabled: false, Threshold: 100000}},
)
testCases := []struct {
name string
requestType qbtypes.RequestType
query qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]
expected qbtypes.Statement
}{
{
name: "AmbiguousKeyFiltersResourceOnly",
requestType: qbtypes.RequestTypeRaw,
query: qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]{
Signal: telemetrytypes.SignalLogs,
Filter: &qbtypes.Filter{Expression: "service.name = 'webapp'"},
Limit: 10,
},
expected: qbtypes.Statement{
Query: "WITH __resource_filter AS (SELECT fingerprint FROM signoz_logs.distributed_logs_v2_resource WHERE (simpleJSONExtractString(labels, 'service.name') = ? AND labels LIKE ? AND labels LIKE ?) AND seen_at_ts_bucket_start >= ? AND seen_at_ts_bucket_start <= ? GROUP BY fingerprint) SELECT timestamp, id, trace_id, span_id, trace_flags, severity_text, severity_number, scope_name, scope_version, body_v2 as body, attributes_string, attributes_number, attributes_bool, resources_string, scope_string FROM signoz_logs.distributed_logs_v2 WHERE resource_fingerprint GLOBAL IN (SELECT fingerprint FROM __resource_filter) AND timestamp >= ? AND ts_bucket_start >= ? AND timestamp < ? AND ts_bucket_start <= ? LIMIT ?",
Args: []any{"webapp", "%service.name%", "%service.name\":\"webapp%", uint64(1747945619), uint64(1747983448), "1747947419000000000", uint64(1747945619), "1747983448000000000", uint64(1747983448), 10},
Warnings: []string{
"Key `service.name` is ambiguous, found 2 different combinations of field context / data type: [name=service.name,context=resource,datatype=string name=service.name,context=body,datatype=string]. Using `resource` context by default. To query another context explicitly, use the fully qualified name (e.g., 'attribute.service.name' or 'body.service.name')",
},
},
},
}
for _, testCase := range testCases {
t.Run(testCase.name, func(t *testing.T) {
q, err := statementBuilder.Build(context.Background(), valuer.UUID{}, 1747947419000, 1747983448000, testCase.requestType, testCase.query, nil)
require.NoError(t, err)
require.Equal(t, testCase.expected.Query, q.Query)
require.Equal(t, testCase.expected.Args, q.Args)
require.Equal(t, testCase.expected.Warnings, q.Warnings)
})
}
}

View File

@@ -19,6 +19,7 @@ var (
aiobservabilitytypes.GenAIUsageOutputTokens: genAIAttribute(aiobservabilitytypes.GenAIUsageOutputTokens, telemetrytypes.FieldDataTypeFloat64),
aiobservabilitytypes.GenAIUsageCacheReadInputTokens: genAIAttribute(aiobservabilitytypes.GenAIUsageCacheReadInputTokens, telemetrytypes.FieldDataTypeFloat64),
aiobservabilitytypes.GenAIUsageCacheCreationInputTokens: genAIAttribute(aiobservabilitytypes.GenAIUsageCacheCreationInputTokens, telemetrytypes.FieldDataTypeFloat64),
aiobservabilitytypes.GenAIUsageReasoningOutputTokens: genAIAttribute(aiobservabilitytypes.GenAIUsageReasoningOutputTokens, telemetrytypes.FieldDataTypeFloat64),
aiobservabilitytypes.SignozGenAITotalCost: genAIAttribute(aiobservabilitytypes.SignozGenAITotalCost, telemetrytypes.FieldDataTypeFloat64),
aiobservabilitytypes.GenAIInputMessages: genAIAttribute(aiobservabilitytypes.GenAIInputMessages, telemetrytypes.FieldDataTypeString),

View File

@@ -15,6 +15,7 @@ const (
GenAIUsageOutputTokens = "gen_ai.usage.output_tokens"
GenAIUsageCacheReadInputTokens = "gen_ai.usage.cache_read.input_tokens"
GenAIUsageCacheCreationInputTokens = "gen_ai.usage.cache_creation.input_tokens"
GenAIUsageReasoningOutputTokens = "gen_ai.usage.reasoning.output_tokens"
GenAIInputMessages = "gen_ai.input.messages"
GenAIOutputMessages = "gen_ai.output.messages"

View File

@@ -27,6 +27,7 @@ type SpanMapperStore interface {
// TraceStore defines the data access interface for trace detail queries.
type TraceStore interface {
GetTraceSummary(ctx context.Context, traceID string) (*TraceSummary, error)
GetTraceStats(ctx context.Context, orgID valuer.UUID, traceID string, summary *TraceSummary) (*TraceStats, error)
GetTraceSpans(ctx context.Context, traceID string, summary *TraceSummary) ([]StorableSpan, error)
GetMinimalSpans(ctx context.Context, traceID string, start, end time.Time) ([]MinimalSpan, error)
GetTraceSpansByIDs(ctx context.Context, traceID string, start, end time.Time, spanIDs []string) ([]StorableSpan, error)

View File

@@ -0,0 +1,76 @@
package spantypes
import "github.com/SigNoz/signoz/pkg/types/aiobservabilitytypes"
// TraceStatsGenAIColumns pairs each summed TraceStats column with the gen_ai attribute it sums.
var TraceStatsGenAIColumns = []TraceStatsGenAIColumn{
{Column: "input_tokens", Key: aiobservabilitytypes.GenAIUsageInputTokens},
{Column: "output_tokens", Key: aiobservabilitytypes.GenAIUsageOutputTokens},
{Column: "cache_read_tokens", Key: aiobservabilitytypes.GenAIUsageCacheReadInputTokens},
{Column: "cache_write_tokens", Key: aiobservabilitytypes.GenAIUsageCacheCreationInputTokens},
{Column: "reasoning_tokens", Key: aiobservabilitytypes.GenAIUsageReasoningOutputTokens},
{Column: "total_cost", Key: aiobservabilitytypes.SignozGenAITotalCost},
}
type TraceStatsGenAIColumn struct {
Column string
Key string
}
// TraceStats is the single-row result of the trace summary aggregate query.
type TraceStats struct {
StartNs uint64
EndNs uint64
RootServiceName string
RootEntryPoint string
TotalSpans uint64
TotalErrorSpans uint64
HasMissingSpans bool
GenAISpanCount uint64
Tokens TraceAITokens
TotalCost *float64
}
// GettableTraceSummary is the response for the trace summary API; the trace-level
// fields match the waterfall response.
type GettableTraceSummary struct {
StartTimestampMillis uint64 `json:"startTimestampMillis"`
EndTimestampMillis uint64 `json:"endTimestampMillis"`
RootServiceName string `json:"rootServiceName"`
RootServiceEntryPoint string `json:"rootServiceEntryPoint"`
TotalSpansCount uint64 `json:"totalSpansCount"`
TotalErrorSpansCount uint64 `json:"totalErrorSpansCount"`
HasMissingSpans bool `json:"hasMissingSpans"`
AI *TraceAISummary `json:"ai,omitempty"`
}
// TraceAISummary is present when any span carries a gen_ai gate key.
type TraceAISummary struct {
Tokens TraceAITokens `json:"tokens"`
// TotalCost is null when no span carries a cost attribute.
TotalCost *float64 `json:"totalCost" nullable:"true"`
}
type TraceAITokens struct {
Input uint64 `json:"input"`
Output uint64 `json:"output"`
CacheRead uint64 `json:"cacheRead"`
CacheWrite uint64 `json:"cacheWrite"`
Reasoning uint64 `json:"reasoning"`
}
func NewGettableTraceSummary(stats *TraceStats) *GettableTraceSummary {
summary := &GettableTraceSummary{
StartTimestampMillis: stats.StartNs / 1_000_000,
EndTimestampMillis: stats.EndNs / 1_000_000,
RootServiceName: stats.RootServiceName,
RootServiceEntryPoint: stats.RootEntryPoint,
TotalSpansCount: stats.TotalSpans,
TotalErrorSpansCount: stats.TotalErrorSpans,
HasMissingSpans: stats.HasMissingSpans,
}
if stats.GenAISpanCount > 0 {
summary.AI = &TraceAISummary{Tokens: stats.Tokens, TotalCost: stats.TotalCost}
}
return summary
}

View File

@@ -895,6 +895,7 @@ _TRACES_TABLES_TO_TRUNCATE = [
"span_attributes_keys",
"signoz_error_index_v2",
"top_level_operations",
"trace_summary",
]

View File

@@ -1,88 +0,0 @@
import json
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.logs import Logs
from fixtures.querier import (
build_raw_query,
get_rows,
make_query_request,
)
def test_resource_body_conflict(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
export_json_types: Callable[[list[Logs]], None],
) -> None:
now = datetime.now(tz=UTC)
start_ms = int((now - timedelta(seconds=10)).timestamp() * 1000)
end_ms = int(now.timestamp() * 1000)
# python's body carries service.name, making the bare key ambiguous across
# resource and body; java's body omits it, so ANDing body in would drop it.
logs_list = [
Logs(
timestamp=now - timedelta(seconds=2),
resources={"service.name": "java"},
body_v2=json.dumps({"msg": "hello"}),
body_promoted="",
),
Logs(
timestamp=now - timedelta(seconds=1),
resources={"service.name": "python"},
body_v2=json.dumps({"service.name": "python"}),
body_promoted="",
),
]
export_json_types(logs_list)
insert_logs(logs_list)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
cases = [
{
"name": "bare_key_resolves_to_resource",
"filter": "service.name = 'java'",
"expected_service_names": ["java"],
"expect_resource_warning": True,
},
{
"name": "qualified_body_key_targets_body",
"filter": "body.service.name = 'python'",
"expected_service_names": ["python"],
"expect_resource_warning": False,
},
]
for case in cases:
response = make_query_request(
signoz,
token,
start_ms,
end_ms,
request_type="raw",
queries=[
build_raw_query(
name="A",
signal="logs",
filter_expression=case["filter"],
limit=100,
step_interval=60,
)
],
)
assert response.status_code == HTTPStatus.OK, f"{case['name']}: {response.text}"
rows = get_rows(response)
assert [row["data"]["resources_string"].get("service.name") for row in rows] == case["expected_service_names"], f"{case['name']}: {response.json()}"
warning = response.json()["data"].get("warning")
if case["expect_resource_warning"]:
assert warning is not None and "Using `resource` context by default" in warning["warnings"][0]["message"], f"{case['name']}: {warning}"
else:
assert warning is None, f"{case['name']}: {warning}"

View File

@@ -64,8 +64,8 @@ def test_resource_default_warning(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
)
assert warning["warnings"] == [
{"message": expected_service_name_warning},
@@ -237,8 +237,8 @@ def test_deduped_warnings_for_single_query(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
)
expected_status_code_warning = "Key `http.status_code` is ambiguous, found 2 different combinations of field context / data type: [name=http.status_code,context=attribute,datatype=number name=http.status_code,context=attribute,datatype=string]."
assert warning["warnings"] == [
@@ -328,8 +328,8 @@ def test_deduped_warnings_for_multiple_queries(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
)
expected_status_code_warning = "Key `http.status_code` is ambiguous, found 2 different combinations of field context / data type: [name=http.status_code,context=attribute,datatype=number name=http.status_code,context=attribute,datatype=string]."
assert warning["warnings"] == [

View File

@@ -0,0 +1,280 @@
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
import pytest
import requests
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.querierai import root_span
from fixtures.traces import TraceIdGenerator, Traces, TracesKind, TracesStatusCode
WATERFALL_FIELDS = (
"startTimestampMillis",
"endTimestampMillis",
"rootServiceName",
"rootServiceEntryPoint",
"totalSpansCount",
"totalErrorSpansCount",
"hasMissingSpans",
)
@pytest.mark.parametrize("attribute_backend", ["map", "json"])
def test_summary_ai_trace(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_traces: Callable[[list[Traces]], None],
use_attribute_backend: Callable[[str], None],
attribute_backend: str,
) -> None:
"""The summary carries the waterfall's trace-level fields and, for a trace with gen_ai
spans, token totals over every LLM span and the cost summed over the spans that carry it.
Spans are written to one layout only, so a read from the wrong column sums to zero."""
use_attribute_backend(attribute_backend)
write_mode = "json_only" if attribute_backend == "json" else "legacy_only"
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
service = f"td-summary-{attribute_backend}"
resources = {"service.name": service}
trace_id = TraceIdGenerator.trace_id()
root_id = TraceIdGenerator.span_id()
insert_traces(
[
root_span(now=now, trace_id=trace_id, span_id=root_id, resources=resources, duration_s=4),
Traces(
timestamp=now - timedelta(seconds=4),
duration=timedelta(seconds=1),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=root_id,
name="chat gpt-4o-mini",
kind=TracesKind.SPAN_KIND_CLIENT,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
attributes={
"gen_ai.request.model": "gpt-4o-mini",
"gen_ai.usage.input_tokens": 100,
"gen_ai.usage.output_tokens": 20,
"gen_ai.usage.cache_read.input_tokens": 7,
"_signoz.gen_ai.total_cost": 0.01,
},
attribute_write_mode=write_mode,
),
# a failed LLM call: counted in tokens and errors, but priced by nobody
Traces(
timestamp=now - timedelta(seconds=3),
duration=timedelta(seconds=0.5),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=root_id,
name="chat gpt-4o-mini",
kind=TracesKind.SPAN_KIND_CLIENT,
status_code=TracesStatusCode.STATUS_CODE_ERROR,
resources=resources,
attributes={
"gen_ai.request.model": "gpt-4o-mini",
"gen_ai.usage.input_tokens": 50,
"gen_ai.usage.output_tokens": 5,
},
attribute_write_mode=write_mode,
),
Traces(
timestamp=now - timedelta(seconds=2),
duration=timedelta(seconds=0.5),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=root_id,
name="execute_tool",
kind=TracesKind.SPAN_KIND_INTERNAL,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
attributes={"gen_ai.tool.name": "get_weather"},
attribute_write_mode=write_mode,
),
]
)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
headers = {"authorization": f"Bearer {token}", "content-type": "application/json"}
summary = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/traces/{trace_id}/summary"), timeout=10, headers=headers)
assert summary.status_code == HTTPStatus.OK, summary.text
summary = summary.json()["data"]
waterfall = requests.post(
signoz.self.host_configs["8080"].get(f"/api/v4/traces/{trace_id}/waterfall"),
timeout=10,
headers=headers,
json={"selectedSpanId": "", "uncollapsedSpans": []},
)
assert waterfall.status_code == HTTPStatus.OK, waterfall.text
waterfall = waterfall.json()["data"]
assert {k: summary[k] for k in WATERFALL_FIELDS} == {k: waterfall[k] for k in WATERFALL_FIELDS}
assert summary["rootServiceName"] == service
assert summary["rootServiceEntryPoint"] == "POST /api/chat"
assert summary["totalSpansCount"] == 4
assert summary["totalErrorSpansCount"] == 1
assert summary["hasMissingSpans"] is False
assert summary["ai"]["tokens"] == {"input": 150, "output": 25, "cacheRead": 7, "cacheWrite": 0, "reasoning": 0}
assert summary["ai"]["totalCost"] == pytest.approx(0.01)
def test_summary_ai_trace_across_json_rollout(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_traces: Callable[[list[Traces]], None],
seed_attribute_evolution: Callable[[str, datetime], None],
) -> None:
"""A trace that straddles the attribute JSON rollout has LLM spans written only to the legacy
maps before it and to the JSON column after it. The summary window covers both, so the gen_ai
reads must fall back across columns and sum every span."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
rollout = now - timedelta(minutes=30)
seed_attribute_evolution("traces", rollout)
service = "td-summary-rollout"
resources = {"service.name": service}
trace_id = TraceIdGenerator.trace_id()
root_id = TraceIdGenerator.span_id()
insert_traces(
[
Traces(
timestamp=rollout - timedelta(minutes=10),
duration=timedelta(minutes=15),
trace_id=trace_id,
span_id=root_id,
parent_span_id="",
name="long agent run",
kind=TracesKind.SPAN_KIND_SERVER,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
attribute_write_mode="legacy_only",
),
Traces(
timestamp=rollout - timedelta(minutes=5),
duration=timedelta(seconds=1),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=root_id,
name="chat gpt-4o-mini",
kind=TracesKind.SPAN_KIND_CLIENT,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
attributes={"gen_ai.request.model": "gpt-4o-mini", "gen_ai.usage.input_tokens": 100, "gen_ai.usage.output_tokens": 20, "_signoz.gen_ai.total_cost": 0.01},
attribute_write_mode="legacy_only",
),
Traces(
timestamp=rollout + timedelta(minutes=4),
duration=timedelta(seconds=1),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=root_id,
name="chat gpt-4o-mini",
kind=TracesKind.SPAN_KIND_CLIENT,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
attributes={"gen_ai.request.model": "gpt-4o-mini", "gen_ai.usage.input_tokens": 50, "gen_ai.usage.output_tokens": 5, "_signoz.gen_ai.total_cost": 0.02},
attribute_write_mode="json_only",
),
]
)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
summary = requests.get(
signoz.self.host_configs["8080"].get(f"/api/v1/traces/{trace_id}/summary"),
timeout=10,
headers={"authorization": f"Bearer {token}"},
)
assert summary.status_code == HTTPStatus.OK, summary.text
summary = summary.json()["data"]
assert summary["totalSpansCount"] == 3
assert summary["rootServiceEntryPoint"] == "long agent run"
assert summary["ai"]["tokens"] == {"input": 150, "output": 25, "cacheRead": 0, "cacheWrite": 0, "reasoning": 0}
assert summary["ai"]["totalCost"] == pytest.approx(0.03)
def test_summary_non_ai_trace_with_missing_root(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_traces: Callable[[list[Traces]], None],
) -> None:
"""A trace whose recorded spans all hang off an unrecorded parent reports the synthetic
"Missing Span" root exactly as the waterfall does, and a trace without gen_ai spans has
no `ai` block."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
resources = {"service.name": "td-summary-orphan"}
trace_id = TraceIdGenerator.trace_id()
missing_parent_id = TraceIdGenerator.span_id()
insert_traces(
[
Traces(
timestamp=now - timedelta(seconds=5),
duration=timedelta(seconds=2),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=missing_parent_id,
name="SELECT users",
kind=TracesKind.SPAN_KIND_CLIENT,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
),
Traces(
timestamp=now - timedelta(seconds=4),
duration=timedelta(seconds=1),
trace_id=trace_id,
span_id=TraceIdGenerator.span_id(),
parent_span_id=missing_parent_id,
name="publish event",
kind=TracesKind.SPAN_KIND_PRODUCER,
status_code=TracesStatusCode.STATUS_CODE_OK,
resources=resources,
),
]
)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
headers = {"authorization": f"Bearer {token}", "content-type": "application/json"}
summary = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/traces/{trace_id}/summary"), timeout=10, headers=headers)
assert summary.status_code == HTTPStatus.OK, summary.text
summary = summary.json()["data"]
waterfall = requests.post(
signoz.self.host_configs["8080"].get(f"/api/v4/traces/{trace_id}/waterfall"),
timeout=10,
headers=headers,
json={"selectedSpanId": "", "uncollapsedSpans": []},
)
assert waterfall.status_code == HTTPStatus.OK, waterfall.text
waterfall = waterfall.json()["data"]
assert {k: summary[k] for k in WATERFALL_FIELDS} == {k: waterfall[k] for k in WATERFALL_FIELDS}
assert summary["hasMissingSpans"] is True
assert summary["rootServiceName"] == ""
assert summary["rootServiceEntryPoint"] == "Missing Span"
assert summary["totalSpansCount"] == 2
assert "ai" not in summary
def test_summary_unknown_trace(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = requests.get(
signoz.self.host_configs["8080"].get(f"/api/v1/traces/{TraceIdGenerator.trace_id()}/summary"),
timeout=10,
headers={"authorization": f"Bearer {token}"},
)
assert response.status_code == HTTPStatus.NOT_FOUND, response.text