refactor: simplify BFS in getAllLevels method

This reverts commit f9222c9930.

conf/example.yaml

@@ -440,6 +440,17 @@ tracedetail:
     max_depth_to_auto_expand: 5
     # Threshold below which all spans are returned without windowing.
     max_limit_to_select_all_spans: 10000
+  flamegraph:
+    # Maximum number of BFS depth levels included in a windowed response.
+    max_selected_levels: 50
+    # Maximum spans per level before sampling is applied.
+    max_spans_per_level: 100
+    # Number of highest-latency spans always included when sampling a level.
+    sampling_top_latency_count: 5
+    # Number of timestamp buckets used for uniform sampling within a level.
+    sampling_bucket_count: 50
+    # Threshold below which all spans are returned without windowing or sampling.
+    select_all_spans_limit: 100000
 
 ##################### Authz #################################
 authz:

docs/api/openapi.yml

@@ -6516,6 +6516,61 @@ components:
       - attribute
       - resource
       type: string
+    SpantypesFlamegraphSpan:
+      properties:
+        attributes:
+          additionalProperties: {}
+          type: object
+        durationNano:
+          minimum: 0
+          type: integer
+        event:
+          items:
+            $ref: '#/components/schemas/SpantypesEvent'
+          type: array
+        hasError:
+          type: boolean
+        level:
+          format: int64
+          type: integer
+        name:
+          type: string
+        parentSpanId:
+          type: string
+        resource:
+          additionalProperties:
+            type: string
+          type: object
+        spanId:
+          type: string
+        timestamp:
+          minimum: 0
+          type: integer
+      required:
+      - event
+      type: object
+    SpantypesGettableFlamegraphTrace:
+      properties:
+        endTimestampMillis:
+          format: int64
+          type: integer
+        hasMore:
+          type: boolean
+        spans:
+          items:
+            items:
+              $ref: '#/components/schemas/SpantypesFlamegraphSpan'
+            type: array
+          type: array
+        startTimestampMillis:
+          format: int64
+          type: integer
+      required:
+      - spans
+      - startTimestampMillis
+      - endTimestampMillis
+      - hasMore
+      type: object
     SpantypesGettableSpanMapperGroups:
       properties:
         items:
@@ -6581,6 +6636,15 @@ components:
         traceId:
           type: string
       type: object
+    SpantypesPostableFlamegraph:
+      properties:
+        selectFields:
+          items:
+            $ref: '#/components/schemas/TelemetrytypesTelemetryFieldKey'
+          type: array
+        selectedSpanId:
+          type: string
+      type: object
     SpantypesPostableSpanMapper:
       properties:
         config:
@@ -20027,6 +20091,75 @@ paths:
       summary: Put profile in Zeus for a deployment.
       tags:
       - zeus
+  /api/v3/traces/{traceID}/flamegraph:
+    post:
+      deprecated: false
+      description: Returns the flamegraph view of spans for a given trace ID.
+      operationId: GetFlamegraph
+      parameters:
+      - in: path
+        name: traceID
+        required: true
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/SpantypesPostableFlamegraph'
+      responses:
+        "200":
+          content:
+            application/json:
+              schema:
+                properties:
+                  data:
+                    $ref: '#/components/schemas/SpantypesGettableFlamegraphTrace'
+                  status:
+                    type: string
+                required:
+                - status
+                - data
+                type: object
+          description: OK
+        "400":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Bad Request
+        "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 flamegraph view for a trace
+      tags:
+      - tracedetail
   /api/v3/traces/{traceID}/waterfall:
     post:
       deprecated: false

frontend/src/api/generated/services/sigNoz.schemas.ts

@@ -7675,6 +7675,77 @@ export enum SpantypesFieldContextDTO {
 	attribute = 'attribute',
 	resource = 'resource',
 }
+export type SpantypesFlamegraphSpanDTOAttributes = { [key: string]: unknown };
+
+export type SpantypesFlamegraphSpanDTOResource = { [key: string]: string };
+
+export interface SpantypesFlamegraphSpanDTO {
+	/**
+	 * @type object
+	 */
+	attributes?: SpantypesFlamegraphSpanDTOAttributes;
+	/**
+	 * @type integer
+	 * @minimum 0
+	 */
+	durationNano?: number;
+	/**
+	 * @type array
+	 */
+	event: SpantypesEventDTO[];
+	/**
+	 * @type boolean
+	 */
+	hasError?: boolean;
+	/**
+	 * @type integer
+	 * @format int64
+	 */
+	level?: number;
+	/**
+	 * @type string
+	 */
+	name?: string;
+	/**
+	 * @type string
+	 */
+	parentSpanId?: string;
+	/**
+	 * @type object
+	 */
+	resource?: SpantypesFlamegraphSpanDTOResource;
+	/**
+	 * @type string
+	 */
+	spanId?: string;
+	/**
+	 * @type integer
+	 * @minimum 0
+	 */
+	timestamp?: number;
+}
+
+export interface SpantypesGettableFlamegraphTraceDTO {
+	/**
+	 * @type integer
+	 * @format int64
+	 */
+	endTimestampMillis: number;
+	/**
+	 * @type boolean
+	 */
+	hasMore: boolean;
+	/**
+	 * @type array
+	 */
+	spans: SpantypesFlamegraphSpanDTO[][];
+	/**
+	 * @type integer
+	 * @format int64
+	 */
+	startTimestampMillis: number;
+}
+
 export type SpantypesSpanMapperGroupConditionDTOAnyOf = {
 	/**
 	 * @type array,null
@@ -7976,6 +8047,17 @@ export interface SpantypesGettableWaterfallTraceDTO {
 	uncollapsedSpans?: string[] | null;
 }
 
+export interface SpantypesPostableFlamegraphDTO {
+	/**
+	 * @type array
+	 */
+	selectFields?: TelemetrytypesTelemetryFieldKeyDTO[];
+	/**
+	 * @type string
+	 */
+	selectedSpanId?: string;
+}
+
 export enum SpantypesSpanMapperOperationDTO {
 	move = 'move',
 	copy = 'copy',
@@ -10277,6 +10359,17 @@ export type GetHosts200 = {
 	status: string;
 };
 
+export type GetFlamegraphPathParameters = {
+	traceID: string;
+};
+export type GetFlamegraph200 = {
+	data: SpantypesGettableFlamegraphTraceDTO;
+	/**
+	 * @type string
+	 */
+	status: string;
+};
+
 export type GetWaterfallPathParameters = {
 	traceID: string;
 };

frontend/src/api/generated/services/tracedetail/index.ts

@@ -12,6 +12,8 @@ import type {
 } from 'react-query';
 
 import type {
+	GetFlamegraph200,
+	GetFlamegraphPathParameters,
 	GetTraceAggregations200,
 	GetTraceAggregationsPathParameters,
 	GetWaterfall200,
@@ -19,6 +21,7 @@ import type {
 	GetWaterfallV4200,
 	GetWaterfallV4PathParameters,
 	RenderErrorResponseDTO,
+	SpantypesPostableFlamegraphDTO,
 	SpantypesPostableTraceAggregationsDTO,
 	SpantypesPostableWaterfallDTO,
 } from '../sigNoz.schemas';
@@ -126,6 +129,105 @@ export const useGetTraceAggregations = <
 > => {
 	return useMutation(getGetTraceAggregationsMutationOptions(options));
 };
+/**
+ * Returns the flamegraph view of spans for a given trace ID.
+ * @summary Get flamegraph view for a trace
+ */
+export const getFlamegraph = (
+	{ traceID }: GetFlamegraphPathParameters,
+	spantypesPostableFlamegraphDTO?: BodyType<SpantypesPostableFlamegraphDTO>,
+	signal?: AbortSignal,
+) => {
+	return GeneratedAPIInstance<GetFlamegraph200>({
+		url: `/api/v3/traces/${traceID}/flamegraph`,
+		method: 'POST',
+		headers: { 'Content-Type': 'application/json' },
+		data: spantypesPostableFlamegraphDTO,
+		signal,
+	});
+};
+
+export const getGetFlamegraphMutationOptions = <
+	TError = ErrorType<RenderErrorResponseDTO>,
+	TContext = unknown,
+>(options?: {
+	mutation?: UseMutationOptions<
+		Awaited<ReturnType<typeof getFlamegraph>>,
+		TError,
+		{
+			pathParams: GetFlamegraphPathParameters;
+			data?: BodyType<SpantypesPostableFlamegraphDTO>;
+		},
+		TContext
+	>;
+}): UseMutationOptions<
+	Awaited<ReturnType<typeof getFlamegraph>>,
+	TError,
+	{
+		pathParams: GetFlamegraphPathParameters;
+		data?: BodyType<SpantypesPostableFlamegraphDTO>;
+	},
+	TContext
+> => {
+	const mutationKey = ['getFlamegraph'];
+	const { mutation: mutationOptions } = options
+		? options.mutation &&
+			'mutationKey' in options.mutation &&
+			options.mutation.mutationKey
+			? options
+			: { ...options, mutation: { ...options.mutation, mutationKey } }
+		: { mutation: { mutationKey } };
+
+	const mutationFn: MutationFunction<
+		Awaited<ReturnType<typeof getFlamegraph>>,
+		{
+			pathParams: GetFlamegraphPathParameters;
+			data?: BodyType<SpantypesPostableFlamegraphDTO>;
+		}
+	> = (props) => {
+		const { pathParams, data } = props ?? {};
+
+		return getFlamegraph(pathParams, data);
+	};
+
+	return { mutationFn, ...mutationOptions };
+};
+
+export type GetFlamegraphMutationResult = NonNullable<
+	Awaited<ReturnType<typeof getFlamegraph>>
+>;
+export type GetFlamegraphMutationBody =
+	| BodyType<SpantypesPostableFlamegraphDTO>
+	| undefined;
+export type GetFlamegraphMutationError = ErrorType<RenderErrorResponseDTO>;
+
+/**
+ * @summary Get flamegraph view for a trace
+ */
+export const useGetFlamegraph = <
+	TError = ErrorType<RenderErrorResponseDTO>,
+	TContext = unknown,
+>(options?: {
+	mutation?: UseMutationOptions<
+		Awaited<ReturnType<typeof getFlamegraph>>,
+		TError,
+		{
+			pathParams: GetFlamegraphPathParameters;
+			data?: BodyType<SpantypesPostableFlamegraphDTO>;
+		},
+		TContext
+	>;
+}): UseMutationResult<
+	Awaited<ReturnType<typeof getFlamegraph>>,
+	TError,
+	{
+		pathParams: GetFlamegraphPathParameters;
+		data?: BodyType<SpantypesPostableFlamegraphDTO>;
+	},
+	TContext
+> => {
+	return useMutation(getGetFlamegraphMutationOptions(options));
+};
 /**
  * Returns the waterfall view of spans for a given trace ID with tree structure, metadata, and windowed pagination
  * @summary Get waterfall view for a trace

frontend/src/pages/TraceDetailsV3/TraceFlamegraph/__tests__/computeVisualLayout.test.ts

@@ -472,98 +472,4 @@ describe('computeVisualLayout', () => {
 		expect(aRow).toBeGreaterThan(1); // must NOT be at row 1
 		expect(aRow).toBe(3); // next free row after B at row 2 (A overlaps B)
 	});
-
-	// --- Wide-group fast path (> WIDE_GROUP_THRESHOLD siblings) ---
-	// Past the threshold the layout switches to exact overlap-only packing to
-	// avoid the O(N^2) connector-avoidance spiral. These lock in correctness and
-	// the no-overlap invariant at scale.
-
-	function noRowHasOverlap(
-		layout: ReturnType<typeof computeVisualLayout>,
-	): void {
-		for (const row of layout.visualRows) {
-			const sorted = [...row].sort((a, b) => a.timestamp - b.timestamp);
-			for (let i = 1; i < sorted.length; i++) {
-				const prevEnd = sorted[i - 1].timestamp + sorted[i - 1].durationNano / 1e6;
-				expect(sorted[i].timestamp).toBeGreaterThanOrEqual(prevEnd);
-			}
-		}
-	}
-
-	it('should pack thousands of sequential leaf siblings into 1 row (wide path)', () => {
-		const root = makeSpan({ spanId: 'root', timestamp: 0, durationNano: 1e12 });
-		const kids: FlamegraphSpan[] = [];
-		// 2000 strictly sequential (non-overlapping) children
-		for (let i = 0; i < 2000; i++) {
-			kids.push(
-				makeSpan({
-					spanId: `k${i}`,
-					parentSpanId: 'root',
-					timestamp: i * 10,
-					durationNano: 5e6, // 5ms, ends before next starts
-				}),
-			);
-		}
-
-		const layout = computeVisualLayout([[root], kids]);
-
-		expect(layout.spanToVisualRow['root']).toBe(0);
-		expect(layout.totalVisualRows).toBe(2); // all siblings share row 1
-		for (const k of kids) {
-			expect(layout.spanToVisualRow[k.spanId]).toBe(1);
-		}
-		noRowHasOverlap(layout);
-	});
-
-	it('should pack thousands of fully-overlapping leaf siblings without violations (wide path)', () => {
-		const root = makeSpan({ spanId: 'root', timestamp: 0, durationNano: 1e12 });
-		const kids: FlamegraphSpan[] = [];
-		// 1000 children all spanning the same window → each needs its own row
-		for (let i = 0; i < 1000; i++) {
-			kids.push(
-				makeSpan({
-					spanId: `k${i}`,
-					parentSpanId: 'root',
-					timestamp: 0,
-					durationNano: 100e6,
-				}),
-			);
-		}
-
-		const layout = computeVisualLayout([[root], kids]);
-
-		expect(layout.totalVisualRows).toBe(1001); // root + 1000 stacked rows
-		expect(Object.keys(layout.spanToVisualRow)).toHaveLength(1001);
-		noRowHasOverlap(layout);
-	});
-
-	it('should keep non-leaf subtrees adjacent within a wide mixed group (wide path)', () => {
-		const root = makeSpan({ spanId: 'root', timestamp: 0, durationNano: 1e12 });
-		const kids: FlamegraphSpan[] = [];
-		for (let i = 0; i < 1000; i++) {
-			kids.push(
-				makeSpan({
-					spanId: `k${i}`,
-					parentSpanId: 'root',
-					timestamp: i * 10,
-					durationNano: 5e6,
-				}),
-			);
-		}
-		// One of the wide siblings has a child of its own
-		const grandchild = makeSpan({
-			spanId: 'gc',
-			parentSpanId: 'k500',
-			timestamp: 5000,
-			durationNano: 2e6,
-		});
-
-		const layout = computeVisualLayout([[root], kids, [grandchild]]);
-
-		const parentRow = layout.spanToVisualRow['k500'];
-		const gcRow = layout.spanToVisualRow['gc'];
-		expect(gcRow - parentRow).toBe(1); // subtree adjacency preserved
-		expect(Object.keys(layout.spanToVisualRow)).toHaveLength(1002);
-		noRowHasOverlap(layout);
-	});
 });

frontend/src/pages/TraceDetailsV3/TraceFlamegraph/computeVisualLayout.ts

@@ -18,81 +18,6 @@ export interface VisualLayout {
 	totalVisualRows: number;
 }
 
-// Above this many siblings under one parent, the connector-avoidance refinement
-// (Checks 2 & 3) is both visually meaningless — the row is already a dense wall —
-// and quadratic: every child deposits a connector point on each intermediate row,
-// which pushes later children even higher, which deposits more points. That
-// feedback loop inflates a layout needing ~50 rows to thousands and never
-// finishes on wide traces. Past the threshold we pack by overlap only.
-const WIDE_GROUP_THRESHOLD = 512;
-
-/**
- * Segment tree over rows that answers "lowest row index >= `from` whose smallest
- * span start-time is >= `end`" in O(log rows). Used to place a large group of
- * leaf siblings by overlap only: because siblings are processed in descending
- * start order, every already-placed span on a row starts at or after the current
- * one, so [start, end] overlaps a row iff some span there starts before `end` —
- * i.e. the row is free iff its minimum start >= end. Each node stores the max of
- * its subtree's per-row minimum starts so a free row can be found by descent.
- */
-class LowestFreeRow {
-	private readonly size: number;
-
-	private readonly tree: Float64Array;
-
-	constructor(rows: number) {
-		let size = 1;
-		while (size < rows) {
-			size *= 2;
-		}
-		this.size = size;
-		this.tree = new Float64Array(size * 2).fill(Infinity);
-	}
-
-	place(row: number, start: number): void {
-		let i = row + this.size;
-		// A row's key is the minimum start among its spans. Children are processed
-		// in descending start order so a leaf's start is the new minimum, but a
-		// non-leaf subtree's descendant can land on a row out of order — take min.
-		if (start >= this.tree[i]) {
-			return;
-		}
-		this.tree[i] = start;
-		for (i >>= 1; i >= 1; i >>= 1) {
-			const next = Math.max(this.tree[2 * i], this.tree[2 * i + 1]);
-			if (this.tree[i] === next) {
-				break;
-			}
-			this.tree[i] = next;
-		}
-	}
-
-	lowestFrom(from: number, end: number): number {
-		return this.descend(1, 0, this.size - 1, from, end);
-	}
-
-	private descend(
-		node: number,
-		lo: number,
-		hi: number,
-		from: number,
-		end: number,
-	): number {
-		if (hi < from || this.tree[node] < end) {
-			return -1;
-		}
-		if (lo === hi) {
-			return lo;
-		}
-		const mid = (lo + hi) >> 1;
-		const left = this.descend(2 * node, lo, mid, from, end);
-		if (left !== -1) {
-			return left;
-		}
-		return this.descend(2 * node + 1, mid + 1, hi, from, end);
-	}
-}
-
 /**
  * Computes an overlap-safe visual layout for flamegraph spans using DFS ordering.
  *
@@ -289,53 +214,7 @@ export function computeVisualLayout(spans: FlamegraphSpan[][]): VisualLayout {
 		arr.push(point);
 	}
 
-	// Fast path for a parent with a very large group of children: pack by overlap
-	// only (descending greedy), skipping the quadratic connector-avoidance that
-	// spirals at this scale. Leaf children — the bulk of a wide trace — are placed
-	// in O(log rows) via the segment tree; the rare non-leaf subtree falls back to
-	// findPlacement against the shared interval map. Both structures are kept in
-	// sync so each placement sees all prior occupancy. Same ShapeEntry[] contract.
-	function computeWideShape(
-		rootSpan: FlamegraphSpan,
-		children: FlamegraphSpan[],
-	): ShapeEntry[] {
-		const shape: ShapeEntry[] = [{ span: rootSpan, relativeRow: 0 }];
-		const localIntervals = new Map<number, Array<[number, number]>>();
-		// Children occupy relative rows 1..children.length in the worst case.
-		const finder = new LowestFreeRow(children.length + 2);
-
-		const occupy = (row: number, span: FlamegraphSpan): void => {
-			const s = span.timestamp;
-			const e = span.timestamp + span.durationNano / 1e6;
-			shape.push({ span, relativeRow: row });
-			addIntervalTo(localIntervals, row, s, e);
-			finder.place(row, s);
-		};
-
-		for (const child of children) {
-			if (childrenMap.has(child.spanId)) {
-				// Non-leaf: place its whole subtree shape as a unit via findPlacement.
-				const childShape = computeSubtreeShape(child);
-				const offset = findPlacement(childShape, 1, localIntervals);
-				for (const entry of childShape) {
-					occupy(entry.relativeRow + offset, entry.span);
-				}
-			} else {
-				const end = child.timestamp + child.durationNano / 1e6;
-				occupy(finder.lowestFrom(1, end), child);
-			}
-		}
-
-		return shape;
-	}
-
 	function computeSubtreeShape(rootSpan: FlamegraphSpan): ShapeEntry[] {
-		const children = childrenMap.get(rootSpan.spanId);
-
-		if (children && children.length > WIDE_GROUP_THRESHOLD) {
-			return computeWideShape(rootSpan, children);
-		}
-
 		const localIntervals = new Map<number, Array<[number, number]>>();
 		const localConnectorPoints = new Map<number, number[]>();
 		const shape: ShapeEntry[] = [];
@@ -346,6 +225,7 @@ export function computeVisualLayout(spans: FlamegraphSpan[][]): VisualLayout {
 		shape.push({ span: rootSpan, relativeRow: 0 });
 		addIntervalTo(localIntervals, 0, rootStart, rootEnd);
 
+		const children = childrenMap.get(rootSpan.spanId);
 		if (children) {
 			for (const child of children) {
 				const childShape = computeSubtreeShape(child);

frontend/src/pages/TraceDetailsV3/TraceFlamegraph/hooks/useVisualLayoutWorker.ts

@@ -94,7 +94,7 @@ export function useVisualLayoutWorker(spans: FlamegraphSpan[][]): {
 			cleanup();
 		};
 
-		// Timeout: if worker doesn't respond in 15s, terminate and error
+		// Timeout: if worker doesn't respond in 30s, terminate and error
 		const WORKER_TIMEOUT_MS = 15000;
 		const timeoutId = setTimeout(() => {
 			if (requestIdRef.current === currentId && isComputingRef.current) {

pkg/apiserver/signozapiserver/tracedetail.go

@@ -67,5 +67,24 @@ func (provider *provider) addTraceDetailRoutes(router *mux.Router) error {
 		return err
 	}
 
+	if err := router.Handle("/api/v3/traces/{traceID}/flamegraph", handler.New(
+		provider.authzMiddleware.ViewAccess(provider.traceDetailHandler.GetFlamegraph),
+		handler.OpenAPIDef{
+			ID:                  "GetFlamegraph",
+			Tags:                []string{"tracedetail"},
+			Summary:             "Get flamegraph view for a trace",
+			Description:         "Returns the flamegraph view of spans for a given trace ID.",
+			Request:             new(spantypes.PostableFlamegraph),
+			RequestContentType:  "application/json",
+			Response:            new(spantypes.GettableFlamegraphTrace),
+			ResponseContentType: "application/json",
+			SuccessStatusCode:   http.StatusOK,
+			ErrorStatusCodes:    []int{http.StatusBadRequest, http.StatusNotFound},
+			SecuritySchemes:     newSecuritySchemes(types.RoleViewer),
+		},
+	)).Methods(http.MethodPost).GetError(); err != nil {
+		return err
+	}
+
 	return nil
 }

pkg/modules/tracedetail/config.go

@@ -6,7 +6,16 @@ import (
 )
 
 type Config struct {
-	Waterfall WaterfallConfig `mapstructure:"waterfall"`
+	Waterfall  WaterfallConfig  `mapstructure:"waterfall"`
+	Flamegraph FlamegraphConfig `mapstructure:"flamegraph"`
+}
+
+type FlamegraphConfig struct {
+	MaxSelectedLevels            int  `mapstructure:"max_selected_levels"`
+	MaxSpansPerLevel             int  `mapstructure:"max_spans_per_level"`
+	SamplingTopLatencySpansCount int  `mapstructure:"sampling_top_latency_count"`
+	SamplingBucketCount          int  `mapstructure:"sampling_bucket_count"`
+	SelectAllSpansLimit          uint `mapstructure:"select_all_spans_limit"`
 }
 
 type WaterfallConfig struct {
@@ -29,6 +38,13 @@ func newConfig() factory.Config {
 			MaxDepthToAutoExpand:     5,
 			MaxLimitToSelectAllSpans: 10_000,
 		},
+		Flamegraph: FlamegraphConfig{
+			MaxSelectedLevels:            50,
+			MaxSpansPerLevel:             100,
+			SamplingTopLatencySpansCount: 5,
+			SamplingBucketCount:          50,
+			SelectAllSpansLimit:          100_000,
+		},
 	}
 }
 
@@ -45,5 +61,25 @@ func (c Config) Validate() error {
 		return errors.NewInvalidInputf(errors.CodeInvalidInput,
 			"tracedetail.waterfall.max_limit_to_select_all_spans must be positive")
 	}
+	if c.Flamegraph.MaxSelectedLevels <= 0 {
+		return errors.NewInvalidInputf(errors.CodeInvalidInput,
+			"tracedetail.flamegraph.level_limit must be positive, got %d", c.Flamegraph.MaxSelectedLevels)
+	}
+	if c.Flamegraph.MaxSpansPerLevel <= 0 {
+		return errors.NewInvalidInputf(errors.CodeInvalidInput,
+			"tracedetail.flamegraph.spans_per_level must be positive, got %d", c.Flamegraph.MaxSpansPerLevel)
+	}
+	if c.Flamegraph.SamplingTopLatencySpansCount < 0 {
+		return errors.NewInvalidInputf(errors.CodeInvalidInput,
+			"tracedetail.flamegraph.top_latency_count cannot be negative, got %d", c.Flamegraph.SamplingTopLatencySpansCount)
+	}
+	if c.Flamegraph.SamplingBucketCount <= 0 {
+		return errors.NewInvalidInputf(errors.CodeInvalidInput,
+			"tracedetail.flamegraph.bucket_count must be positive, got %d", c.Flamegraph.SamplingBucketCount)
+	}
+	if c.Flamegraph.SelectAllSpansLimit == 0 {
+		return errors.NewInvalidInputf(errors.CodeInvalidInput,
+			"tracedetail.flamegraph.max_limit_to_select_all_spans must be positive")
+	}
 	return nil
 }

pkg/modules/tracedetail/impltracedetail/handler.go

@@ -80,3 +80,19 @@ func (h *handler) GetTraceAggregations(rw http.ResponseWriter, r *http.Request)
 
 	render.Success(rw, http.StatusOK, result)
 }
+
+func (h *handler) GetFlamegraph(rw http.ResponseWriter, r *http.Request) {
+	req := new(spantypes.PostableFlamegraph)
+	if err := binding.JSON.BindBody(r.Body, req); err != nil {
+		render.Error(rw, err)
+		return
+	}
+
+	result, err := h.module.GetFlamegraph(r.Context(), mux.Vars(r)["traceID"], req.SelectedSpanID, req.SelectFields)
+	if err != nil {
+		render.Error(rw, err)
+		return
+	}
+
+	render.Success(rw, http.StatusOK, result)
+}

pkg/modules/tracedetail/impltracedetail/module.go

@@ -7,6 +7,7 @@ import (
 	"github.com/SigNoz/signoz/pkg/factory"
 	"github.com/SigNoz/signoz/pkg/modules/tracedetail"
 	"github.com/SigNoz/signoz/pkg/types/spantypes"
+	"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
 )
 
 type module struct {
@@ -148,6 +149,17 @@ func (m *module) GetTraceAggregations(ctx context.Context, traceID string, req *
 	return &spantypes.GettableTraceAggregations{Aggregations: results}, nil
 }
 
+func (m *module) GetFlamegraph(ctx context.Context, traceID string, selectedSpanID string, selectFields []telemetrytypes.TelemetryFieldKey) (*spantypes.GettableFlamegraphTrace, error) {
+	summary, err := m.store.GetTraceSummary(ctx, traceID)
+	if err != nil {
+		return nil, err
+	}
+	if summary.NumSpans <= uint64(m.config.Flamegraph.SelectAllSpansLimit) {
+		return m.getFullFlamegraph(ctx, traceID, summary, selectFields)
+	}
+	return m.getWindowedFlamegraph(ctx, traceID, selectedSpanID, summary, selectFields)
+}
+
 // getWindowedWaterfall builds the waterfall tree with minimal data and then returns only a window of full spans.
 func (m *module) getWindowedWaterfall(ctx context.Context, traceID, selectedSpanID string, uncollapsedSpans []string, start, end time.Time) (*spantypes.GettableWaterfallTrace, error) {
 	// Step 1: minimal fetch → build full tree → select visible window
@@ -188,3 +200,47 @@ func (m *module) getWindowedWaterfall(ctx context.Context, traceID, selectedSpan
 		waterfallTrace, selectedSpans, uncollapsedSpans, false, nil,
 	), nil
 }
+
+func (m *module) getFullFlamegraph(ctx context.Context, traceID string, summary *spantypes.TraceSummary, selectFields []telemetrytypes.TelemetryFieldKey) (*spantypes.GettableFlamegraphTrace, error) {
+	fullSpans, err := m.store.GetFlamegraphSpans(ctx, traceID, summary.Start, summary.End, nil)
+	if err != nil {
+		return nil, err
+	}
+	if len(fullSpans) == 0 {
+		return nil, spantypes.ErrTraceNotFound
+	}
+	flamegraphTrace := spantypes.NewFlamegraphTraceFromStorable(fullSpans, selectFields)
+	return spantypes.NewGettableFlamegraphTrace(flamegraphTrace.GetAllLevels(), summary.Start.UnixMilli(), summary.End.UnixMilli(), false), nil
+}
+
+// getWindowedFlamegraph returns a window of a max levels and max sampled spans per level around the selected span.
+func (m *module) getWindowedFlamegraph(ctx context.Context, traceID, selectedSpanID string, summary *spantypes.TraceSummary, selectFields []telemetrytypes.TelemetryFieldKey) (*spantypes.GettableFlamegraphTrace, error) {
+	minimalSpans, err := m.store.GetMinimalSpans(ctx, traceID, summary.Start, summary.End)
+	if err != nil {
+		return nil, err
+	}
+	if len(minimalSpans) == 0 {
+		return nil, spantypes.ErrTraceNotFound
+	}
+
+	flamegraphTrace := spantypes.NewFlamegraphTraceFromMinimal(minimalSpans)
+	minimalSpans = nil //nolint:ineffassign,wastedassign // release backing array before further db calls
+
+	cfg := m.config.Flamegraph
+	selectedSpans := flamegraphTrace.GetSelectedLevels(selectedSpanID, cfg.MaxSelectedLevels, cfg.MaxSpansPerLevel, cfg.SamplingTopLatencySpansCount, cfg.SamplingBucketCount)
+	if len(selectedSpans) == 0 {
+		return nil, spantypes.ErrTraceNotFound
+	}
+
+	fullSpans, err := m.store.GetFlamegraphSpans(ctx, traceID, summary.Start, summary.End, spantypes.FlamegraphWindowSpanIDs(selectedSpans))
+	if err != nil {
+		return nil, err
+	}
+
+	return spantypes.NewGettableFlamegraphTrace(
+		flamegraphTrace.EnrichSelectedSpans(selectedSpans, fullSpans, selectFields),
+		summary.Start.UnixMilli(),
+		summary.End.UnixMilli(),
+		true,
+	), nil
+}

pkg/modules/tracedetail/impltracedetail/store.go

@@ -154,6 +154,46 @@ func (s *traceStore) GetTraceSpansByIDs(ctx context.Context, traceID string, sta
 	return spans, nil
 }
 
+func (s *traceStore) GetFlamegraphSpans(ctx context.Context, traceID string, start, end time.Time, spanIDs []string) ([]spantypes.StorableSpan, error) {
+	sb := sqlbuilder.NewSelectBuilder()
+	sb.Select(
+		"span_id",
+		"any(parent_span_id) AS parent_span_id",
+		"any(timestamp) AS timestamp",
+		"any(duration_nano) AS duration_nano",
+		"any(has_error) AS has_error",
+		"any(name) AS name",
+		"any(events) AS events",
+		"any(attributes_string) AS attributes_string",
+		"any(attributes_number) AS attributes_number",
+		"any(attributes_bool) AS attributes_bool",
+		"any(resources_string) AS resources_string",
+	)
+	sb.From(fmt.Sprintf("%s.%s", spantypes.TraceDB, spantypes.TraceTable))
+	conditions := []string{
+		sb.E("trace_id", traceID),
+		sb.GE("ts_bucket_start", start.Unix()-1800),
+		sb.LE("ts_bucket_start", end.Unix()),
+	}
+	if len(spanIDs) > 0 {
+		ids := make([]any, len(spanIDs))
+		for i, id := range spanIDs {
+			ids[i] = id
+		}
+		conditions = append(conditions, sb.In("span_id", ids...))
+	}
+	sb.Where(conditions...)
+	sb.GroupBy("span_id")
+	sb.OrderBy("timestamp ASC", "name ASC")
+	query, args := sb.BuildWithFlavor(sqlbuilder.ClickHouse)
+
+	var spans []spantypes.StorableSpan
+	if err := s.telemetryStore.ClickhouseDB().Select(ctx, &spans, query, args...); err != nil {
+		return nil, errors.WrapInternalf(err, errors.CodeInternal, "error querying flamegraph spans")
+	}
+	return spans, nil
+}
+
 func (s *traceStore) GetSpanCountByField(ctx context.Context, traceID string, summary *spantypes.TraceSummary, fieldKey telemetrytypes.TelemetryFieldKey) (map[string]uint64, error) {
 	fieldExpr, err := buildFieldExpr(fieldKey)
 	if err != nil {

pkg/modules/tracedetail/tracedetail.go

@@ -5,6 +5,7 @@ import (
 	"net/http"
 
 	"github.com/SigNoz/signoz/pkg/types/spantypes"
+	"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
 )
 
 // Handler exposes HTTP handlers for trace detail APIs.
@@ -12,6 +13,7 @@ type Handler interface {
 	GetWaterfall(http.ResponseWriter, *http.Request)
 	GetWaterfallV4(http.ResponseWriter, *http.Request)
 	GetTraceAggregations(http.ResponseWriter, *http.Request)
+	GetFlamegraph(http.ResponseWriter, *http.Request)
 }
 
 // Module defines the business logic for trace detail operations.
@@ -19,4 +21,5 @@ type Module interface {
 	GetWaterfall(ctx context.Context, traceID string, req *spantypes.PostableWaterfall) (*spantypes.GettableWaterfallTrace, error)
 	GetWaterfallV4(ctx context.Context, traceID string, selectedSpanID string, uncollapsedSpans []string, selectAllLimit uint) (*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)
 }

pkg/telemetrytraces/statement_builder.go

@@ -124,10 +124,8 @@ func (b *traceQueryStatementBuilder) Build(
 		-------------------------------- End of tech debt ----------------------------
 	*/
 
-	for _, action := range adjustTraceKeys(keys, &query, requestType) {
-		// TODO: change to debug level once we are confident about the behavior
-		b.logger.InfoContext(ctx, "key adjustment action", slog.String("action", action))
-	}
+	query = b.adjustKeys(ctx, keys, query, requestType)
+
 	// Create SQL builder
 	q := sqlbuilder.NewSelectBuilder()
 
@@ -195,30 +193,24 @@ func getKeySelectors(query qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation])
 	return keySelectors
 }
 
-// mergeDeprecatedTraceKeys prepends deprecated intrinsic/calculated trace field
-// definitions to the keys map. We do this during statement building, not at
-// metadata fetch time, because:
-//  1. Filter expressions that reference deprecated columns must continue to
-//     resolve — otherwise they fail with "key not found".
-//  2. Doing it at metadata fetch time would also surface deprecated keys in
-//     autocomplete suggestions, which we don't want.
-//  3. We prepend (not append) so the intrinsic/calculated entry wins ordering
-//     in the multi_if SQL expression.
-func mergeDeprecatedTraceKeys(keys map[string][]*telemetrytypes.TelemetryFieldKey) {
+func (b *traceQueryStatementBuilder) adjustKeys(ctx context.Context, keys map[string][]*telemetrytypes.TelemetryFieldKey, query qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation], requestType qbtypes.RequestType) qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation] {
+
+	// add deprecated fields only during statement building
+	// why?
+	// 1. to not fail filter expression that use deprecated cols
+	// 2. this could have been moved to metadata fetching itself, however, that
+	// would mean, they also show up in suggestions we we don't want to do
+	// 3. reason for not doing a simple append is to keep intrinsic/calculated field first so that it gets
+	// priority in multi_if sql expression
 	for fieldKeyName, fieldKey := range IntrinsicFieldsDeprecated {
 		keys[fieldKeyName] = append([]*telemetrytypes.TelemetryFieldKey{&fieldKey}, keys[fieldKeyName]...)
 	}
 	for fieldKeyName, fieldKey := range CalculatedFieldsDeprecated {
 		keys[fieldKeyName] = append([]*telemetrytypes.TelemetryFieldKey{&fieldKey}, keys[fieldKeyName]...)
 	}
-}
-
-func adjustTraceKeys(keys map[string][]*telemetrytypes.TelemetryFieldKey, query *qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation], requestType qbtypes.RequestType) []string {
-
-	mergeDeprecatedTraceKeys(keys)
 
 	// Adjust keys for alias expressions in aggregations
-	actions := querybuilder.AdjustKeysForAliasExpressions(query, requestType)
+	actions := querybuilder.AdjustKeysForAliasExpressions(&query, requestType)
 
 	/*
 		Check if user is using multiple contexts or data types for same field name
@@ -236,7 +228,7 @@ func adjustTraceKeys(keys map[string][]*telemetrytypes.TelemetryFieldKey, query
 		and make it just http.status_code and remove the duplicate entry.
 	*/
 
-	actions = append(actions, querybuilder.AdjustDuplicateKeys(query)...)
+	actions = append(actions, querybuilder.AdjustDuplicateKeys(&query)...)
 
 	/*
 		Now adjust each key to have correct context and data type
@@ -244,20 +236,24 @@ func adjustTraceKeys(keys map[string][]*telemetrytypes.TelemetryFieldKey, query
 		Reason for doing this is to not create an unexpected behavior for users
 	*/
 	for idx := range query.SelectFields {
-		actions = append(actions, adjustTraceKey(&query.SelectFields[idx], keys)...)
+		actions = append(actions, b.adjustKey(&query.SelectFields[idx], keys)...)
 	}
 	for idx := range query.GroupBy {
-		actions = append(actions, adjustTraceKey(&query.GroupBy[idx].TelemetryFieldKey, keys)...)
+		actions = append(actions, b.adjustKey(&query.GroupBy[idx].TelemetryFieldKey, keys)...)
 	}
 	for idx := range query.Order {
-		actions = append(actions, adjustTraceKey(&query.Order[idx].Key.TelemetryFieldKey, keys)...)
+		actions = append(actions, b.adjustKey(&query.Order[idx].Key.TelemetryFieldKey, keys)...)
 	}
 
-	return actions
+	for _, action := range actions {
+		// TODO: change to debug level once we are confident about the behavior
+		b.logger.InfoContext(ctx, "key adjustment action", slog.String("action", action))
+	}
+
+	return query
 }
 
-// adjustTraceKey resolves a single TelemetryFieldKey against the keys map.
-func adjustTraceKey(key *telemetrytypes.TelemetryFieldKey, keys map[string][]*telemetrytypes.TelemetryFieldKey) []string {
+func (b *traceQueryStatementBuilder) adjustKey(key *telemetrytypes.TelemetryFieldKey, keys map[string][]*telemetrytypes.TelemetryFieldKey) []string {
 
 	// for recording actions taken
 	actions := []string{}

pkg/telemetrytraces/stmt_builder_test.go

@@ -1125,13 +1125,28 @@ func TestAdjustKey(t *testing.T) {
 		},
 	}
 
+	fm := NewFieldMapper()
+	cb := NewConditionBuilder(fm)
+	mockMetadataStore := telemetrytypestest.NewMockMetadataStore()
+	fl := flaggertest.New(t)
+	aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, nil, fl)
+	statementBuilder := NewTraceQueryStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		aggExprRewriter,
+		nil,
+		fl,
+	)
+
 	for _, c := range cases {
 		t.Run(c.name, func(t *testing.T) {
 			// Create a copy of the input key to avoid modifying the original
 			key := c.inputKey
 
 			// Call adjustKey
-			adjustTraceKey(&key, c.keysMap)
+			statementBuilder.adjustKey(&key, c.keysMap)
 
 			// Verify the key was adjusted as expected
 			require.Equal(t, c.expectedKey.Name, key.Name, "key name should match")
@@ -1384,6 +1399,21 @@ func TestAdjustKeys(t *testing.T) {
 		},
 	}
 
+	fm := NewFieldMapper()
+	cb := NewConditionBuilder(fm)
+	mockMetadataStore := telemetrytypestest.NewMockMetadataStore()
+	fl := flaggertest.New(t)
+	aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, nil, fl)
+	statementBuilder := NewTraceQueryStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		aggExprRewriter,
+		nil,
+		fl,
+	)
+
 	for _, c := range cases {
 		t.Run(c.name, func(t *testing.T) {
 			// Create a deep copy of the keys map to avoid modifying the original
@@ -1394,7 +1424,7 @@ func TestAdjustKeys(t *testing.T) {
 			}
 
 			// Call adjustKeys
-			adjustTraceKeys(keysMapCopy, &c.query, qbtypes.RequestTypeScalar)
+			c.query = statementBuilder.adjustKeys(context.Background(), keysMapCopy, c.query, qbtypes.RequestTypeScalar)
 
 			// Verify select fields were adjusted
 			if c.expectedSelectFields != nil {

pkg/telemetrytraces/trace_operator_cte_builder.go

@@ -216,13 +216,6 @@ func (b *traceOperatorCTEBuilder) buildQueryCTE(ctx context.Context, queryName s
 	}
 	b.stmtBuilder.logger.DebugContext(ctx, "Retrieved keys for query", slog.String("query_name", queryName), slog.Int("keys_count", len(keys)))
 
-	// The CTE only selects spans matching the filter. Aggregations, group by
-	// and order by run later in buildFinalQuery, so RequestTypeRaw is fine here.
-	for _, action := range adjustTraceKeys(keys, query, qbtypes.RequestTypeRaw) {
-		// TODO: change to debug level once we are confident about the behavior
-		b.stmtBuilder.logger.InfoContext(ctx, "key adjustment action", slog.String("action", action))
-	}
-
 	// Build resource filter CTE for this specific query
 	resourceFilterCTEName := fmt.Sprintf("__resource_filter_%s", cteName)
 	resourceStmt, err := b.buildResourceFilterCTE(ctx, *query)
@@ -424,28 +417,21 @@ func (b *traceOperatorCTEBuilder) buildNotCTE(leftCTE, rightCTE string) (string,
 }
 
 func (b *traceOperatorCTEBuilder) buildFinalQuery(ctx context.Context, selectFromCTE string, requestType qbtypes.RequestType) (*qbtypes.Statement, error) {
-	keySelectors := b.getKeySelectors()
-	keys, _, err := b.stmtBuilder.metadataStore.GetKeysMulti(ctx, keySelectors)
-	if err != nil {
-		return nil, err
-	}
-	b.adjustOperatorKeys(ctx, keys, requestType)
-
 	switch requestType {
 	case qbtypes.RequestTypeRaw:
-		return b.buildListQuery(ctx, selectFromCTE, keys)
+		return b.buildListQuery(ctx, selectFromCTE)
 	case qbtypes.RequestTypeTimeSeries:
-		return b.buildTimeSeriesQuery(ctx, selectFromCTE, keys)
+		return b.buildTimeSeriesQuery(ctx, selectFromCTE)
 	case qbtypes.RequestTypeTrace:
-		return b.buildTraceQuery(ctx, selectFromCTE, keys)
+		return b.buildTraceQuery(ctx, selectFromCTE)
 	case qbtypes.RequestTypeScalar:
-		return b.buildScalarQuery(ctx, selectFromCTE, keys)
+		return b.buildScalarQuery(ctx, selectFromCTE)
 	default:
 		return nil, errors.NewInvalidInputf(errors.CodeInvalidInput, "unsupported request type: %s", requestType)
 	}
 }
 
-func (b *traceOperatorCTEBuilder) buildListQuery(ctx context.Context, selectFromCTE string, keys map[string][]*telemetrytypes.TelemetryFieldKey) (*qbtypes.Statement, error) {
+func (b *traceOperatorCTEBuilder) buildListQuery(ctx context.Context, selectFromCTE string) (*qbtypes.Statement, error) {
 	sb := sqlbuilder.NewSelectBuilder()
 
 	// Select core fields
@@ -467,6 +453,22 @@ func (b *traceOperatorCTEBuilder) buildListQuery(ctx context.Context, selectFrom
 		"parent_span_id": true,
 	}
 
+	// Get keys for selectFields
+	keySelectors := b.getKeySelectors()
+	for _, field := range b.operator.SelectFields {
+		keySelectors = append(keySelectors, &telemetrytypes.FieldKeySelector{
+			Name:          field.Name,
+			Signal:        telemetrytypes.SignalTraces,
+			FieldContext:  field.FieldContext,
+			FieldDataType: field.FieldDataType,
+		})
+	}
+
+	keys, _, err := b.stmtBuilder.metadataStore.GetKeysMulti(ctx, keySelectors)
+	if err != nil {
+		return nil, err
+	}
+
 	// Add selectFields using ColumnExpressionFor since we now have all base table columns
 	for _, field := range b.operator.SelectFields {
 		if selectedFields[field.Name] {
@@ -516,44 +518,6 @@ func (b *traceOperatorCTEBuilder) buildListQuery(ctx context.Context, selectFrom
 	}, nil
 }
 
-// adjustOperatorKeys runs the same key adjustments as adjustTraceKeys, but on
-// the operator's own fields. The operator has a different struct shape than
-// QueryBuilderQuery, so we copy the relevant fields into a temp query, run
-// the shared helpers, and copy the results back.
-func (b *traceOperatorCTEBuilder) adjustOperatorKeys(ctx context.Context, keys map[string][]*telemetrytypes.TelemetryFieldKey, requestType qbtypes.RequestType) {
-	mergeDeprecatedTraceKeys(keys)
-
-	tmp := qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation]{
-		Aggregations: b.operator.Aggregations,
-		SelectFields: b.operator.SelectFields,
-		GroupBy:      b.operator.GroupBy,
-		Order:        b.operator.Order,
-	}
-
-	actions := querybuilder.AdjustKeysForAliasExpressions(&tmp, requestType)
-	actions = append(actions, querybuilder.AdjustDuplicateKeys(&tmp)...)
-
-	for idx := range tmp.SelectFields {
-		actions = append(actions, adjustTraceKey(&tmp.SelectFields[idx], keys)...)
-	}
-	for idx := range tmp.GroupBy {
-		actions = append(actions, adjustTraceKey(&tmp.GroupBy[idx].TelemetryFieldKey, keys)...)
-	}
-	for idx := range tmp.Order {
-		actions = append(actions, adjustTraceKey(&tmp.Order[idx].Key.TelemetryFieldKey, keys)...)
-	}
-
-	// Copy back the slices the helpers can rewrite.
-	b.operator.Aggregations = tmp.Aggregations
-	b.operator.SelectFields = tmp.SelectFields
-	b.operator.GroupBy = tmp.GroupBy
-	b.operator.Order = tmp.Order
-
-	for _, action := range actions {
-		b.stmtBuilder.logger.InfoContext(ctx, "key adjustment action", slog.String("action", action))
-	}
-}
-
 func (b *traceOperatorCTEBuilder) getKeySelectors() []*telemetrytypes.FieldKeySelector {
 	var keySelectors []*telemetrytypes.FieldKeySelector
 
@@ -581,15 +545,6 @@ func (b *traceOperatorCTEBuilder) getKeySelectors() []*telemetrytypes.FieldKeySe
 		})
 	}
 
-	for _, sf := range b.operator.SelectFields {
-		keySelectors = append(keySelectors, &telemetrytypes.FieldKeySelector{
-			Name:          sf.Name,
-			Signal:        telemetrytypes.SignalTraces,
-			FieldContext:  sf.FieldContext,
-			FieldDataType: sf.FieldDataType,
-		})
-	}
-
 	for i := range keySelectors {
 		keySelectors[i].Signal = telemetrytypes.SignalTraces
 	}
@@ -597,7 +552,7 @@ func (b *traceOperatorCTEBuilder) getKeySelectors() []*telemetrytypes.FieldKeySe
 	return keySelectors
 }
 
-func (b *traceOperatorCTEBuilder) buildTimeSeriesQuery(ctx context.Context, selectFromCTE string, keys map[string][]*telemetrytypes.TelemetryFieldKey) (*qbtypes.Statement, error) {
+func (b *traceOperatorCTEBuilder) buildTimeSeriesQuery(ctx context.Context, selectFromCTE string) (*qbtypes.Statement, error) {
 	sb := sqlbuilder.NewSelectBuilder()
 
 	sb.Select(fmt.Sprintf(
@@ -605,6 +560,12 @@ func (b *traceOperatorCTEBuilder) buildTimeSeriesQuery(ctx context.Context, sele
 		int64(b.operator.StepInterval.Seconds()),
 	))
 
+	keySelectors := b.getKeySelectors()
+	keys, _, err := b.stmtBuilder.metadataStore.GetKeysMulti(ctx, keySelectors)
+	if err != nil {
+		return nil, err
+	}
+
 	var allGroupByArgs []any
 
 	for _, gb := range b.operator.GroupBy {
@@ -683,7 +644,8 @@ func (b *traceOperatorCTEBuilder) buildTimeSeriesQuery(ctx context.Context, sele
 	combinedArgs := append(allGroupByArgs, allAggChArgs...)
 
 	// Add HAVING clause if specified
-	if err := b.addHavingClause(sb); err != nil {
+	err = b.addHavingClause(sb)
+	if err != nil {
 		return nil, err
 	}
 
@@ -710,11 +672,17 @@ func (b *traceOperatorCTEBuilder) buildTraceSummaryCTE(selectFromCTE string) {
 	b.addCTE("trace_summary", sql, args, []string{"all_spans", selectFromCTE})
 }
 
-func (b *traceOperatorCTEBuilder) buildTraceQuery(ctx context.Context, selectFromCTE string, keys map[string][]*telemetrytypes.TelemetryFieldKey) (*qbtypes.Statement, error) {
+func (b *traceOperatorCTEBuilder) buildTraceQuery(ctx context.Context, selectFromCTE string) (*qbtypes.Statement, error) {
 	b.buildTraceSummaryCTE(selectFromCTE)
 
 	sb := sqlbuilder.NewSelectBuilder()
 
+	keySelectors := b.getKeySelectors()
+	keys, _, err := b.stmtBuilder.metadataStore.GetKeysMulti(ctx, keySelectors)
+	if err != nil {
+		return nil, err
+	}
+
 	var allGroupByArgs []any
 
 	for _, gb := range b.operator.GroupBy {
@@ -796,7 +764,8 @@ func (b *traceOperatorCTEBuilder) buildTraceQuery(ctx context.Context, selectFro
 		sb.GroupBy(groupByKeys...)
 	}
 
-	if err := b.addHavingClause(sb); err != nil {
+	err = b.addHavingClause(sb)
+	if err != nil {
 		return nil, err
 	}
 
@@ -852,9 +821,15 @@ func (b *traceOperatorCTEBuilder) buildTraceQuery(ctx context.Context, selectFro
 	}, nil
 }
 
-func (b *traceOperatorCTEBuilder) buildScalarQuery(ctx context.Context, selectFromCTE string, keys map[string][]*telemetrytypes.TelemetryFieldKey) (*qbtypes.Statement, error) {
+func (b *traceOperatorCTEBuilder) buildScalarQuery(ctx context.Context, selectFromCTE string) (*qbtypes.Statement, error) {
 	sb := sqlbuilder.NewSelectBuilder()
 
+	keySelectors := b.getKeySelectors()
+	keys, _, err := b.stmtBuilder.metadataStore.GetKeysMulti(ctx, keySelectors)
+	if err != nil {
+		return nil, err
+	}
+
 	var allGroupByArgs []any
 
 	for _, gb := range b.operator.GroupBy {
@@ -936,7 +911,8 @@ func (b *traceOperatorCTEBuilder) buildScalarQuery(ctx context.Context, selectFr
 	combinedArgs := append(allGroupByArgs, allAggChArgs...)
 
 	// Add HAVING clause if specified
-	if err := b.addHavingClause(sb); err != nil {
+	err = b.addHavingClause(sb)
+	if err != nil {
 		return nil, err
 	}
 

pkg/telemetrytraces/trace_operator_cte_builder_test.go

@@ -14,24 +14,6 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-func newTestTraceOperatorStatementBuilder(t *testing.T) *traceOperatorStatementBuilder {
-	t.Helper()
-	fm := NewFieldMapper()
-	cb := NewConditionBuilder(fm)
-	mockMetadataStore := telemetrytypestest.NewMockMetadataStore()
-	mockMetadataStore.KeysMap = buildCompleteFieldKeyMap()
-	fl := flaggertest.New(t)
-	aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, nil, fl)
-	traceStmtBuilder := NewTraceQueryStatementBuilder(
-		instrumentationtest.New().ToProviderSettings(),
-		mockMetadataStore, fm, cb, aggExprRewriter, nil, fl,
-	)
-	return NewTraceOperatorStatementBuilder(
-		instrumentationtest.New().ToProviderSettings(),
-		mockMetadataStore, fm, cb, traceStmtBuilder, aggExprRewriter, fl,
-	)
-}
-
 func TestTraceOperatorStatementBuilder(t *testing.T) {
 	cases := []struct {
 		name           string
@@ -481,7 +463,32 @@ func TestTraceOperatorStatementBuilder(t *testing.T) {
 		},
 	}
 
-	statementBuilder := newTestTraceOperatorStatementBuilder(t)
+	fm := NewFieldMapper()
+	cb := NewConditionBuilder(fm)
+	mockMetadataStore := telemetrytypestest.NewMockMetadataStore()
+	mockMetadataStore.KeysMap = buildCompleteFieldKeyMap()
+	fl := flaggertest.New(t)
+	aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, nil, fl)
+
+	traceStmtBuilder := NewTraceQueryStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		aggExprRewriter,
+		nil,
+		fl,
+	)
+
+	statementBuilder := NewTraceOperatorStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		traceStmtBuilder,
+		aggExprRewriter,
+		fl,
+	)
 
 	for _, c := range cases {
 		t.Run(c.name, func(t *testing.T) {
@@ -572,7 +579,32 @@ func TestTraceOperatorStatementBuilderErrors(t *testing.T) {
 		},
 	}
 
-	statementBuilder := newTestTraceOperatorStatementBuilder(t)
+	fm := NewFieldMapper()
+	cb := NewConditionBuilder(fm)
+	mockMetadataStore := telemetrytypestest.NewMockMetadataStore()
+	mockMetadataStore.KeysMap = buildCompleteFieldKeyMap()
+	fl := flaggertest.New(t)
+	aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, nil, fl)
+
+	traceStmtBuilder := NewTraceQueryStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		aggExprRewriter,
+		nil,
+		fl,
+	)
+
+	statementBuilder := NewTraceOperatorStatementBuilder(
+		instrumentationtest.New().ToProviderSettings(),
+		mockMetadataStore,
+		fm,
+		cb,
+		traceStmtBuilder,
+		aggExprRewriter,
+		fl,
+	)
 
 	for _, c := range cases {
 		t.Run(c.name, func(t *testing.T) {
@@ -594,142 +626,3 @@ func TestTraceOperatorStatementBuilderErrors(t *testing.T) {
 		})
 	}
 }
-
-func TestTraceOperatorStatementBuilderAdjustsKeys(t *testing.T) {
-	cases := []struct {
-		name          string
-		requestType   qbtypes.RequestType
-		operator      qbtypes.QueryBuilderTraceOperator
-		builderFilter string
-		wantSQL       string
-		wantArgs      []any
-	}{
-		{
-			name:        "deprecated duration filter in referenced builder query",
-			requestType: qbtypes.RequestTypeRaw,
-			operator: qbtypes.QueryBuilderTraceOperator{
-				Expression: "A",
-				Limit:      10,
-			},
-			builderFilter: "durationNano = '3s'",
-			wantSQL:       "duration_nano = ?",
-			wantArgs:      []any{int64(3000000000)},
-		},
-		{
-			name:        "context-prefixed aggregation alias in order by",
-			requestType: qbtypes.RequestTypeScalar,
-			operator: qbtypes.QueryBuilderTraceOperator{
-				Expression: "A",
-				Aggregations: []qbtypes.TraceAggregation{
-					{
-						Expression: "count()",
-						Alias:      "span.count_",
-					},
-				},
-				Order: []qbtypes.OrderBy{
-					{
-						Key: qbtypes.OrderByKey{
-							TelemetryFieldKey: telemetrytypes.TelemetryFieldKey{
-								Name:         "count_",
-								FieldContext: telemetrytypes.FieldContextSpan,
-							},
-						},
-						Direction: qbtypes.OrderDirectionDesc,
-					},
-				},
-			},
-			wantSQL: "ORDER BY __result_0 desc",
-		},
-	}
-
-	statementBuilder := newTestTraceOperatorStatementBuilder(t)
-
-	for _, c := range cases {
-		t.Run(c.name, func(t *testing.T) {
-			err := c.operator.ParseExpression()
-			require.NoError(t, err)
-
-			filter := c.builderFilter
-			if filter == "" {
-				filter = "service.name = 'frontend'"
-			}
-
-			q, err := statementBuilder.Build(
-				context.Background(),
-				1747947419000,
-				1747983448000,
-				c.requestType,
-				c.operator,
-				&qbtypes.CompositeQuery{
-					Queries: []qbtypes.QueryEnvelope{
-						{
-							Type: qbtypes.QueryTypeBuilder,
-							Spec: qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation]{
-								Name:   "A",
-								Signal: telemetrytypes.SignalTraces,
-								Filter: &qbtypes.Filter{Expression: filter},
-							},
-						},
-					},
-				},
-			)
-
-			require.NoError(t, err)
-			require.Contains(t, q.Query, c.wantSQL)
-			for _, arg := range c.wantArgs {
-				require.Contains(t, q.Args, arg)
-			}
-		})
-	}
-}
-
-// TestTraceOperatorStatementBuilderDeduplicatesKeys checks that a trace
-// operator with the same field name listed twice in GroupBy (once with a
-// context, once without) ends up with a single column in the outer SELECT
-// and a single entry in GROUP BY.
-func TestTraceOperatorStatementBuilderDeduplicatesKeys(t *testing.T) {
-	statementBuilder := newTestTraceOperatorStatementBuilder(t)
-
-	operator := qbtypes.QueryBuilderTraceOperator{
-		Expression: "A",
-		Aggregations: []qbtypes.TraceAggregation{
-			{Expression: "count()"},
-		},
-		GroupBy: []qbtypes.GroupByKey{
-			{TelemetryFieldKey: telemetrytypes.TelemetryFieldKey{
-				Name:         "http.method",
-				FieldContext: telemetrytypes.FieldContextAttribute,
-			}},
-			// Same name, no context — should be merged with the entry above.
-			{TelemetryFieldKey: telemetrytypes.TelemetryFieldKey{
-				Name: "http.method",
-			}},
-		},
-	}
-	require.NoError(t, operator.ParseExpression())
-
-	q, err := statementBuilder.Build(
-		context.Background(),
-		1747947419000,
-		1747983448000,
-		qbtypes.RequestTypeScalar,
-		operator,
-		&qbtypes.CompositeQuery{
-			Queries: []qbtypes.QueryEnvelope{
-				{
-					Type: qbtypes.QueryTypeBuilder,
-					Spec: qbtypes.QueryBuilderQuery[qbtypes.TraceAggregation]{
-						Name:   "A",
-						Signal: telemetrytypes.SignalTraces,
-						Filter: &qbtypes.Filter{Expression: "service.name = 'frontend'"},
-					},
-				},
-			},
-		},
-	)
-
-	require.NoError(t, err)
-
-	require.Contains(t, q.Query,
-		"SELECT toString(multiIf(mapContains(attributes_string, 'http.method') = ?, attributes_string['http.method'], NULL)) AS `http.method`, count() AS __result_0 FROM A GROUP BY `http.method` ORDER BY __result_0 DESC")
-}

pkg/types/spantypes/flamegraph_span.go

@@ -0,0 +1,95 @@
+package spantypes
+
+import (
+	"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
+)
+
+type FlamegraphSpan struct {
+	SpanID       string            `json:"spanId"`
+	ParentSpanID string            `json:"parentSpanId"`
+	Timestamp    uint64            `json:"timestamp"`
+	DurationNano uint64            `json:"durationNano"`
+	HasError     bool              `json:"hasError"`
+	Name         string            `json:"name"`
+	Level        int64             `json:"level"`
+	Events       []Event           `json:"event" required:"true" nullable:"false"`
+	Attributes   map[string]any    `json:"attributes,omitempty"`
+	Resource     map[string]string `json:"resource,omitempty"`
+	Children     []*FlamegraphSpan `json:"-"` // internal tree use only
+}
+
+// FlamegraphLevel groups span IDs at a single level within the selected window.
+type FlamegraphLevel struct {
+	Level   int64
+	SpanIDs []string
+}
+
+type PostableFlamegraph struct {
+	SelectedSpanID string                             `json:"selectedSpanId"`
+	SelectFields   []telemetrytypes.TelemetryFieldKey `json:"selectFields,omitempty"`
+}
+
+// GettableFlamegraphTrace is the response for the v3 flamegraph API.
+type GettableFlamegraphTrace struct {
+	Spans                [][]*FlamegraphSpan `json:"spans" required:"true" nullable:"false"`
+	StartTimestampMillis int64               `json:"startTimestampMillis" required:"true"`
+	EndTimestampMillis   int64               `json:"endTimestampMillis" required:"true"`
+	HasMore              bool                `json:"hasMore" required:"true"`
+}
+
+func NewGettableFlamegraphTrace(spans [][]*FlamegraphSpan, startMs, endMs int64, hasMore bool) *GettableFlamegraphTrace {
+	return &GettableFlamegraphTrace{
+		Spans:                spans,
+		StartTimestampMillis: startMs,
+		EndTimestampMillis:   endMs,
+		HasMore:              hasMore,
+	}
+}
+
+func NewFlamegraphSpanFromStorable(s *StorableSpan, level int64, selectFields []telemetrytypes.TelemetryFieldKey) *FlamegraphSpan {
+	span := &FlamegraphSpan{
+		SpanID:       s.SpanID,
+		ParentSpanID: s.ParentSpanID,
+		Timestamp:    uint64(s.StartTime.UnixNano()),
+		DurationNano: s.DurationNano,
+		HasError:     s.HasError,
+		Name:         s.Name,
+		Level:        level,
+		Events:       s.UnmarshalledEvents(),
+	}
+	if len(selectFields) == 0 {
+		return span
+	}
+	for _, field := range selectFields {
+		switch field.FieldContext {
+		case telemetrytypes.FieldContextResource:
+			if v, ok := s.ResourcesString[field.Name]; ok && v != "" {
+				if span.Resource == nil {
+					span.Resource = make(map[string]string)
+				}
+				span.Resource[field.Name] = v
+			}
+		case telemetrytypes.FieldContextAttribute:
+			if v := s.AttributeValue(field.Name); v != nil {
+				if span.Attributes == nil {
+					span.Attributes = make(map[string]any)
+				}
+				span.Attributes[field.Name] = v
+			}
+		}
+	}
+	return span
+}
+
+// FlamegraphWindowSpanIDs collects all span IDs from a level window into a flat slice.
+func FlamegraphWindowSpanIDs(window []FlamegraphLevel) []string {
+	total := 0
+	for _, lvl := range window {
+		total += len(lvl.SpanIDs)
+	}
+	ids := make([]string, 0, total)
+	for _, lvl := range window {
+		ids = append(ids, lvl.SpanIDs...)
+	}
+	return ids
+}

pkg/types/spantypes/flamegraph_trace.go

@@ -0,0 +1,213 @@
+package spantypes
+
+import (
+	"sort"
+
+	"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
+)
+
+// FlamegraphTrace holds the level wise tree built from minimal spans.
+type FlamegraphTrace struct {
+	roots     []*FlamegraphSpan
+	nodeByID  map[string]*FlamegraphSpan
+	startTime uint64
+	endTime   uint64
+}
+
+func NewFlamegraphTraceFromMinimal(spans []MinimalSpan) *FlamegraphTrace {
+	t := &FlamegraphTrace{
+		nodeByID: make(map[string]*FlamegraphSpan, len(spans)),
+	}
+	for i := range spans {
+		node := spans[i].ToFlamegraphSpan()
+		t.updateTimeRange(node.Timestamp, node.DurationNano)
+		t.nodeByID[node.SpanID] = node
+	}
+	t.buildSpanTree()
+	return t
+}
+
+func NewFlamegraphTraceFromStorable(spans []StorableSpan, selectFields []telemetrytypes.TelemetryFieldKey) *FlamegraphTrace {
+	t := &FlamegraphTrace{
+		nodeByID: make(map[string]*FlamegraphSpan, len(spans)),
+	}
+	for i := range spans {
+		node := NewFlamegraphSpanFromStorable(&spans[i], 0, selectFields) // level is set later by BFS
+		t.updateTimeRange(node.Timestamp, node.DurationNano)
+		t.nodeByID[node.SpanID] = node
+	}
+	t.buildSpanTree()
+	return t
+}
+
+func (t *FlamegraphTrace) GetAllLevels() [][]*FlamegraphSpan {
+	var result [][]*FlamegraphSpan
+	for _, root := range t.roots {
+		currentLevel := []*FlamegraphSpan{root}
+		for depth := int64(0); len(currentLevel) > 0; depth++ {
+			var nextLevel []*FlamegraphSpan
+			for _, node := range currentLevel {
+				node.Level = depth
+				nextLevel = append(nextLevel, node.Children...)
+			}
+			result = append(result, currentLevel)
+			currentLevel = nextLevel
+		}
+	}
+
+	return result
+}
+
+// GetSelectedLevels returns the window of levels around selectedSpanID with sampling applied to dense levels.
+func (t *FlamegraphTrace) GetSelectedLevels(selectedSpanID string, levelLimit, spansPerLevel, topLatencyCount, bucketCount int) []FlamegraphLevel {
+	allLevels := t.GetAllLevels()
+
+	selectedIndex := getLevelIndex(allLevels, selectedSpanID)
+
+	// 40% window above level with selected span and 60% below that
+	beforeSelectedLevel := int(float64(levelLimit) * 0.4)
+	startLevel := max(0, selectedIndex-beforeSelectedLevel)
+	endLevel := min(len(allLevels), startLevel+levelLimit)
+
+	result := make([]FlamegraphLevel, 0, endLevel-startLevel)
+	for i := startLevel; i < endLevel; i++ {
+		spans := allLevels[i]
+		sampled := spans
+		if len(spans) > spansPerLevel {
+			sampled = t.sampleLevel(spans, selectedSpanID, i == selectedIndex, topLatencyCount, bucketCount)
+		}
+		if len(sampled) == 0 {
+			continue
+		}
+		spanIDs := make([]string, len(sampled))
+		for j, s := range sampled {
+			spanIDs[j] = s.SpanID
+		}
+		result = append(result, FlamegraphLevel{Level: spans[0].Level, SpanIDs: spanIDs})
+	}
+
+	return result
+}
+
+func (t *FlamegraphTrace) EnrichSelectedSpans(selectedSpans []FlamegraphLevel, fullSpans []StorableSpan, selectFields []telemetrytypes.TelemetryFieldKey) [][]*FlamegraphSpan {
+	fullByID := make(map[string]*StorableSpan, len(fullSpans))
+	for i := range fullSpans {
+		fullByID[fullSpans[i].SpanID] = &fullSpans[i]
+	}
+
+	result := make([][]*FlamegraphSpan, len(selectedSpans))
+	for i, lvl := range selectedSpans {
+		result[i] = make([]*FlamegraphSpan, 0, len(lvl.SpanIDs))
+		for _, spanID := range lvl.SpanIDs {
+			if full, ok := fullByID[spanID]; ok {
+				result[i] = append(result[i], NewFlamegraphSpanFromStorable(full, lvl.Level, selectFields))
+			} else if lean, ok := t.nodeByID[spanID]; ok {
+				result[i] = append(result[i], lean)
+			}
+		}
+	}
+	return result
+}
+
+func (t *FlamegraphTrace) updateTimeRange(timestamp, durationNano uint64) {
+	if t.startTime == 0 || timestamp < t.startTime {
+		t.startTime = timestamp
+	}
+	if end := timestamp + durationNano; end > t.endTime {
+		t.endTime = end
+	}
+}
+
+func (t *FlamegraphTrace) buildSpanTree() {
+	for _, node := range t.nodeByID {
+		if node.ParentSpanID != "" {
+			if parent, ok := t.nodeByID[node.ParentSpanID]; ok {
+				parent.Children = append(parent.Children, node)
+			} else {
+				missing := &FlamegraphSpan{
+					SpanID:       node.ParentSpanID,
+					Name:         "Missing Span",
+					Timestamp:    node.Timestamp,
+					DurationNano: node.DurationNano,
+					Children:     []*FlamegraphSpan{node},
+				}
+				t.nodeByID[missing.SpanID] = missing
+				t.roots = append(t.roots, missing)
+			}
+		} else if flamegraphSpanIndex(t.roots, node.SpanID) == -1 {
+			t.roots = append(t.roots, node)
+		}
+	}
+
+	sort.Slice(t.roots, func(i, j int) bool {
+		if t.roots[i].Timestamp == t.roots[j].Timestamp {
+			return t.roots[i].SpanID < t.roots[j].SpanID
+		}
+		return t.roots[i].Timestamp < t.roots[j].Timestamp
+	})
+}
+
+func (t *FlamegraphTrace) sampleLevel(spans []*FlamegraphSpan, selectedSpanID string, isSelectedLevel bool, topLatencyCount, bucketCount int) []*FlamegraphSpan {
+	sorted := make([]*FlamegraphSpan, len(spans))
+	copy(sorted, spans)
+	sort.Slice(sorted, func(i, j int) bool {
+		return sorted[i].DurationNano > sorted[j].DurationNano
+	})
+
+	topK := min(topLatencyCount, len(sorted))
+	sampled := make([]*FlamegraphSpan, topK, topK+1)
+	copy(sampled, sorted[:topK])
+
+	if isSelectedLevel {
+		for _, span := range sorted {
+			if span.SpanID == selectedSpanID {
+				sampled = append(sampled, span)
+				break
+			}
+		}
+	}
+
+	return append(sampled, t.bucketSampleSpans(sorted, bucketCount)...)
+}
+
+func (t *FlamegraphTrace) bucketSampleSpans(sorted []*FlamegraphSpan, bucketCount int) []*FlamegraphSpan {
+	bucketSize := (t.endTime - t.startTime) / uint64(bucketCount)
+	if bucketSize == 0 {
+		bucketSize = 1
+	}
+	buckets := make([][]*FlamegraphSpan, bucketCount)
+	for _, span := range sorted {
+		if span.Timestamp < t.startTime || span.Timestamp > t.endTime {
+			continue
+		}
+		idx := min(int((span.Timestamp-t.startTime)/bucketSize), bucketCount-1)
+		if len(buckets[idx]) < 2 {
+			buckets[idx] = append(buckets[idx], span)
+		}
+	}
+	var result []*FlamegraphSpan
+	for _, bucket := range buckets {
+		result = append(result, bucket...)
+	}
+	return result
+}
+
+func getLevelIndex(levels [][]*FlamegraphSpan, spanID string) int {
+	for i, lvl := range levels {
+		for _, span := range lvl {
+			if span.SpanID == spanID {
+				return i
+			}
+		}
+	}
+	return 0
+}
+
+func flamegraphSpanIndex(spans []*FlamegraphSpan, spanID string) int {
+	for i, s := range spans {
+		if s != nil && s.SpanID == spanID {
+			return i
+		}
+	}
+	return -1
+}

pkg/types/spantypes/store.go

@@ -30,6 +30,7 @@ type TraceStore interface {
 	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)
+	GetFlamegraphSpans(ctx context.Context, traceID string, start, end time.Time, spanIDs []string) ([]StorableSpan, error)
 
 	GetSpanCountByField(ctx context.Context, traceID string, summary *TraceSummary, fieldKey telemetrytypes.TelemetryFieldKey) (map[string]uint64, error)
 	GetSpanDurationByField(ctx context.Context, traceID string, summary *TraceSummary, fieldKey telemetrytypes.TelemetryFieldKey) (map[string]uint64, error)

pkg/types/spantypes/waterfall_span.go

@@ -164,6 +164,17 @@ func (item *MinimalSpan) ToWaterfallSpan(traceID string) *WaterfallSpan {
 	}
 }
 
+func (item *MinimalSpan) ToFlamegraphSpan() *FlamegraphSpan {
+	return &FlamegraphSpan{
+		SpanID:       item.SpanID,
+		ParentSpanID: item.ParentSpanID,
+		Timestamp:    uint64(item.StartTime.UnixNano()),
+		DurationNano: item.DurationNano,
+		HasError:     item.HasError,
+		Children:     make([]*FlamegraphSpan, 0),
+	}
+}
+
 // NewMissingWaterfallSpan creates a synthetic placeholder span for a parent that has no recorded data.
 func NewMissingWaterfallSpan(spanID, traceID string, timeUnixNano, durationNano uint64) *WaterfallSpan {
 	return &WaterfallSpan{
@@ -267,6 +278,19 @@ func (ws *WaterfallSpan) getPathToSelectedSpanID(selectedSpanID string) ([]strin
 	return nil, false
 }
 
+func (item *StorableSpan) AttributeValue(name string) any {
+	if v, ok := item.AttributesString[name]; ok {
+		return v
+	}
+	if v, ok := item.AttributesNumber[name]; ok {
+		return v
+	}
+	if v, ok := item.AttributesBool[name]; ok {
+		return v
+	}
+	return nil
+}
+
 func (item *StorableSpan) Attributes() map[string]any {
 	attributes := make(map[string]any, len(item.AttributesString)+len(item.AttributesNumber)+len(item.AttributesBool))
 	for k, v := range item.AttributesString {

tests/fixtures/querier.py

@@ -459,57 +459,6 @@ def find_named_result(
     )
 
 
-def assert_scalar_value(
-    response: requests.Response,
-    name: str,
-    expected: Any,
-    *,
-    row: int = 0,
-    col: int = 0,
-) -> None:
-    """Assert that the named scalar result has `expected` at data[row][col]."""
-    result = find_named_result(response.json()["data"]["data"]["results"], name)
-    assert result is not None, f"no result for query {name}"
-    assert result["data"][row][col] == expected, f"expected {expected} at [{row}][{col}], got {result['data'][row][col]}"
-
-
-def assert_grouped_scalar(
-    response: requests.Response,
-    name: str,
-    *,
-    expected_groups: int,
-    expected_columns: int,
-    last_col_value: Any | None = None,
-) -> None:
-    """Assert grouped scalar result has the expected column count and group count.
-    If `last_col_value` is set and there is exactly one group, also assert the
-    last column of that single row equals it (a common aggregation-value check)."""
-    result = find_named_result(response.json()["data"]["data"]["results"], name)
-    assert result is not None, f"no result for query {name}"
-    columns = result["columns"]
-    rows = result["data"]
-    assert len(columns) == expected_columns, f"expected {expected_columns} columns, got {len(columns)}: {columns}"
-    assert len(rows) == expected_groups, f"expected {expected_groups} groups, got {len(rows)}: {rows}"
-    if last_col_value is not None and expected_groups == 1:
-        assert rows[0][-1] == last_col_value, f"expected last col {last_col_value}, got row {rows[0]}"
-
-
-def assert_raw_row_subset(
-    response: requests.Response,
-    name: str,
-    expected: dict[str, Any],
-    *,
-    row: int = 0,
-) -> None:
-    """Assert that the named raw result's rows[row]['data'] is a superset of `expected`."""
-    result = find_named_result(response.json()["data"]["data"]["results"], name)
-    assert result is not None, f"no result for query {name}"
-    rows = result["rows"]
-    assert rows is not None, f"no rows for query {name}"
-    data = rows[row]["data"]
-    assert expected.items() <= data.items(), f"expected subset {expected}, got data {data}"
-
-
 def build_scalar_query(
     name: str,
     signal: str,

tests/integration/tests/querier/15_trace_operator.py

@@ -25,22 +25,13 @@ returnSpansFrom="A"
 from collections.abc import Callable
 from datetime import UTC, datetime, timedelta
 from http import HTTPStatus
-from typing import Any
 
 import pytest
 import requests
 
 from fixtures import types
 from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
-from fixtures.querier import (
-    assert_grouped_scalar,
-    assert_raw_row_subset,
-    assert_scalar_value,
-    format_timestamp,
-    generate_traces_with_corrupt_metadata,
-    get_rows,
-    make_query_request,
-)
+from fixtures.querier import get_rows
 from fixtures.traces import TraceIdGenerator, Traces, TracesKind, TracesStatusCode
 
 
@@ -443,173 +434,3 @@ def test_trace_operator(
     )
     assert response.status_code == HTTPStatus.OK, f"HTTP {response.status_code}: {response.text}"
     assert case["validate"](response), f"validation failed: {response.json()}"
-
-
-def _expected_trace_subset(trace: Traces) -> dict[str, Any]:
-    return {
-        "duration_nano": trace.duration_nano,
-        "name": trace.name,
-        "parent_span_id": trace.parent_span_id,
-        "span_id": trace.span_id,
-        "timestamp": format_timestamp(trace.timestamp),
-        "trace_id": trace.trace_id,
-    }
-
-
-@pytest.mark.parametrize(
-    "payload_factory,request_type,assert_result",
-    [
-        # Case 1: CTE filter uses the deprecated intrinsic field `durationNano`.
-        pytest.param(
-            lambda traces: [
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "A",
-                        "signal": "traces",
-                        "filter": {"expression": 'durationNano = "3s"'},
-                    },
-                },
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "B",
-                        "signal": "traces",
-                        "filter": {"expression": 'durationNano = "5s"'},
-                    },
-                },
-                {
-                    "type": "builder_trace_operator",
-                    "spec": {
-                        "name": "C",
-                        "expression": "A => B",
-                        "limit": 1,
-                    },
-                },
-            ],
-            "raw",
-            lambda response, traces: assert_raw_row_subset(response, "C", _expected_trace_subset(traces[0])),
-            id="deprecated-intrinsic-filter",
-        ),
-        # Case 2: CTE filter uses the deprecated calculated field `responseStatusCode`.
-        pytest.param(
-            lambda traces: [
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "A",
-                        "signal": "traces",
-                        "filter": {"expression": 'responseStatusCode = "200"'},
-                    },
-                },
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "B",
-                        "signal": "traces",
-                        "filter": {"expression": 'durationNano = "5s"'},
-                    },
-                },
-                {
-                    "type": "builder_trace_operator",
-                    "spec": {
-                        "name": "C",
-                        "expression": "A => B",
-                        "limit": 1,
-                    },
-                },
-            ],
-            "raw",
-            lambda response, traces: assert_raw_row_subset(response, "C", _expected_trace_subset(traces[0])),
-            id="deprecated-calculated-filter",
-        ),
-        # Case 3: order by uses `count_` with fieldContext `span`, which has
-        # to be rewritten to the aggregation alias `span.count_`.
-        pytest.param(
-            lambda traces: [
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "A",
-                        "signal": "traces",
-                        "aggregations": [{"expression": "count()"}],
-                    },
-                },
-                {
-                    "type": "builder_trace_operator",
-                    "spec": {
-                        "name": "C",
-                        "expression": "A",
-                        "aggregations": [{"expression": "count()", "alias": "span.count_"}],
-                        "order": [{"key": {"name": "count_", "fieldContext": "span"}, "direction": "desc"}],
-                    },
-                },
-            ],
-            "scalar",
-            lambda response, traces: assert_scalar_value(response, "C", len(traces)),
-            id="context-prefixed-aggregation-alias-order",
-        ),
-        # Case 4: group by lists `cloud.provider` twice (once with a resource
-        # context, once without).
-        pytest.param(
-            lambda traces: [
-                {
-                    "type": "builder_query",
-                    "spec": {
-                        "name": "A",
-                        "signal": "traces",
-                        "disabled": True,
-                        "aggregations": [{"expression": "count()"}],
-                    },
-                },
-                {
-                    "type": "builder_trace_operator",
-                    "spec": {
-                        "name": "C",
-                        "expression": "A",
-                        "aggregations": [{"expression": "count()"}],
-                        "groupBy": [
-                            {"name": "cloud.provider", "fieldContext": "resource"},
-                            {"name": "cloud.provider"},
-                        ],
-                    },
-                },
-            ],
-            "scalar",
-            lambda response, traces: assert_grouped_scalar(response, "C", expected_groups=1, expected_columns=2, last_col_value=len(traces)),
-            id="duplicate-group-by-deduplicated",
-        ),
-    ],
-)
-def test_trace_operator_with_adjusted_keys(
-    signoz: types.SigNoz,
-    create_user_admin: None,  # pylint: disable=unused-argument
-    get_token: Callable[[str, str], str],
-    insert_traces: Callable[[list[Traces]], None],
-    payload_factory: Callable[[list[Traces]], list[dict[str, Any]]],
-    request_type: str,
-    assert_result: Callable[[requests.Response, list[Traces]], None],
-) -> None:
-    """
-    Trace operators build a CTE per referenced builder query and an outer
-    query on top. Both layers need the same key adjustment as regular trace
-    queries, otherwise deprecated keys and context-prefixed aliases don't
-    resolve.
-    """
-    traces = generate_traces_with_corrupt_metadata()
-    insert_traces(traces)
-    payload = payload_factory(traces)
-
-    token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
-
-    response = make_query_request(
-        signoz,
-        token,
-        start_ms=int((datetime.now(tz=UTC) - timedelta(minutes=5)).timestamp() * 1000),
-        end_ms=int(datetime.now(tz=UTC).timestamp() * 1000),
-        request_type=request_type,
-        queries=payload,
-    )
-
-    assert response.status_code == HTTPStatus.OK, response.text
-    assert_result(response, traces)