Add flow packet clustering

This commit is contained in:
dirtydishes 2025-12-27 20:25:32 -05:00
parent a21d513f32
commit 6c376b26dc
8 changed files with 347 additions and 11 deletions

View file

@ -2,3 +2,5 @@ export const STREAM_OPTION_PRINTS = "OPTIONS_PRINTS";
export const SUBJECT_OPTION_PRINTS = "options.prints";
export const STREAM_EQUITY_PRINTS = "EQUITY_PRINTS";
export const SUBJECT_EQUITY_PRINTS = "equities.prints";
export const STREAM_FLOW_PACKETS = "FLOW_PACKETS";
export const SUBJECT_FLOW_PACKETS = "flow.packets";

View file

@ -1,6 +1,6 @@
import { createClient, type ClickHouseClient } from "@clickhouse/client";
import { EquityPrintSchema, OptionPrintSchema } from "@islandflow/types";
import type { EquityPrint, OptionPrint } from "@islandflow/types";
import { EquityPrintSchema, FlowPacketSchema, OptionPrintSchema } from "@islandflow/types";
import type { EquityPrint, FlowPacket, OptionPrint } from "@islandflow/types";
import {
normalizeOptionPrint,
optionPrintsTableDDL,
@ -11,6 +11,13 @@ import {
EQUITY_PRINTS_TABLE,
normalizeEquityPrint
} from "./equity-prints";
import {
FLOW_PACKETS_TABLE,
flowPacketsTableDDL,
fromFlowPacketRecord,
toFlowPacketRecord,
type FlowPacketRecord
} from "./flow-packets";
export type ClickHouseOptions = {
url: string;
@ -44,6 +51,14 @@ export const ensureEquityPrintsTable = async (
});
};
export const ensureFlowPacketsTable = async (
client: ClickHouseClient
): Promise<void> => {
await client.exec({
query: flowPacketsTableDDL()
});
};
export const insertOptionPrint = async (
client: ClickHouseClient,
print: OptionPrint
@ -68,6 +83,18 @@ export const insertEquityPrint = async (
});
};
export const insertFlowPacket = async (
client: ClickHouseClient,
packet: FlowPacket
): Promise<void> => {
const record = toFlowPacketRecord(packet);
await client.insert({
table: FLOW_PACKETS_TABLE,
values: [record],
format: "JSONEachRow"
});
};
const clampLimit = (limit: number): number => {
if (!Number.isFinite(limit)) {
return 100;
@ -149,6 +176,26 @@ const normalizeEquityRow = (row: unknown): unknown => {
return row;
};
const normalizeFlowPacketRow = (row: unknown): FlowPacketRecord | null => {
if (!row || typeof row !== "object") {
return null;
}
const record = row as Record<string, unknown>;
return {
source_ts: coerceNumber(record.source_ts) as number,
ingest_ts: coerceNumber(record.ingest_ts) as number,
seq: coerceNumber(record.seq) as number,
trace_id: String(record.trace_id ?? ""),
id: String(record.id ?? ""),
members: Array.isArray(record.members)
? record.members.map((value) => String(value))
: [],
features_json: String(record.features_json ?? "{}"),
join_quality_json: String(record.join_quality_json ?? "{}")
};
};
export const fetchRecentOptionPrints = async (
client: ClickHouseClient,
limit: number
@ -177,6 +224,24 @@ export const fetchRecentEquityPrints = async (
return EquityPrintSchema.array().parse(rows.map(normalizeEquityRow));
};
export const fetchRecentFlowPackets = async (
client: ClickHouseClient,
limit: number
): Promise<FlowPacket[]> => {
const safeLimit = clampLimit(limit);
const result = await client.query({
query: `SELECT * FROM ${FLOW_PACKETS_TABLE} ORDER BY source_ts DESC, seq DESC LIMIT ${safeLimit}`,
format: "JSONEachRow"
});
const rows = await result.json<unknown[]>();
const records = rows
.map(normalizeFlowPacketRow)
.filter((record): record is FlowPacketRecord => record !== null);
const packets = records.map(fromFlowPacketRecord);
return FlowPacketSchema.array().parse(packets);
};
export const fetchOptionPrintsAfter = async (
client: ClickHouseClient,
afterTs: number,

View file

@ -0,0 +1,70 @@
import type { FlowPacket } from "@islandflow/types";
export const FLOW_PACKETS_TABLE = "flow_packets";
export type FlowPacketRecord = {
source_ts: number;
ingest_ts: number;
seq: number;
trace_id: string;
id: string;
members: string[];
features_json: string;
join_quality_json: string;
};
export const flowPacketsTableDDL = (): string => {
return `
CREATE TABLE IF NOT EXISTS ${FLOW_PACKETS_TABLE} (
source_ts UInt64,
ingest_ts UInt64,
seq UInt64,
trace_id String,
id String,
members Array(String),
features_json String,
join_quality_json String
)
ENGINE = MergeTree
ORDER BY (source_ts, seq)
`;
};
export const toFlowPacketRecord = (packet: FlowPacket): FlowPacketRecord => {
return {
source_ts: packet.source_ts,
ingest_ts: packet.ingest_ts,
seq: packet.seq,
trace_id: packet.trace_id,
id: packet.id,
members: packet.members,
features_json: JSON.stringify(packet.features),
join_quality_json: JSON.stringify(packet.join_quality)
};
};
const safeJson = (value: string, fallback: Record<string, unknown>): Record<string, unknown> => {
try {
const parsed = JSON.parse(value);
if (parsed && typeof parsed === "object") {
return parsed as Record<string, unknown>;
}
} catch {
// ignore
}
return fallback;
};
export const fromFlowPacketRecord = (record: FlowPacketRecord): FlowPacket => {
return {
source_ts: record.source_ts,
ingest_ts: record.ingest_ts,
seq: record.seq,
trace_id: record.trace_id,
id: record.id,
members: record.members,
features: safeJson(record.features_json, {}),
join_quality: safeJson(record.join_quality_json, {}) as Record<string, number>
};
};

View file

@ -1,3 +1,4 @@
export * from "./clickhouse";
export * from "./flow-packets";
export * from "./equity-prints";
export * from "./option-prints";

View file

@ -0,0 +1,39 @@
import { describe, expect, it } from "bun:test";
import {
flowPacketsTableDDL,
FLOW_PACKETS_TABLE,
fromFlowPacketRecord,
toFlowPacketRecord
} from "../src/flow-packets";
const packet = {
source_ts: 10,
ingest_ts: 20,
seq: 1,
trace_id: "fp-1",
id: "fp-1",
members: ["p1", "p2"],
features: {
option_contract_id: "SPY-2025-01-17-450-C",
count: 2,
total_size: 30
},
join_quality: {
nbbo_age_ms: 5
}
};
describe("flow-packets storage helpers", () => {
it("includes the correct table name in the DDL", () => {
const ddl = flowPacketsTableDDL();
expect(ddl).toContain(FLOW_PACKETS_TABLE);
expect(ddl).toContain("CREATE TABLE IF NOT EXISTS");
});
it("round-trips flow packet records", () => {
const record = toFlowPacketRecord(packet);
const restored = fromFlowPacketRecord(record);
expect(restored.features).toEqual(packet.features);
expect(restored.join_quality).toEqual(packet.join_quality);
});
});