sparklesfinance kit

Two host-native bundles · three exact market tracks

Give the model evidence, not a ticker and a guess.

Sparkles is an all-in-one collection of read-only market sources, typed calculations, research workflows, simulations, and audit receipts for CN, US, and HK equities—available through distinct Pi and DeepSeek Harness host lanes.

Open the manual →Choose a hostBrowse 135 plugins
Sparkles · finance context

track cn · SSE / SZSE / BSE

release Pi 0.1.13 · DSH 0.1.15

identity 600519 · XSHG · Main Board


◇ evidence envelope

source eastmoney

as_of 2026-08-11

freshness completed_daily

adjustment explicitly_selected

quality provider_observation

entitlement rights_unknown


✓ source receipt retained

✓ rule + calendar context retained

✓ no executable order surface


›

135T1–T6 proposals
3Isolated equity tracks
7Known runtime variables
0Order-mutation tools

Choose the host lane

Shared finance core. Native host surfaces.

The evidence laws, provider adapters, calculations, and receipts are shared. Installation, lifecycle, commands, and status UI follow each host's own contracts.

Pi lane

One extension entrypoint for the Pi TUI.

Tested with Pi 1.0.2. Pi 0.1.13 loads on that host.

Install the published @pi-sparkles/pi-sparkles@latest aggregate. Pi owns the extension lifecycle, slash commands, compact finance statusline, and responsive Unicode OHLCV charts rendered inline in the terminal.

Pi can also offer to remember your stock picks across chats. Accept through conversation, then choose any bounded reviews separately. Use the Pi Durable watchlist →

pi install npm:@pi-sparkles/pi-sparkles@latest
Compare installation and configuration →

Choose the controlling context

Start with the market, not the tool.

Each route keeps listing identity, venue rules, calendars, currencies, source receipts, and persisted state in its own track. Shared calculations are reused without erasing those differences.

CN↗

CN equity track

The most complete starting route: exact A-share identity, reviewed market calendars and rules, explicitly selected Eastmoney or supported Sina daily evidence, CNINFO/Tushare event inputs, research, planning, simulation, and journaling.

CNYAsia/Shanghai
US↗

US equity track

A primary-document route from exact US listing and venue context through SEC filings, exact XBRL facts, fundamentals, valuation, ownership, thesis, portfolio review, and point-in-time research.

USDAmerica/New_York
HK↗

HK equity track

A separately labelled HK route for listing context, HKEX calendar and rules, daily evidence, disclosures, board lots, ownership, Stock Connect facts, and A/H comparisons. It does not claim a duplicated end-to-end tier journey.

HKDAsia/Hong_Kong

Canonical product order

A research stack that grows with the question.

The tier order is a dependency order and a trader journey—not six interchangeable bundles. The current packages cover the complete T1–T6 inventory while keeping each host's lifecycle and presentation native.

T1

Observe & plan

Swing trader

Resolve the listing, establish market time and rules, acquire daily evidence, calculate technicals, and record a simulated plan.

T2

Research

Long-term investor

Move from primary filings and ownership evidence to fundamentals, comparisons, valuation, thesis, and a cited report.

T3

Monitor

Portfolio manager

Import holdings, measure risk and attribution, maintain watchlists, and audit alerts and state changes.

T4

Test

Quant researcher

Bind point-in-time data, calculate declared factors, replay events, compare trials, and reproduce backtests.

T5

Broaden

Multi-asset researcher

Add rates, FX, funds, fixed income, derivatives, commodities, crypto, and separately labelled global comparisons.

T6

Review the tape

Three-track day trader · execution review

Review separately labelled CN, HK, and US transaction-tape capability packets, apply shared sequence, gap, reset, correction, and bounded-stream laws, simulate named possible fills locally, evaluate compliance, create a non-executable handoff, and reconcile content-bound external receipts.

Read the six role journeys →Inspect current releases →

What enters model context

Structured facts with their limits still attached.

Sparkles turns provider and user-owned inputs into bounded context that the LLM can inspect and cite. The model receives provenance and unknowns alongside values in either host.

01

Validated identity

Track, MIC or venue, board, listing, share class, status, currency, timezone, and effective date—never just an ambiguous ticker.

02

Canonical observation

The source value plus as-of and retrieval time, freshness, unit, adjustment, quality, and entitlement. Unknown fields remain unknown.

03

Evidence and provenance

Content receipts, provider and document identifiers, source versions, hashes, correction lineage, gaps, duplicates, and conflicts.

04

Typed calculation

Inputs, formula or expression tree, period and filing coherence, output unit, rounding assumptions, and complete source leaves.

05

Decision record

A plan, scenario, thesis, replay, portfolio change, or journal entry with its assumptions and source receipts—not an executable instruction.

Read the context contract →

Hard product boundary

Research here. Trade elsewhere.

The plugin can help inspect evidence, calculate, plan, simulate, journal, and reconcile caller-owned receipts. It never becomes a broker console.

Safety manual →
  • No plugin can place, route, cancel, replace, or otherwise mutate a paper or live order.
  • Bid/offer interaction and brokerage order management stay outside the plugin boundary; users act through separate means.
  • Credentials are read at runtime from the caller environment and are never product files, defaults, logs, receipts, or persisted plugin state.
  • A provider must be selected explicitly. Failure never triggers silent fallback to another provider or market track.
  • Market observations and research are evidence, not investment, legal, accounting, or tax advice.