quantskills/agent-quantspace/ingest

PandaData Ingest

Use this skill when a task needs stock, listed-fund/ETF, index, or futures market data; reference data; adjustment factors; or futures tick data from the PandaData SDK.

Prerequisites

• Install optional SDK dependencies with uv sync --extra panda_data.
• Set PANDA_DATA_USERNAME and PANDA_DATA_PASSWORD in the environment.
• PandaDataClient fetches data only. Persist normalized OHLCV with

skills.store.data_manager.DataManager.

Public API

from skills.ingest import PandaDataClient
from skills.ingest import to_panda_data_symbol, to_quantspace_symbol

PandaDataClient accepts QuantSpace symbols such as SHSE.510300 and panda_data native symbols such as 510300.SH. Returned symbol columns are converted back to QuantSpace format by default.

Wrapped Endpoints

Bars

• fetch_market_data(symbol, start_date, end_date, type="stock")
• fetch_market_min_data(symbol, start_date, end_date, symbol_type="stock", frequency="1m")
• fetch_hk_daily(symbol, start_date, end_date)
• fetch_us_daily(symbol, start_date, end_date)
• get_fund_daily(start_date, end_date, symbol=...)
• get_fund_daily_pre(start_date, end_date, symbol=...)
• get_fund_daily_post(start_date, end_date, symbol=...)

Reference

• get_stock_detail
• get_fund_detail
• get_index_detail
• get_index_indicator
• get_index_weights
• get_industry_detail
• get_industry_constituents
• get_stock_industry
• get_concept_list
• get_concept_constituents
• get_adj_factor

ETF Creation/Redemption

• get_fund_etf_cr_limits
• get_fund_etf_cr_net
• get_fund_etf_constituents
• get_fund_etf_cr

Listed ETFs and LOFs are funds, not A-share stocks. Fetch their bars with get_fund_daily*; do not send them to get_stock_daily, get_factor, or fetch_market_data(..., type="stock"). The get_fund_etf_* methods provide creation/redemption data and are not price-bar replacements.

The three get_fund_daily* methods accept an inclusive YYYYMMDD range of any length. PandaDataClient automatically sends contiguous requests of at most 365 calendar days and concatenates the returned frames in request order. This handling applies only to listed-fund daily bars, not get_fund_etf_* creation/redemption endpoints.

Futures Tick Utility

skills.ingest.panda_future_tick contains offline-testable helpers and CLI building blocks for PandaData futures tick downloads.

Progressive References

Detailed PandaAI docs are split by task under references/. Open only the specific file needed for the endpoint you are using.

Recipes

Daily A-share bars

from skills.ingest import PandaDataClient
client = PandaDataClient()
df = client.fetch_market_data("SHSE.600000", "20230101", "20231231", type="stock")

Daily ETF bars

from skills.ingest import PandaDataClient
client = PandaDataClient()
df = client.get_fund_daily(
    "20250610",
    "20250613",
    symbol="SHSE.510300",
    fields=["open", "high", "low", "close", "volume", "amount"],
)

Use get_fund_daily_pre or get_fund_daily_post when the research explicitly requires forward- or backward-adjusted fund prices.

The market-data API frequency remains the real bar interval. For example, adjusted daily bars still use freq="1d" (or the endpoint's daily API); do not send "1d_adj" to an ingest endpoint as a frequency. 1d_adj is only the QuantSpace storage directory/data-set name for adjusted 1d bars. The same rule applies to other intervals: a directory such as 5m_adj contains adjusted 5m bars, while the ingest frequency is still 5m.

Normalize and save bars

import pandas as pd
from skills.ingest import PandaDataClient
from skills.store.data_manager import DataManager
client = PandaDataClient()
raw = client.fetch_market_data("SHSE.600000", "20230101", "20231231", type="stock")
bars = raw.copy()
bars["eob"] = pd.to_datetime(bars["date"])
bars = bars.set_index("eob")[["open", "high", "low", "close", "volume"]].sort_index()
DataManager().save_symbol("SHSE.600000", bars, frequency="1d", source="panda_data")

When saving adjusted bars, pass the storage directory key, for example frequency="1d_adj", to DataManager; the fetch frequency stays 1d.

The same store boundary applies to listed funds: normalize the returned date column to the timezone-naive eob index, keep OHLCV columns, and call DataManager.save_symbol. PandaDataClient never writes local files.

Symbol conversion

from skills.ingest import to_panda_data_symbol, to_quantspace_symbol
assert to_panda_data_symbol("SHSE.510300") == "510300.SH"
assert to_quantspace_symbol("RB_DOMINANT.SHF") == "SHFE.RB99"

CLI command reference

<!-- BEGIN GENERATED: qs commands -->

Transport: --input-json FILE|-, --preview N, and --log-level LEVEL; flags override JSON. Use --print-template [FIELD] for fillable templates.

Shared parameter groups

OutputOptions

uv run qs ingest bars

Fetch stock, index, future, fund, HK or US bars; optionally store them.

Priority: P0. Side effects: network, writes-output, writes-store.

Shared parameters: OutputOptions.

Output data fields: asset, freq, adjust, start, end, requested_symbols, dry_run, saved, rows, columns, symbols, no_data, output, target_frequency, planned_paths, written_paths, preview.

Inspect schemas: uv run qs schema ingest.bars (add --part input|output or --file-field FIELD).

uv run qs ingest adj-factor

Fetch stock adjustment-factor events; optionally store them per symbol.

Priority: P1. Side effects: network, writes-output, writes-store.

Shared parameters: OutputOptions.

Output data fields: scope, requested_symbols, start, end, dry_run, saved, rows, columns, symbols, no_data, output, planned_paths, path_template, written_paths, preview.

Inspect schemas: uv run qs schema ingest.adj-factor (add --part input|output or --file-field FIELD).

uv run qs ingest ref stock-detail

Fetch stock basic information (get_stock_detail).

Priority: P1. Side effects: network, writes-output.

Shared parameters: OutputOptions.

Output data fields: endpoint, rows, columns, output, preview.

Inspect schemas: uv run qs schema ingest.ref.stock-detail (add --part input|output or --file-field FIELD).

uv run qs ingest ref fund-detail

Fetch fund and ETF metadata (get_fund_detail).

Priority: P1. Side effects: network, writes-output.

Shared parameters: OutputOptions.