Field manual

How an agent should work

Follow SKILL.md, then open only the reference the decision tree names. Prefer bundled scripts. Summarize DataFrames; never paste a 2,000-row pandas repr.

Download history

import pytse_client as tse

frames = tse.download(symbols=["فولاد", "وبملت"], adjust=True, include_jdate=True)
print(frames["فولاد"].tail())

Ticker board

ticker = tse.Ticker("نوری")
print(ticker.title, ticker.eps, ticker.p_e_ratio, ticker.float_shares)
rt = ticker.get_ticker_real_time_info_response()
for bid in rt.buy_orders:
    print(bid.volume, bid.count, bid.price)

Index

idx = tse.FinancialIndex("شاخص کل")
print(idx.last_value, idx.last_update, idx.history.tail())

Intraday

from datetime import date
df = tse.get_trade_details("اهرم", date(2023,3,19), date(2023,4,22),
                           aggregate=True, timeframe="1m")

HTTP for hosts

GET /api/skill

Catalog: name, description, files, tools

GET /api/skill/md

Reconstructed SKILL.md

GET /api/tools

JSON Schema for every tool

POST /api/tools/invoke

{ tool, input, agent }

GET /api/mcp/tools

MCP-shaped list

POST /api/mcp/call

{ name, arguments }

GET /api/tickers

Studio universe

GET /api/health

Postgres ping

All HTTP data endpoints serve the synthetic studio corpus (demo: true) — never live TSETMC.

References

SKILL.mdYAML frontmatter + activation rules, decision tree, hard constraints.
references/ticker-api.mdEvery Ticker property, realtime board, shareholders, float, trade ticks.
references/download-api.mdBatch history, adjust=True algorithm, CSV layout, jdate.
references/orderbook-and-trades.mdget_orderbook, get_trade_details, timeframes, diff_orderbook.
references/financial-indexes.mdشاخص کل and sector indexes, FinancialIndex class, intraday.
references/filter-stats.mdget_stats, market watch, flow codes, YVal instrument types.
references/pitfalls.mdIP bans, asyncio, Arabic/Persian letters, halted symbols, missing EPS.
references/symbols-and-markets.mdFlow codes, ISIN, delisted index override, all_symbols().
references/agent-protocol.mdJSON tool schemas, HTTP invoke contract, progressive disclosure.
references/data-dictionary.mdExact columns, dtypes, and CSV/cache conventions for every DataFrame pytse-client returns.

Units & calendar

Prices and values are Iranian Rials. Toman = Rial / 10.
Cash session 09:00–12:30 Tehran, Saturday–Wednesday.
Export dates are Gregorian; pass include_jdate=True for Jalali.
adjust=True before any return calculation.
16 tools cover the public pytse-client surface.