Repository navigation
Tools Reference
35 tools. providers is the tool's possible provider set, not a promise about any single call:
an individual response can degrade to fewer providers if some fail. The per-call truth is always that
response's data_provider plus its providers_answered.
Ask the server for this list at runtime: tapebasis_tool_map() returns it without any upstream
call.
Every provider that can serve the field is asked, in parallel and natively; overlapping fields are
compared. These can return ALL.
Quote plus fundamentals snapshot for any instrument class, equity, index, FX, commodity or crypto.
| Providers | TradingView scanner · Finviz quote page · Yahoo chart API · StockAnalysis quote API · Bloomberg quote page |
| Compared | price, change %, previous close, volume, session open/high/low, 52-week high/low, P/E, RSI(14) |
| Tolerances | price/prev-close/52-week/session range 0.5 % · change % 2 % · P/E 2 % · RSI 5 % · volume 1 % |
| Observation guards | price, change %, previous close, volume, session high/low/open: max_skew_seconds = 900 and same US/Eastern session date; change %, volume and the session range also carry an 8 h age ceiling |
| Declared delay | Bloomberg's 15-minute equity/index delay is itself a comparability boundary: a pair that far apart is not_comparable, with the delay named |
| Never compared | change % on FX, where no shared previous close exists (reported from every provider) |
Each compared field resolves to one value, the newest print, while every reading stays in
comparisons. Finviz-only material rides along, labelled per field in fields (each descriptor's
source is FINVIZ): the 84-field
fundamentals snapshot, company profile, peers, the after-hours print, the classification chips
(sector, industry, country, market_cap_class, exchange, etf_type) and the
income_breakdown island. resolutions records the ticker each provider was asked for.
day_open comes from TradingView, Yahoo and StockAnalysis: Finviz has no Open row and Bloomberg's
page carries one only sometimes, so the field is compared where both sides publish it rather than
against a provider that never had it. feed_mode (streaming, delayed_streaming_900) states the
scanner's own delay, the same fact Bloomberg's exchange_delay reports for its leg. pe_ratio has no
StockAnalysis getter: its quote API carries price, range, volume and the 52-week range, and P/E lives
on a separate stock page this tool does not fetch.
Symbol resolution only decides who can be asked. Once a symbol resolves, an empty provider response
remains a visible NO_DATA error; it is not reclassified as NOT_ATTEMPTED merely because the
symbol is absent from that provider's payload.
data.session (also at the envelope level) names the window data.price is a print from:
pre_market, regular, post_market or overnight, and data.price_field names the provider field it
was read from. Each provider's live print is resolved before comparison, because the obvious price field
goes stale the moment the regular session ends:
TradingView's close is the regular feed's last trade while its premarket_close keeps streaming,
Yahoo's regularMarketPrice freezes while fulldayPrice does not, and StockAnalysis keeps the regular
close in price and the current extended print in extended_hours. Reading the obvious field is what
let a quote sit hours behind the market and still be labelled current. The day change and the previous
close move with the price, so a row's three numbers describe one session rather than mixing a live
price with a baseline from two sessions back.
The session is taken from the winning provider's own print, never from another leg's label. A delayed
feed states the session it is in, and while it lags that statement describes a print the row is not
serving: measured at 09:42 ET, the scanner still said pre_market because its feed is 15 minutes
delayed, while the winning print was a regular-session one. Labelling from the scanner made
stock_quote and stock_prices name different sessions for the same symbol at the same moment.
The Finviz "Revenue & Expenses Breakdown" table and the product / region / segment revenue split ship
as a JSON island on the quote page stock_quote already fetched, so reading it costs no extra
request. It is not a separate tool: it is served inside stock_quote's payload as
income_breakdown, labelled FINVIZ per field.
statement holds up to two fiscal years of revenue, gross profit, operating income, net income and
the expense lines between them. Its amounts are published in millions and are scaled to absolute
currency units here; statement.scale_applied states the factor, and statement.currency the unit.
Bank filers swap the row set (interest income, loan loss provision, non-interest expense) and have no
Total Revenue row, which is served as it arrives.
revenue.axes holds up to three annual series: products_and_services, regions and segment. Which
exist varies by company, and their meaning is company-specific (Apple's segment is geographic,
NVIDIA's is operating segments), so the keys stay as Finviz sends them. Each axis carries unit and
converted; converted: true means Finviz FX-converted the amounts into the filing currency. An ETF
has no island at all, so the field is null and the quote is unaffected.
Because symbols are resolved per provider, XAUUSD reaches OANDA:XAUUSD on TradingView and
XAUUSD:CUR on Bloomberg, is correctly reported as not carried by Yahoo, Finviz and StockAnalysis,
and the response is ALL over the two that answered rather than NONE. Indices and commodities
(gold, silver, WTI, Brent, copper, natural gas, spot and front-month futures) are addressed through
the instrument registry in symbols.py, so WTI reaches TradingView as FX:USOIL, never as
NYSE:WTI (an equity whose ticker happens to match), and CL1! reaches Yahoo as CL=F and
Bloomberg as CL1:COM.
Bloomberg and StockAnalysis are asked in parallel like the others, so providers_answered can hold
every label and comparisons[].values every reading. Bloomberg's delayed print is bounded two ways:
it is last in the freshness tie-break, and a pair a delay length apart is reported as
not_comparable instead of as a disagreement. exchange_delay / print_time ride along in data,
so the delay is visible without reading the comparison detail. StockAnalysis covers US equities and
ETFs only, so it is NOT_ATTEMPTED for indices, FX, crypto, commodities and foreign listings; its
price surface is the same consolidated tape Yahoo serves, so that pair is weak corroboration. Its
extended-hours block rides along as stockanalysis_extended_hours.
Bulk quotes for a ticker list.
Each row carries its own data_provider, providers_answered and an agrees verdict, one ticker can
be ALL while another is single-source or NONE in the same response. agrees is true only when
at least one compared field agreed and none disagreed, false on a real conflict, and null when
both price and change percent reached no verdict. Multi-source rows also carry price_source (whose
print the single price came from), a fields map and their own comparisons and freshness
blocks.
Every row exposes misses and errors for its failed or uncarried providers, not just rows where
every provider came up empty. Error entries include provider, symbol, the original code and the
message; an omitted batch row is NO_DATA. A provider-wide empty response is also NO_DATA, never
NOT_ATTEMPTED. A returned payload without a usable price is also NO_DATA, never coverage.
The envelope aggregates the symbol-tagged failures, and its coverage and comparison notice are
derived from the actual provider-symbol outcomes. A complete ALL row does not mask an attempted
miss for another ticker: the envelope is PARTIAL while every provider supplying a row still
appears in providers_answered. The FX day change is reported but never compared, exactly as in
stock_quote.
Each row also names its session, the window its price is a print from, and the price_field it was
read from. The price itself is the live print rather than the obvious field: TradingView's scanner
close stops moving when the regular feed's last trade falls behind the market while premarket_close
keeps streaming, Yahoo's regularMarketPrice freezes while fulldayPrice does not, and StockAnalysis
keeps the current extended print in its own block. The row's change_percent and previous_close are
measured against the same session's baseline, so the row's own three numbers are consistent. Before
this, a pre-market price could be paired with a close two sessions old, implying a day change that
neither the change_percent beside it nor any provider reported.
fields states provenance per value, because data_provider describes the row rather than each
number in it:
A row can read ALL while previous_close was shaped by one provider. Only price and
change_percent are compared, so only those carry an agreement verdict and a compared list of every
provider that supplied a reading. The rest name the one provider whose value was used, chosen
freshest-print-first: an uncontested field comes from whichever provider has it and can date its
print, with the documented order breaking a tie on the same instant. A provider with no usable print
time cannot outrank one that has it. price_source remains as the row-level shorthand for the
provider the compared values were taken from.
Bloomberg is asked about at most four tickers per call (TAPEBASIS_BLOOMBERG_BATCH_MAX), because its
page path is paced at 10 s per request. The rest are listed in warnings and are not counted as
Bloomberg-answered for those rows: coverage stays truthful per row. StockAnalysis is asked about every
ticker: its pacing absorbs a short list, so it needs no cap, and a symbol it does not carry is
recorded as NOT_ATTEMPTED for that row.
US equity screener, cross-checked on the same instruments.
TradingView ranks the whole market server-side (sort_by ∈ market_cap, dividend_yield, change,
price). Finviz is then asked about exactly the tickers TradingView returned, and the price is
compared at 0.5 %. The optional finviz_filters expression is forwarded to Finviz to narrow this
selected universe; its usual filter validation applies. The option never replaces the TradingView
ranking with an independent Finviz universe.
Asking Finviz for its own independent ranking would compare two unrelated universes: a market-cap
ranking against a day-%-change ranking shared zero rows on every call, so the "multi-source" screen
was silently single-source. cross_checked_rows, agreeing_rows and not_comparable_rows summarise
the result.
Finviz screener rows carry no print time, so a per-row match cannot be proven to describe the same
session. Those rows are reported as not_comparable (not as agreement), and the envelope warns that
no row could be confirmed. Yahoo has no screener endpoint, so it is not asked.
S&P 500, Dow, Nasdaq Composite, Russell 2000, IWM and VIX levels, keyed by canonical name (SPX,
DJI, COMP, RUT, IWM, VIX).
| Providers | TradingView SP:SPX/DJ:DJI/NASDAQ:IXIC/TVC:RUT/TVC:VIX · Finviz indices feed · Yahoo ^GSPC/^DJI/^IXIC/^RUT/^VIX
|
| Compared | index level, 0.5 % (VIX 3 %, because a volatility gauge moves percent-level within a minute) |
| Not compared | day-change %, reported from every provider, derived from different reference closes |
Each level resolves by the same newest-print rule as a quote, and levels are only compared between providers whose prints share a session. A provider that has rolled into the next session is newer not wrong; its level is the one shown, and the session-date guard keeps yesterday's close from being reported as a data conflict.
Finviz quotes indices only through its indices feed, not a per-ticker page, /stock?t=SPX is a 404
and COMP there is Compass Inc, a real equity. So resolutions reports Finviz as not carrying an
index ticker, while market_indices still uses Finviz's indices feed for the level.
RUT (index) and IWM (its tracking ETF) are different instruments at different price levels and
are never cross-compared.
The list is fixed (INDEX_SNAPSHOT), and the registry being broader does not extend it: NDX is a
registry row addressable through single-instrument tools (stock_quote NDX), but it is not part of
the snapshot.
Each index carries its own fields map and data_provider. The day-quote block (level, previous
close, change, change percent) is filled from providers whose print belongs to the winner's session
date, so a row is always one observation: a provider from another session never lends a number. When
the winning provider's session has no previous close to report, the field is null rather than
borrowed from the other session, which is the behaviour that stopped a level and its previous close
contradicting each other on the page.
Ticker headlines, matched by title across the feeds.
Finviz supplies price reactions around each item; Yahoo supplies its own search feed. The TradingView
leg is the provider's licensed Marketaux feed and needs MARKETAUX_API_TOKEN; without it that leg is
NOT_ATTEMPTED and the response is labelled with the feeds that did answer.
Sentiment does not need the token, see market_sentiment.
Options chain for one expiry, from Finviz and Yahoo, with the underlying price confirmed by TradingView's scanner (TradingView publishes no chain).
Contracts are matched by side, expiry and strike; volume is compared at 5 %, open interest at 1 % when both report it. Finviz carries full greeks. A missing side, expiry or strike cannot establish a match and never counts as agreement. A contract-level expiry takes precedence over its chain-level expiry, which supplies the date only when the contract does not carry one.
With no explicit expiry, each provider chooses its nearest expiry independently. If those dates differ,
data.expiry is null and data.overlap carries not_comparable: true with both dates in its reason;
the contracts are not compared. When both chains establish the same date, including an explicitly
requested shared expiry, comparison proceeds normally. data.expiries names the providers separately
when their available-expiry lists differ.
Missing activity fields stay null and a reported zero stays zero in normalisation and comparison;
neither falls through to another field or becomes a synthetic zero.
Yahoo reports zero open interest for contracts expiring the same day, so a 0 vs N pair is counted
under open_interest_unavailable, not as a mismatch.
Earnings dates, reported-quarter history, estimate revisions and post-earnings price reactions.
Finviz owns the detailed tables; its earnings_date is the last reported timestamp, which is why
it is returned as finviz_earnings_date, separate from next_earnings_date (the upcoming release
from TradingView or Yahoo). Finviz never serves as a fallback for the next date: when both calendars
have no upcoming timestamp, were not attempted, or failed, next_earnings_date is null. A
historical report cannot be reported as a future event.
Pre-/post-market price from every independent source that publishes one: TradingView's
premarket_*/postmarket_* scanner columns, Finviz's after-hours print, Yahoo's 1-minute
extended-hours candles, and StockAnalysis's extended-hours block. The after-hours price is compared
at 0.5 %. Each provider's after_hours_readings entry carries price, session, timestamp, a
normalized UTC as_of_utc when parseable, reference_close, change, change_percent and
change_reference. A missing baseline or print time
stays null. Only like pre-market or post-market windows can be judged. Differing named sessions are
not_comparable, and agreement on a shared window requires a usable as_of_utc or price_timestamp
from both providers in that pair. An undated TradingView print therefore cannot veto a dated
Yahoo/StockAnalysis pair; it remains visible with match: null only in its own pairs. StockAnalysis's
pre/post label is retained even if it supplies no timestamp. Yahoo's as_of_utc is seconds-preserving,
timezone-aware ISO from the candle, not a string the freshness parser has to guess.
Each session block carries the session's volume when the feed reports one, so a pre-market print
can be told apart from a single lot. Yahoo's pre-market reference comes from reference_close, its
post-market print from the regular block alongside it. StockAnalysis reads its extended
previous_close, Finviz its snapshot's Prev Close before falling back to its own extended
change arithmetic, and TradingView recovers the close from its extended print and reported change.
A missing Finviz session or print clock stays null without suppressing that available baseline;
its pairs remain unjudged. These are provider-owned baselines, not an overnight Yahoo baseline
borrowed for a pre/post print. StockAnalysis's block is also served whole as
stockanalysis_extended_hours, next to stockanalysis_regular_close.
stock_quote does not itself return the independent BOATS overnight block. Request
stock_extended_hours when that session is needed; adding those separate upstream calls to every
quote remains a distinct contract expansion, not part of this metadata repair.
data.session names the window the instrument is in right now (pre_market, regular,
post_market, overnight), so a caller asking "pre or post" does not have to infer it from which block
happens to be populated. It comes from TradingView's own current_session statement, with the newest
Yahoo block as the fallback.
The overnight block is the 20:00-04:00 ET window that Blue Ocean ATS trades. It is read natively
through TradingView's BOATS: quote websocket and Yahoo's /v7/finance/quote request with
overnightPrice=true. TradingView's scanner has no BOATS row, and Yahoo's 1-minute chart stops at
after-hours, so neither is substituted for the venue's print. Yahoo uses its overnightMarketPrice,
overnightMarketChange, overnightMarketChangePercent and overnightMarketTime quote fields, with
regularMarketPrice as the overnight change reference. Yahoo publishes no overnight volume on that
header, so its volume is null rather than inferred from regular trading.
overnight_source names the provider whose newest eligible print is in overnight. Each provider block
stays in overnight_readings, including print_time, seconds-preserving ISO as_of_utc, the 20:00 ET
window start in session_window_start_utc, is_current, and venue identity when reported. On prints
sharing that evidenced occurrence, every block's change and change_percent are calculated from
Yahoo's reference_close; change_reference states YAHOO.reference_close. TradingView's original
BOATS ch and chp readings remain separate as reported_change and reported_change_percent, since
they use a different venue baseline. If Yahoo is absent, the reference is invalid, or the occurrences
cannot be shown to match, the normalized change and reference fields are null rather than borrowed from
another close. The reported TradingView fields are still retained.
overnight_comparison compares only prints from that same overnight occurrence with usable,
sufficiently close timestamps. These are two readers of one BOATS venue, not independent market venues.
overnight_errors reports a failed secondary
endpoint; those errors are appended to the envelope's errors, so a null overnight block is not
silently indistinguishable from venue no-data. The primary pre/post comparisons remain independent.
Every overnight block carries is_current and the aggregate has overnight_is_current. After 04:00 ET
the last BOATS print remains available as the last overnight observation with is_current: false,
but is not called the live print and cannot displace a pre-market price. TradingView's out_of_session
and streaming labels alone do not assert that the venue is still trading.
TradingView quote qsd frames are deltas. They are merged into the preceding symbol snapshot rather
than replacing it, so a later frame carrying only lp_time, volume or another field cannot delete a
price already received. A quiet empty socket is not cached, and receive or rate-limit failures keep
a typed provider error instead of silently becoming overnight: null.
Yahoo's candles are requested over two days, not one: range=1d stopped returning a timestamp
array, so every call reported NO_DATA with "unexpected response shape: 'timestamp'" and the session
timeline silently disappeared. Each candle is classified against the session windows of the day it
falls in, which keeps yesterday's regular prints out of today's pre-market block.
Full indicator stack from TradingView (RSI, MACD, EMAs, Bollinger, ATR, ADX, structure, trade setup).
The request uses the resolved TradingView symbol and venue, including an EXCHANGE:TICKER
resolution; the caller's default exchange cannot route a resolved ticker to the wrong venue. If a
resolved symbol has no venue and no exchange prefix, the exchange argument remains the fallback.
Finviz's snapshot and Yahoo's chart API independently confirm four headline readings in cross_check:
price (0.5 %), RSI(14) (12 %), ATR(14) (10 %), volume (2 %).
The three price-and-price-derived readings are guarded by print time: TradingView's scanner
update_time, Finviz's rendered print time and Yahoo's market_time are compared, so two prints from
different sessions are reported as not_comparable rather than as a conflict. RSI and ATR are
lookback statistics and are judged without a timestamp guard.
Volume is reported from every provider but never compared: Finviz's snapshot counts the current
partial session while the consolidated feeds report the last completed one, so the two numbers answer
different questions. That is stated in cross_check.volume.pairs[].not_comparable rather than
presented as a disagreement.
A single strategy, run twice, once on TradingView's chart-websocket OHLCV, once on Yahoo's chart OHLCV, then compared. Finviz is not asked: it has no history endpoint.
Compared: total return (1 %), win rate (2 %), trade count (exact), max drawdown (2 %), Sharpe (5 %), buy-and-hold return (1 %). Identical candles produce identical metrics; a divergence means the two feeds disagree about the instrument's history, which is worth knowing before trusting either.
For RSI, Bollinger, MACD, EMA cross, Supertrend, RSI pullback, Keltner breakout and triple EMA, a signal is evaluated only after its bar closes and the entry or exit fills at the following bar's open. A signal on the final close cannot open a new position. A final bar with no following execution bar still force-closes an open position at its close and flags that exit in the raw trade list. The optional per-trade log carries the same execution prices; per-bar equity marking includes the entry bar's close and applies net costs at exit.
Donchian retains its separate assumption: an intrabar high/low breakout fills at that bar's close.
This is not a close-confirmed signal model, but final-bar entries are suppressed there too. The
execution_model block names the mode, signal-to-fill rule, final-bar rule and forced-exit treatment
instead of requiring callers to infer them. It is present on single backtests, comparisons and
walk-forward results.
The full strategy leaderboard, also computed on both feeds. Adds winner_comparison, the winning
strategy on TradingView, the winner on Yahoo, and whether the two feeds pick the same one.
Global FX, crypto, index and ETF snapshot from TradingView's scanner and Yahoo's chart API, group by group. Finviz has no equivalent breadth, so it is not asked.
Rows join by the requested ticker, not by the symbol the upstream reports back: Yahoo answers both
JPY=X and USDJPY=X with meta.symbol: "JPY=X", which used to split one instrument into two rows.
Each row carries both providers' prints and an agree flag. all_groups_agree summarises the whole
snapshot.
For a cross-checked index read, use market_indices.
TradingView technical read + sentiment + news, with the underlying price cross-checked against Finviz's snapshot and Yahoo's chart API. One request, not one per leg.
The price leg is genuinely multi-source; the analysis and sentiment legs remain single-source and are
labelled per field in fields (their source is TRADINGVIEW). sentiment_source states whether
the sentiment came from Marketaux or
local scoring.
Served, labelled FINVIZ, explicitly single-source. No other provider has a comparable dataset, so
those legs are marked NOT_ATTEMPTED, which is not an error and is excluded from errors.
| Tool | Capability | Notes |
|---|---|---|
stock_filings |
SEC filing history | Finviz's SEC feed |
stock_ownership |
Insider + institutional ownership, short interest | Several datasets; a failing one is reported without losing the others |
stock_peers |
Peer-group companies | No provider has a curated peer list |
stock_why_moving |
Narrative explainer for a price move | Finviz editorial feed |
stock_chart |
Rendered PNG candlestick chart | TradingView produces text analysis, not images |
stock_dilution |
Dilution risk model (share count, float, cash runway) | Finviz only models dilution-prone companies; others return NONE with that reason |
stock_screener_filters |
Finviz screener filter catalogue | Finviz's proprietary filter vocabulary |
The normal MCP JSON envelope contains data.image_base64, data.mime_type: "image/png",
data.bytes and chart dimensions. Decode image_base64 to obtain the PNG. A remote client does not
need filesystem access. data.path and the convenience field image_path carry the same
server-local path when a local copy can be written; that path is not a URL and is secondary to the
inline image. A write failure returns the inline image, a null path and a warning.
fv_access.chart calls chart_service.get_chart_base64, which makes one Finviz request and writes
the resulting bytes to local storage when possible. It does not call download_chart for a second
copy. At most 32 service-generated PNGs are retained, ordered by modification time. Unknown files in
the chart directory are never deleted. The existing download_chart path-only service remains
available for callers that explicitly do not want base64 in the response.
Two backends, selected at runtime, same payload shape:
| Condition | Backend | Label |
|---|---|---|
MARKETAUX_API_TOKEN set and usable |
Marketaux entity sentiment (licensed) | TRADINGVIEW |
| no token, or Marketaux rejected/quota/outage | Finviz headlines scored locally (VADER) | FINVIZ |
| neither available | labelled empty payload, sentiment_score: null
|
NONE |
Both paths are single-source by nature: there is nothing to cross-check a sentiment score against, so the label names whichever feed supplied the headlines.
The local path is labelled FINVIZ because the headlines are Finviz's, only the scoring is
ours. Labelling it TRADINGVIEW would claim a provenance the data does not have.
sentiment_score: null (not 0.0) means "nothing could be scored". 0.0 means "the tone is
balanced". Conflating them would dress a data gap up as a finding. Full detail:
Sentiment.
Served, labelled TRADINGVIEW, explicitly single-source, no other provider has the dataset.
| Tool | Capability | Why nobody else can confirm it |
|---|---|---|
crypto_top_movers |
Crypto gainers/losers | Finviz does not cover crypto; Yahoo has no movers endpoint |
crypto_bollinger_scan |
Bollinger squeeze scan | Same |
crypto_volume_breakout |
Volume + price breakout scan | Same |
crypto_multi_timeframe |
Multi-timeframe crypto analysis | Same |
crypto_multi_agent |
Multi-agent bull/bear debate | Same |
futures_market_overview |
Futures overview by category | Neither other provider has futures |
futures_top_movers |
Futures gainers/losers | Same |
futures_watchlist |
Curated futures watchlist with levels | Same |
bitcoin_market_pulse |
Bitcoin trend and volatility pulse | Composite of the crypto scans above |
egx_market_overview |
Egyptian Exchange overview | No EGX coverage elsewhere |
Note that crypto and futures quotes are multi-source, stock_quote("BTCUSD") asks TradingView and
Yahoo. It is the scan/analysis endpoints that are TradingView-only.
These tools ask Bloomberg and nobody else, so their responses are single-source by construction:
data_provider: BLOOMBERG, cross_checked: false, reconciliation_leg: false. The provider itself
is a full leg of stock_quote / stock_prices, where it is asked in parallel and its readings
appear in comparisons. What these two tools add is Bloomberg's own quote page and its market board.
| Tool | Capability | Notes |
|---|---|---|
bloomberg_quote |
Bloomberg's own quote page for one instrument |
print_time from the last intraday point, never lastUpdate; exchange_delay states the delay |
bloomberg_markets |
Index levels, government benchmark yields, commodities, FX | The only source here of a government yield curve; the board has no per-instrument print time, so nothing is aged or ordered |
For a corroborated price use stock_quote, which asks Bloomberg too and compares it. Use this when
the caller wants Bloomberg's reading specifically, or a figure only Bloomberg publishes.
| Providers |
bloomberg.com/quote/<id> HTML, __NEXT_DATA__ payload |
| Compared | nothing: this tool asks one provider |
| Delay |
exchangeDelay (15 Minutes on equities), echoed as exchange_delay / is_exchange_delayed
|
| Print time |
priceMovements1Day[-1].dateTime, and only when that point's value matches the quoted price |
| Never used |
lastUpdate, a session sentinel (...T01:00:00Z) that is later than the real print |
print_time: null with print_time_missing_reason is the honest answer when the intraday series
ends somewhere other than the quoted price: a timestamp describing a different number is not a print
time. Symbols resolve through the instrument registry plus a curated alias table (SPX to SPX:IND,
EURUSD to EURUSD:CUR, BRK.B to BRK/B:US), because Bloomberg answers a bad id with a 403 bot
page rather than a 404.
An instrument the registry states Bloomberg does not carry is refused with that reason instead of
being guessed. WTI is a spot series with no Bloomberg quote page, and the mechanical fallback
WTI:US is W&T Offshore Inc, a different company; USOIL resolved to USOIL:US, which answers a
robot page. The response is NONE with a NO_DATA error naming the boundary, and no request is
made. The same identity function decides what the cross-checked tools report, so the id in
resolutions is always the id that was fetched.
The board carries 20 rows on one page: index levels, ten government benchmark yields, crude, gold and FX. The yield curve is the capability no other provider here publishes. There is no per-instrument print time on the board, so the response says so and no reading is aged or ordered. Treat every level as a snapshot, not a live tape.
| Tool | Purpose |
|---|---|
tapebasis_health |
Read the cached drift/outage verdict and per-provider rate-limit state, refresh=True to probe now, or accept_baseline=True to accept the shape being served |
tapebasis_tool_map |
Return this catalogue with no upstream call |
| You want | Use | Multi-source? |
|---|---|---|
| One stock's price and fundamentals | stock_quote |
Yes, on the quote fields |
| A crypto or FX quote | stock_quote |
Yes where both carry it (FX day change excepted) |
| Many stocks' prices | stock_prices |
Per row |
| A market scan | stock_screener |
On price, per row |
| Index levels | market_indices |
Yes |
| Headlines | stock_news |
Overlap only |
| How the market feels | market_sentiment |
No (single-source by nature) |
| Options positioning | stock_options_chain |
On volume/OI |
| Filings, ownership, peers |
stock_filings, stock_ownership, stock_peers
|
No |
| Government bond yields | bloomberg_markets |
No, Bloomberg-only |
| Bloomberg's own quote reading | bloomberg_quote |
No: one provider asked, delay-labelled |
| Crypto / futures scans |
crypto_*, futures_*
|
No |
| Strategy validation |
stock_backtest, stock_backtest_compare
|
Yes, two OHLCV feeds |
| Everything about one stock | combined_analysis |
Price leg only |
- Response Contract: how to read any of these
- Design Decisions: why some tools are single-source by nature
{ "price": { "source": "YAHOO", "compared": ["TRADINGVIEW", "FINVIZ", "YAHOO"], "agreement": true }, "change_percent": { "source": "YAHOO", "compared": ["TRADINGVIEW", "FINVIZ", "YAHOO"], "agreement": true }, "previous_close": { "source": "YAHOO", "compared": ["YAHOO"], "agreement": null }, "volume": { "source": "TRADINGVIEW", "compared": ["TRADINGVIEW"], "agreement": null } }