LikeFolio Data
LikeFolio Data · API docs

Errors & limits v1

Errors are small JSON bodies with a stable message and, for a bad parameter, the parameter's name. Limits exist to stop runaway scripts, not to meter normal use.

Error body

{ "error": "unknown correction category", "field": "correct" }

error is always present; field names the offending parameter when there is one. An unmapped ticker's 404 also carries ticker, kdate and ticker_history (each company that used the ticker, and when). A rare unhandled server error can be plain text.

Status codes

CodeMeaningExamples
400A parameter is invalid, missing or not supported on this route.start must be on or before end (field start, HTTP 400), invalid format (field format), invalid rollup (field rollup), invalid cursor (field cursor), unknown correction category (field correct), unknown metric, expected YYYY-MM-DD, window must be one of [7, 28], ticker is required, a cursor with parquet or arrow, cursor does not match this query and publication; restart pagination, a time-travel parameter on a route that serves the latest only (parameter is not supported by the deployed snapshot API), invalid history_mode, kdate on or after the first vintage requires recorded selection (Google search), kdate=all on Google search (expected YYYY-MM-DD).
401No key, or a key we do not know.missing or unknown key
403The key is revoked, or the route or file is not for your tier.key revoked (also a rotated key once its 10 minutes are up), a subscription-only route called with a trial key, key is not a drop for this tier.
404Nothing there for you.not covered: TICKER (not in coverage, or a search company below the 25% floor or with no verified brand), dataset not published, no such endpoint (also a wrong HTTP method), an identifier /v1/resolve cannot map, search_series_held at any kdate read from a snapshot that holds the company, not covered at kdate: TICKER (Google search; also a trend before its first date), no recorded vintage exists before the first capture; use reconstructed history.
409An identifier maps to more than one company.Ambiguous identifier mapping.
416The dates cannot be served.requested range is entirely inside the embargo window (trial), a future kdate, a kdate before the company's ownership coverage, kdate precedes search coverage (Google search: before 2020-09-01 on company routes, before 2015-12-15 on trend routes), a range after a company's series ended.
429Over a limit. Wait Retry-After seconds.rate limited, too many concurrent bulk jobs.
503Temporarily unavailable. Retry with backoff.key service unavailable (your key could not be checked, so the request was refused), a company's data unavailable, parquet/arrow unavailable (JSON and CSV still answer), /v1/quality without a current passing check, a Google search snapshot that fails its check at a kdate (GT vintage object integrity check failed and similar; no fallback).

Telling "not covered" from "wrong date"

On an identifier 404 from the company routes, the x-identifier-status header says which: not_in_universe (no covered company has ever used the identifier) or not_mapped_at_kdate (a covered company used it, but not at that kdate; the body's ticker_history says when). A ticker resolves only inside its dated span; use company_id for history. A Google search kdate read uses a third value, not_covered_at_kdate: the company or trend is not in the selected snapshot, or the trend has not started by that kdate.

Limits

LimitValueOver it
Burst guardA bucket of 300 requests per key, refilled at 50 requests a second. Trial and subscription alike.429 {"error": "rate limited"}, Retry-After in whole seconds (at least 1).
Bulk jobsAt most 2 running at once per key, and 4 overall, of which trial keys together hold at most 2. A bulk job is a /v1/bulk/download or a format=parquet/arrow response, counted until its body has been sent.429 {"error": "too many concurrent bulk jobs"}, Retry-After: 30.
Page sizeUp to 50,000 rows per JSON or CSV page.Follow next_cursor / x-next-cursor.

There is no per-minute or daily quota and no X-RateLimit-* headers. requests_total in /v1/status is a lifetime counter, not a quota.

Retrying

import time, requests

def get(session, url, **params):
    """GET with the API's retry rules: honour Retry-After on 429, back off on 503."""
    for attempt in range(6):
        r = session.get(url, params=params, timeout=120)
        if r.status_code == 429:
            time.sleep(int(r.headers.get("Retry-After", "1")))
            continue
        if r.status_code == 503:
            time.sleep(min(2 ** attempt, 30))
            continue
        r.raise_for_status()
        return r
    r.raise_for_status()

Do not retry 400, 401, 403, 404 or 416: they will not change on their own.

Company-local search hold (404). search_series_held means no complete approved active basket within the bounded recent window. The response has complete_through=null and points to historical_route; other companies remain available. It is distinct from an unknown or unpublished series. See held-series fields.

Hold versus technical failure. Approved inactive-current 404 responses use company_registry_ended and the approved company_registry_end, preserve the history route, and omit missing-member names. Invalid READY metadata (including an endpoint outside the approved closed-day freshness bound) returns 503; it must not silently serve stale data.

Publication and pagination recovery

Company-search cursor mismatches return HTTP 400 with field=cursor. Keep the endpoint, query parameters, format and key tier unchanged across pages; current and explicit-history routes cannot exchange cursors. After a publication or embargo-cutoff change, discard the unfinished pagination result and restart without a cursor. Cursors are opaque; never edit or synthesize them.

A short GT publication swap may return HTTP 503. Retry the whole request with backoff; the API does not silently substitute historical or partially published data. This guard preserves transport consistency, not a new runtime freshness rule or final data-quality clearance.