ObligacjeIO

By Apify

Overview

Research active Polish corporate bonds listed on the GPW regulated market (GPW RR). Search and filter series, compare issuer information, review cached source-document facts and summaries with page citations and evidence gaps, and open links to the relevant obligacje.io pages. Read Polish or English bond education, including retail Treasury bond concepts; Treasury instruments are outside the catalog. Sign in to obligacje.io and review the requested permissions before giving consent. Catalog, document, education and navigation tools are read-only. A focused question sent to the obligacje.io assistant creates a private conversation in your obligacje.io account and uses its existing assistant allowance. Send only the question and necessary context. Information and education, not investment advice. Answers may contain errors; check original sources. Missing cached evidence does not establish that a clause is absent. Contractual terms do not establish current covenant compliance, completed payments, creation of security, today's benchmark fixing or current issuer health. No live prices, personalized investment recommendations, trading or brokerage.

Tools

app_link

ChatGPT

Create a validated in-app link. Use route type bonds_query with query.filters, query.sort and query.dir to preserve EXACT bond_query/catalog_stats predicates. Copy all selected filters, including strict comparisons, issuer IDs, multiple alternatives and missing/present fields. Never reduce an advanced query to q alone. Other route types navigate to existing app pages.

ask_assistant

ChatGPT

Ask one focused bond question in Polish or English. Uses the account's obligacje.io assistant, creates a new private account conversation and consumes its existing assistant allowance. Supply only the question, locale and optional page context, never an external chat transcript. Answers are informational, not personalized investment advice. Long calls may exceed the host's tool timeout; do not automatically retry.

bond_get_many

ChatGPT

Get details for selected exact series/ISIN keys in one request. Choose fields; optionally include document metadata or includeDocumentaryTerms=true for independently reviewed documentary nominal and coupon formula with citations/freshness, separate from catalog observations. Null reviewed terms mean unavailable, stale or ambiguous review, not zero. Returns explicit missingKeys and paginated bounded results. Use a one-element keys list for a single known bond.

bond_query

ChatGPT

Search, filter and sort the COMPLETE visible bond catalog. Numeric predicates support eq/ne/gt/gte/lt/lte, coupon values are percentage points (5 = 5%). Alternatives in lists are OR; predicates and filter categories are AND. Exact issuerIds/keys, multiple coupon types/currencies/markets/kinds, dates and missing/present filters. totalMatches covers all matches. Follow nextCursor for further rows; changed snapshot/query rejects stale cursors. Choose fields to retrieve only needed groups. For global counts/means/grouping use catalog_stats.

catalog_stats

ChatGPT

Compute exact series and distinct-issuer counts, known/missing coverage, minimum/maximum/unweighted mean annual coupon over ALL matching visible bonds, independent of result limits. Coupon statistics are percentage points. Optional groupBy issuer/currency/couponType/maturityYear/market with paginated groups. Market memberships overlap; never sum overlapping groups as unique series. Unknown values are excluded from means and comparisons, not treated as zero. No debt/outstanding-value or investment ranking.

digest_issues

ChatGPT

Retrieve one shared list of open documentary questions, source conflicts, model limitations and pending/stale document reviews. Target key or keys (mutually exclusive, keys maximum10); omitted target uses the viewed bond. Current state/fixings/active margins/payment execution/covenant compliance are outside this workflow. Supported answers remain available beside model limits. Rejected is a legacy status, not proof of extraction error. Issues include exact unresolved and known typed source evidence plus current reviewed answers. An oversized evidence collection is returned in ordered evidencePart rows: collect the same (bondKey, issue.id) from partIndex1 through partCount, reconstruct each complete source pool (all nested sourcePart slices first), then concatenate distinct evidence. All other issue fields and knownAnswers repeat unchanged. Oversized individual assertions add sourcePart: collect all ordered slices for the same snapshot/bondKey/issue.id/evidenceIndex with identical assertion headers and knownAnswers. issue.sourceEvidence is a whole-source pool slice at zero-based sourceOffset; sourceIndexes address the full pool. Require contiguous offsets, every partIndex1..partCount and totalSourceCount before expansion, then append the assertion once. evidencePart counts these transport rows too; totalEvidenceCount counts distinct assertions. Missing, conflicting or incomplete parts are not complete evidence. Verify totalEvidenceCount before treating it as complete. Page counters count rows/parts; targetSummaries.openIssueCount counts distinct issues. When reviewContext.sourcesFromEvidence=true, its sources are identical to the evidence entry sources and represented there once; use that full list for both. If sourcePart is absent and issue.sourceEvidence is present, sourceIndexes arrays in evidence, reviewContext or legalIssue refer to that issue-local pool in their original order. Reconstruct those sources before applying sourcesFromEvidence. Exact quotes and metadata remain complete. No quotes or source metadata are truncated. includeKnownAnswers=false omits only additional full answer texts to reduce duplication; source evidence remains complete. Full fact resolutions are available through REST/CLI and document_fact_search. No new extraction. No issues is not automatic semantic certification or payload omission accounting. Copy nextCursor with unchanged arguments while hasMore=true; inventory/facts/review changes invalidate stale cursors.

document_context

ChatGPT

Retrieve COMPLETE cached topic matches as whole summary/keypoint/section chunks and typed facts. Oversized fact rows use the same ordered fact_evidence_part source protocol as document_fact_search; reconstruct all parts before treating the fact as authoritative. query is required. Supports one key, keys, issuerId, document/state/language filters and fact validation filters. Uncited summary is context only. Coverage and totalMatches apply before page limit; follow nextCursor. No new PDF extraction; no-match is not proof of no clause.

document_fact_search

ChatGPT

Search COMPLETE cached typed facts across one key, multiple keys or an exact issuerId. Use factType for one exact type or factTypes for related types in one request (mutually exclusive). For coupon formula prefer factTypes=[reference_rate,coupon_margin,coupon_type] without a broad query; dates use issue_date/maturity_date, denomination uses nominal_value. Broad query matches are ranked by typed relevance before incidental quote matches. Optional query and typed validationStatus; rejected facts excluded unless explicitly requested. Returns complete facts, page citations, provenance, gaps, selected extraction coverage and missingKeys. Oversized facts use kind=fact_evidence_part with evidencePart(partIndex,partCount,totalSourceCount,sourceOffset). Collect all ordered parts for the same snapshot/bondKey/factId/document/run before assigning authority. sourceEvidence slices form one exact fact-local pool; expand sourceIndex/sourceIndexes in fact, reviewContext and legalIssue after completion. No quotes are shortened or normalized. distinctFactMatches counts facts; page counters count rows/parts. When hasMore=true, copy nextCursor into cursor with unchanged target/filters; never repeat the same first-page request. A returned page is not the full search. Absence of cached evidence is not absence of a clause in the original document.

document_list

ChatGPT

List document metadata. For a bond provide targetType=bond and key (series or ISIN); for an issuer provide targetType=issuer and issuer_id. This does not retrieve document contents.

document_summary

ChatGPT

Retrieve cached digest summaries with extraction availability and whole-corpus coverage. Target one key, keys or exact issuerId; omitted target uses bond context. Filter documentKinds/languages/states and optional title query. Follow nextCursor; no new extraction. Coverage is document-to-series associations, not unique PDFs.

issuer_get

ChatGPT

Fetch one issuer, its listed bonds, document count and current published business profile. Use for what the company does. profile contains PL/EN paragraphs, issuer/group subject, sourceIds, source URLs/locations and review/publication dates. Null means missing or withdrawn; never infer business activity from model memory. Cite the sources and preserve issuer versus group distinctions. This is not financial-health or credit-risk evidence.

issuer_query

ChatGPT

Search and rank issuers having bonds matching filters. Counts describe matching scoped series, not debt value or all bonds of a company. Sort by name or activeBondCount (active matching series). totalMatches is the full issuer count, follow nextCursor for further rows.

knowledge_get

ChatGPT

Get actual published article text as whole paragraphs in the selected language. Follow nextCursor for remaining paragraphs. Educational curated content, not current market data.