> ## Documentation Index > Fetch the complete documentation index at: https://docs.lerian.studio/llms.txt > Use this file to discover all available pages before exploring further. # List DICT poll feed states > Reports which DICT poll feeds (claim discovery, MED infraction/refund discovery, the fraud-marker inventory sweep, funds-recovery event notifications) are frozen on a durable blocked marker, with the closed-vocabulary reason and the instant the freeze started. It also reports the funds-recovery feed's replay list: BACEN records it permanently stepped over and never stored, named by BACEN resource id, which are replayed through POST /api/v1/dict/funds-recoveries/intake. It also names the records a feed is currently stuck on and the ones a named operator has permanently abandoned: a held record listed here is the id an operator supplies to POST /api/v1/dict/discovery-feeds/{feed}/abandon-record, which is how a walk pinned on a record this deployment can never store is made to move on. A frozen feed deliberately does NOT degrade readiness — this rail also serves Pix keys, entries and claims, and the routes an operator needs to resolve a freeze are served by the same replicas — so this route, the brspi_dict_feed_blocked gauge and the operational event are the alarm. ## OpenAPI ````yaml /es/openapi/v3-current/spi-dict.yaml get /api/v1/dict/discovery-feeds openapi: 3.1.0 info: contact: email: contact@lerian.studio name: Lerian Studio url: https://lerian.studio description: >- OpenAPI 3.1 surface for the DICT (Diretório de Identificadores de Contas Transacionais) capability of Lerian SPI, the direct integration between the institution and the Brazilian Instant Payment System (SPI/Pix). It covers the PIX key lifecycle, portability and ownership claims, the MED 2.0 dispute surface (infraction reports, refunds, fraud markers, and funds recoveries), antifraud statistics, synchronous DICT reports, and the internal operations that keep the key directory consistent. license: name: Lerian Studio General License title: Lerian SPI — DICT API version: 1.0.0 servers: - url: https://spi.sandbox.lerian.net security: [] tags: - description: 'PIX key lifecycle: register, list, search, lookup, delete, and statistics.' name: Keys - description: >- PIX key portability and ownership claims through their full lifecycle (initiate, confirm, reject, cancel, acknowledge, complete). name: Claims - description: 'DICT infraction reports: list and retrieve.' name: Infractions - description: 'DICT refund requests: create, list, and retrieve.' name: Refunds - description: 'PIX fraud markers: create, list, and statistics.' name: Fraud Markers - description: >- Durable MED operation-intent status: the operationId every create/lifecycle-transition backlog route returns or references on every outcome, queryable independently of the business resource. name: Operations - description: >- Synchronous reports over Lerian SPI's own DICT record: COUNT(*) summaries of keys (by type/status) and claims (by type/status + open-past-deadline). COUNT-only — DICT holds no money. name: Reports - description: >- Internal DICT operations: reconciliation runs, claim deadline processing, and orphaned-key cleanup. name: Internal paths: /api/v1/dict/discovery-feeds: get: tags: - Operations summary: List DICT poll feed states description: >- Reports which DICT poll feeds (claim discovery, MED infraction/refund discovery, the fraud-marker inventory sweep, funds-recovery event notifications) are frozen on a durable blocked marker, with the closed-vocabulary reason and the instant the freeze started. It also reports the funds-recovery feed's replay list: BACEN records it permanently stepped over and never stored, named by BACEN resource id, which are replayed through POST /api/v1/dict/funds-recoveries/intake. It also names the records a feed is currently stuck on and the ones a named operator has permanently abandoned: a held record listed here is the id an operator supplies to POST /api/v1/dict/discovery-feeds/{feed}/abandon-record, which is how a walk pinned on a record this deployment can never store is made to move on. A frozen feed deliberately does NOT degrade readiness — this rail also serves Pix keys, entries and claims, and the routes an operator needs to resolve a freeze are served by the same replicas — so this route, the brspi_dict_feed_blocked gauge and the operational event are the alarm. operationId: listDICTDiscoveryFeeds responses: '200': content: application/json: schema: $ref: '#/components/schemas/DICTDiscoveryFeedListResponse' description: State of every DICT poll feed that can freeze. description: OK '500': content: application/problem+json: schema: $ref: '#/components/schemas/Detail' description: Internal Server Error default: content: application/problem+json: schema: $ref: '#/components/schemas/Detail' description: Error security: - BearerAuth: [] components: schemas: DICTDiscoveryFeedListResponse: additionalProperties: false properties: abandonedCount: description: >- How many regulated records named operators have abandoned across all listed feeds. Permanent. examples: - 0 format: int64 type: integer blockedCount: description: How many of the listed feeds are currently frozen. examples: - 0 format: int64 type: integer heldCount: description: >- How many of the listed feeds are running but not covering everything — pinned on a record this deployment permanently refuses, on an entry whose BACEN resource id could not be read, or (the fraud-marker sweep) on a window left edge that stopped advancing. Each one names what it is stuck on in heldBacenRecordIds. examples: - 0 format: int64 type: integer items: description: Every DICT poll feed that can freeze, blocked or not. items: $ref: '#/components/schemas/DICTDiscoveryFeedResponse' type: array steppedOverCount: description: >- How many of the listed feeds carry BACEN records they permanently stepped over and never stored, awaiting replay. examples: - 0 format: int64 type: integer required: - items - blockedCount - heldCount - steppedOverCount - abandonedCount type: object Detail: additionalProperties: false properties: code: description: >- Stable, machine-readable domain error code scoped to the emitting service (format: -NNNN). type: string correlationId: description: Request-scoped correlation identifier echoing X-Request-ID. examples: - req-7a3f9c2e type: string detail: description: >- A human-readable explanation specific to this occurrence of the problem. examples: - Property foo is required but is missing. type: string errors: description: Optional list of individual error details items: $ref: '#/components/schemas/ErrorDetail' type: - array - 'null' instance: description: >- A URI reference that identifies the specific occurrence of the problem. examples: - https://example.com/error-log/abc123 format: uri type: string status: description: HTTP status code examples: - 400 format: int64 type: integer title: description: >- A short, human-readable summary of the problem type. This value should not change between occurrences of the error. examples: - Bad Request type: string type: default: about:blank description: A URI reference to human-readable documentation for the error. examples: - https://example.com/errors/example format: uri type: string upstream: $ref: '#/components/schemas/Upstream' description: >- RFC 9457 extension member: the error a proxied third-party provider reported. Absent unless the emitting service explicitly surfaced one. required: - correlationId type: object DICTDiscoveryFeedResponse: additionalProperties: false properties: abandonedBacenRecordIds: description: >- The abandoned records, oldest first. Permanent: nothing in this service removes an entry, so an abandonment stays visible long after the log line that recorded it is gone. Who abandoned each one, and why, is in the audit trail. items: type: string type: - array - 'null' abandonedRecords: description: >- How many regulated records a named operator has permanently abandoned on this feed. The walk moves past exactly these records and no others; this deployment will never hold them. examples: - 1 format: int64 type: integer abandonedRecordsLimit: description: >- How many records may ever be written off on one feed. Fixed, and deliberately not raised to make a refusal go away: reaching it means a human personally gave up on this many regulated records on this one feed, which is a systemic refusal rather than that many accidents. examples: - 500 format: int64 type: integer abandonedRecordsRemaining: description: >- How many more records may be written off on this feed before POST /api/v1/dict/discovery-feeds/{feed}/abandon-record refuses. Zero means a record this feed is held on has no in-band ending left: it stays held and stays listed, the feed stays pinned on it, and the way out is to resolve what makes those records unstorable or to raise them with BACEN. The gauge brspi_dict_feed_abandoned_records_remaining carries the same number for alerting. It is a budget for the write-off, not a promise that this feed's current hold can use one: a feed held under window_left_edge_held or participant_not_served lists no record, so there is nothing to spend it on however many remain. Always 500 on dict_funds_recovery_events, which steps records over instead of holding them and has no write-off at all. examples: - 500 format: int64 type: integer abandonedSince: description: >- When the abandoned list last grew (RFC 3339, UTC); absent when it is empty. examples: - '2026-08-14T09:00:00Z' type: string blocked: description: >- True while the feed is frozen on its durable blocked marker. A frozen feed discovers nothing and does not clear on restart; readiness deliberately stays up so this deployment's operator routes remain reachable. examples: - false type: boolean blockedSince: description: >- When the freeze started (RFC 3339, UTC); absent when the feed is running. examples: - '2026-08-13T12:00:00Z' type: string feed: description: >- Feed name — the same string as the feed's blocked gauge, refusal counters and operational events, so every operator surface names a feed identically. It matches the feed's /readyz check name for the MED discovery and fraud-marker feeds; the two claim feeds share one readiness leg (dict_claim_feed_pagination) and so do not. examples: - dict_infraction_discovery type: string held: description: >- True while the feed's window is pinned and it is not covering everything: it sits on a record this deployment permanently refuses (record_quarantined), or on an entry whose BACEN resource id could not be read at all (malformed_entry_unreadable), or the fraud-marker sweep closed a cycle without covering it so its window's left edge stays where it is (window_left_edge_held), or BACEN served entries no participant this deployment serves is party to (participant_not_served), or the donor's receipt of an OPEN claim could not be sent and the window is pinned until it lands (claim_receipt_deferred), or that receipt is refused by a judicial block on the donor key and stays owed until the court order lifts (claim_receipt_blocked) — the one hold that does NOT pin the window, because an ofício lasts weeks and the receipt is written down and re-attempted every tick instead, so the feed goes on discovering while it stands. Distinct from blocked: a hold on its own freezes nothing, and the unblock route does not apply to it because there is no durable blocked marker to clear. A hold that survives the tolerated ticks becomes a block, and then both are true. Whether an operator has anything to abandon depends on WHICH hold it is, so read heldReason before reaching for the write-off: record_quarantined and malformed_entry_unreadable name their records in heldBacenRecordIds and the write-off is their sanctioned ending; claim_receipt_blocked names its claim too, but a court order lifts on its own and the tick after the lift sends the receipt, so the write-off is a last resort there, not the ending to reach for; window_left_edge_held, participant_not_served and claim_receipt_deferred name none and the write-off does not apply to them at all. examples: - false type: boolean heldBacenRecordIds: description: >- The records this feed is currently stuck on, oldest first: a BACEN resource id this deployment permanently refuses, the raw identifier BACEN sent for an entry whose id is not a UUID, or an 'unreadable-' handle standing in for an identifier too long or too strange to list verbatim. Anything listed here can be abandoned with POST /api/v1/dict/discovery-feeds/{feed}/abandon-record, and nothing else can — but read heldReason first: under claim_receipt_blocked the listed claim is one a judicial block postponed, its receipt is re-attempted on every tick without holding the feed up, and it is receivable again the day the ofício lifts with no operator action at all, so writing it off there discards a claim this deployment would otherwise have kept. Three holds list nothing at all and have no write-off: claim_receipt_deferred, whose next tick retries the receipt for free; window_left_edge_held, which is stuck on a watched document that is kept off this list and whose only lever is resetModifiedAfter on the unblock route once the hold has escalated to a block; and participant_not_served, which clears by itself as this deployment's participant registry catches up with BACEN and must never be written off. Otherwise capped per feed, so it can be shorter than heldEntries. items: type: string type: - array - 'null' heldEntries: description: >- How many entries the feed is stuck on. For a record this deployment permanently refuses it is how many such records are named in heldBacenRecordIds. For an entry whose BACEN resource id could not be read it is how many the feed could not read — the last tick, for the walks that stop at the first such page; the current CYCLE, for the fraud-marker sweep, whose cycle spans several sweeps of the book: it rises as those sweeps land and settles when the cycle closes. A document a run bound cut mid-walk counts on the sweep that finishes it, never on both. A book meeting the same unreadable entry cycle after cycle keeps reporting that one entry, and heldSince keeps answering how long it has been that way. Or, for that sweep's window hold, how many documents its last dirty cycle could not cover. examples: - 1 format: int64 type: integer heldReason: description: Closed-vocabulary reason the feed is pinned; absent when it is not. examples: - malformed_entry_unreadable type: string heldSince: description: >- When the held count last CHANGED (RFC 3339, UTC) — a tick that finds the same count does not re-stamp it, so this reads as how long the feed has been stuck on this many entries. Absent when the window is not pinned. examples: - '2026-08-13T12:00:00Z' type: string reason: description: >- Closed-vocabulary reason the feed froze; absent when the feed is running. examples: - inclusive_modified_after_no_progress type: string steppedOverBacenResourceIds: description: >- The BACEN resource ids awaiting replay, oldest first. Replay them with POST /api/v1/dict/funds-recoveries/intake, which runs the identical pipeline; attesting on that request clears both this list and the block it raised. items: type: string type: - array - 'null' steppedOverEntries: description: >- How many BACEN records this feed permanently stepped over and never stored, awaiting replay. examples: - 2 format: int64 type: integer steppedOverSince: description: >- When the replay list last grew (RFC 3339, UTC); absent when it is empty. examples: - '2026-08-13T12:00:00Z' type: string steppedOverUnlisted: description: >- How many stepped-over records this feed cannot name — BACEN's entity id was not a resource id, or the replay list was full when they arrived. They are missing exactly as the listed ones are, but cannot be replayed by id; raise them with BACEN. The intake attestation clears this count together with the list. examples: - 1 format: int64 type: integer required: - feed - blocked - held - abandonedRecordsLimit - abandonedRecordsRemaining type: object ErrorDetail: additionalProperties: false properties: location: description: >- Where the error occurred, e.g. 'body.items[3].tags' or 'path.thing-id' type: string message: description: Error message text type: string value: description: The value at the given location type: object Upstream: additionalProperties: false properties: code: description: The upstream provider's own error code, verbatim. examples: - E4001 type: string message: description: >- The upstream provider's own error message, verbatim (bounded, never its raw response body). examples: - account not found at provider type: string type: object securitySchemes: BearerAuth: bearerFormat: JWT description: JWT bearer token issued by the identity provider. scheme: bearer type: http ````