T3.Board (T3 v0.3.0)

Copy Markdown View Source

Pure board projection for ordinary orchestration shell/thread streams.

Fold chunk arrays and catch-up results with apply/3. The returned map is keyed by {host, thread_id}, so IDs from different environments cannot collide. A full shell snapshot replaces only that host's rows. Project-enrichment-only snapshots do not clear rows. Sparse thread events preserve fields they do not carry. Unknown event types are ignored.

rows = T3.Board.apply(%{}, chunk_values, host: "local")
rows = T3.Board.apply(rows, next_chunk_values, host: "local")

normalize/2 exposes individual upserts/removals for callers keeping their own cache. Its upserts can be partial (nil fields mean unavailable); use apply/3 when retaining prior fields and authoritative snapshot replacement matters.

Inputs must use ordinary, uncompressed projection bodies. Do not opt into acceptCompactIds, acceptDedupedBodies, acceptCompactTurnItems, or deferred work without restoring T3's projection format first. Timestamps and statuses are wire strings. Snippets collapse whitespace and default to 160 graphemes.

Summary

Types

key()

Functions

apply(rows, values, opts \\ [])

Fold a chunk, shell snapshot, or thread projection into a board cache.

normalize(values, opts \\ [])

Return {:upsert, row} / {:delete, thread_id} operations for wire items.

Types

key()

@type key() :: {term(), String.t()}

Functions

apply(rows, values, opts \\ [])

@spec apply(%{optional(key()) => T3.Board.Thread.t()}, map() | [map()], keyword()) ::
  map()

Fold a chunk, shell snapshot, or thread projection into a board cache.

normalize(values, opts \\ [])

Return {:upsert, row} / {:delete, thread_id} operations for wire items.