base_dialog.dbss

Binary layout, byte coverage, field semantics, and validation evidence for the base_dialog decoder used by calpheonlabs.

Investigation status

Status	Value
Coverage	100%
Iteration	2
Review state	No open blockers
Last reviewed	2026-07-20
Captures examined	installed client 2923, prior-build ledger
Decoder	`packages/scraper/src/tables/base_dialog.dbss.ts`
Remaining blockers	None

See Decoder methodology for the confidence scale and evidence rules.

base_dialogoffset

Open the companion pointer directory used to frame and validate this payload.

Field ledger

Field/path	Offset and width	Representation	Claimed meaning	Physical confidence	Representation confidence	Semantic confidence	Evidence	Status
`rows` + encoded DBSS count	file `+0` count, then variable rows	array retained in physical order; count omitted	base-dialogue row sequence	Certain	Certain	Certain (structural)	The count drives complete sequential decoding to EOF in both captures; client 2923 has 58,537 retained rows, and a `count + 1` mutation rejects	Closed
`dialogKey`	row `+0`, `u32`	retained losslessly	packed `(dialogId << 16) \| npcId` identity	Certain	Certain	Certain	Every key agrees with the companion and `detail_dialog`; 20,215 archived `(Npc, DialogIndex)` pairs overlap, and representative packed key `2,425,414` decomposes to `582/37`	Closed
`speakerLabel` length	row `+4` then variable per row, `u32` UTF-16 code-unit count	consumed to size the string; omitted	string framing	Certain	Certain	Certain (structural)	Independent raw scan reaches the next field exactly for every row	Closed
`speakerLabel` reserved word	after its length, `u32 == 0`	validated and omitted	reserved serializer word	Certain	Certain	Certain (reserved)	All applicable words are zero in both captures; a nonzero mutation rejects	Closed
`speakerLabel`	counted UTF-16LE code units	decoded and retained	speaker role or name	Certain	Certain	High	20,113/20,186 exact archived `Dialog_Table.Name` root matches; 73 recognizable historical/localization drifts; BDOCodex identifies representative Andro `582/37`	Closed
`lines` + encoded count	after `speakerLabel`, `u32` then variable strings	ordered array retained; count omitted	generic-line list and its physical length	Certain	Certain	Certain (structural)	The counted elements plus footer end at the independently addressed boundary for all 58,537 rows	Closed
`lines[]` length	each line `+0`, `u32` UTF-16 code-unit count	consumed and omitted	line-string framing	Certain	Certain	Certain (structural)	Independent raw scan covers every line without gaps or overrun	Closed
`lines[]` reserved word	each line `+4`, `u32 == 0`	validated and omitted	reserved serializer word	Certain	Certain	Certain (reserved)	Together with the speaker fields, all 68,702 current string-control words are zero; a line mutation rejects	Closed
`lines[]`	counted UTF-16LE code units	decoded and retained in order	generic dialogue text	Certain	Certain	High	20,131/20,186 exact archived `Dialog_Table.Bubble1..Bubble10` lists; 55 recognizable historical text drifts	Closed
row `footer`	final 5 bytes	required zeros; omitted	reserved row terminator	Certain	Certain	Certain (reserved)	Every row ends with five zeros at the companion-confirmed boundary; a targeted mutation rejects	Closed
derived `npcId`	no additional bytes; low 16 bits of `dialogKey`	retained alongside the physical key	NPC identifier	Certain	Certain	Certain	Exact bit identity for every row, cross-table equality, archived `Dialog_Table.Npc`, and BDOCodex `582/37`	Closed
derived `dialogId`	no additional bytes; high 16 bits of `dialogKey`	retained alongside the physical key	dialogue variant/index	Certain	Certain	Certain	Exact bit identity for every row, cross-table equality, archived `Dialog_Table.DialogIndex`, and BDOCodex `582/37`	Closed

Active assumptions

Findings

ID	Finding	Evidence	Schema/documentation impact	Status
BD-F01	The formerly skipped word in every string envelope is required zero framing.	An independent scanner examined all 68,702 current words; every value is zero, the prior capture has the same domain, and speaker/line violations reject.	`.skip(4)` was replaced by `reserved(4)`; no serialized byte is silently ignored.	Closed
BD-F02	Intrinsic row framing and the companion directory are exact, independent descriptions of the same data.	The 58,537 unique pointer spans partition `[4, 2,038,989)` without gaps/overlap; every pointer key equals its physical and decoded key. Key, offset, and length corruptions are detected by the companion audit.	Production decoding remains intrinsic; the companion is exhaustive verification, not a runtime dependency.	Closed
BD-F03	Archived dialogue data corroborates meaning but not current binary layout.	The archive has 27,424 rows/20,474 unique pairs; 20,215 pairs overlap the current 58,537, while 259 are archive-only and 38,322 current-only. It contains no offsets, string controls, footer bytes, or current-domain guarantee.	Exact archived columns and representative keys appear in TSDoc; no archive data is used by the decoder or tests.	Closed
BD-F04	Captured totals and string ranges are observations, not format invariants.	Client 2923 has 58,537 rows, 10,165 lines, and 2,038,989 bytes; observed rows have 0–10 lines, but counts are self-describing.	No literal row, line, size, or string-length cap appears in the schema.	Closed