Mental model

Provenance and traceability

The source object on every record – kinds, spans, trust, derivation, and traces.

Spectron’s accuracy story depends on one invariant: no fact-bearing record is anonymous. Provenance is a structured field, not an afterthought in application logs.

Every fact-bearing record (entity, attribute, relation, instruction, uncertainty, …) carries a source object:

FieldRole
source.kindturn, document, upsert, reflect, elaboration, consolidation, …
source.refOriginating turn, document, or trace identifier
source.session_id, source.turn_atWhen the record came from conversation
source.valid_from, source.span: { start, end }Valid-time anchor and quote position in the originating message or passage
source.locationOptional geometry for where the fact was captured
source.trustSource prior – admin documents rank above casual turns
source.derived_fromLineage for reflections, elaborations, consolidations

Some JSON examples in the docs show a flat source_turn field — that is the conversational shorthand for source.ref.

source.span stores character positions in the originating turn or document passage. Citations are stored data, not best-effort model prose. That powers “jump to quote” in UIs and defensible audit; tiered reads are described in Coherence, retrieval, and cost tiers.

Traces complete the story

Extraction and reconciliation emit decision_trace nodes. Ranked reads emit retrieval_trace; /chat and /reflect emit response_trace. Together they answer “which source produced this belief?” and “which retrieval path led to this answer?” Full detail: Traces and memory evolution.

Was this page helpful?