Connect any AI agent, personal assistant, or coding tool via MCP or HTTPS
Connect →
TELEON ARCHITECTURE MASTER BLUEPRINT & SEED PORTAL
← All 320 Atomic Work Queues STAGE 167 OF 320 context_decomposition Async enhancement

0-Call Page Segmentation (aq167_segment_document_into_pages_0call)

Splits Store S42 documents at heading boundaries and bounded character caps into numbered Store S43 DocumentPage rows with exact UTF-8 byte offsets.

Numbered DocumentPagesExact Byte OffsetsPer-Page SimHash64
TIER 1 · DETERMINISTIC (0 MODEL CALLS)

0-Call Engine & Verification Slot

Heading-aware UTF-8 byte-offset page slicer & SimHash64 builder (0 calls)

Substrate: db_atomic · p99 Claim Target: ≤ 4 ms · Phase: 166 - 170 Document-Based Store, Segmented Numbered Pages, BM25 & SWHID Revisions

AI Agent (1..6)

Model-Guided Escalation & Synthesis Role

Semantic section boundary & table-preservation reviewer

5-Role Component Pod: 0-Call Slot / Solver + OpenSearch / Split-Supervisor (≤4k tok) / Diagnostic Repair + Twin Shim / CPOP + Sister Distiller.

Enforced Category State Machine: context_decomposition

Commit Invariant: Splits large epics into <=4k-token leaf leases with typed input/output contracts while keeping the parent vertex parked until leaf merge.

Architectural Problem Solved: Allows smaller, cheaper models to execute large overnight tasks reliably without context window degradation.

POST https://api.teleon.dev/v1/tenants/:tenantId/queues/aq167_segment_document_into_pages_0call/enqueue
{
  "queue_category": "context_decomposition",
  "priority": 80,
  "payload": { "stage": "167", "objective": "0-Call Page Segmentation" }
}
← Stage 166: Living Document Ingest View in Single-Page Master Blueprint Stage 168: Page BM25 & Facet Index →