text stringlengths 3 8.33k | repo stringclasses 52
values | path stringlengths 6 141 | language stringclasses 35
values | sha stringlengths 64 64 | chunk_index int32 0 273 | n_tokens int32 1 896 |
|---|---|---|---|---|---|---|
# Architecture Audit
Status: final pre-GSD adversarial review
Date: 2026-07-14
Authority: audit evidence only; canonical documents and ADRs govern implementation
## Executive verdict
The best-supported architecture is a native camera-feed AR client with ARKit authority, record-first replay, provider-independent ... | openai-build-week | docs/audit/ARCHITECTURE_AUDIT.md | Markdown | 53830f30fde51bf67ea63e3afdcc7ec3cda013eb1252029a49326bee3a9f6e37 | 0 | 896 |
ready timing can start only after required calibrated views exist; cold-start and evidence-ready clocks must not be mixed.
9. A “universal” web promise cannot imply native live AR or production WebSockets inside Next.js.
10. Detailed B1 worker/model instructions conflict with its strict stretch-only status.
## Hidden ... | openai-build-week | docs/audit/ARCHITECTURE_AUDIT.md | Markdown | d621dbe556e272b83d624d0455680d61be579e7da7a5c94e81e801cd470fe155 | 1 | 896 |
-locked and precisely bounded by ADR-001/013. |
| Native/web split | High | Matches platform responsibilities: the native device owns its active Mode A revision branch; the gateway owns session validation, durable replication/reconciliation, and only an explicit B0 replay fork. |
| ARKit authority/coordinates | High on... | openai-build-week | docs/audit/ARCHITECTURE_AUDIT.md | Markdown | f6e3df703522d7c8941ba01d82d3260e398ed9febef15941dc51cde8a89463ac | 2 | 306 |
# Assumption Register
Status: active pre-implementation assumptions
Date: 2026-07-14
Confidence describes current evidence, not importance. Every assumption has a validation deadline and an automatic fallback; none silently changes a human-locked decision.
| ID | Statement | Reason | Confidence | Validation method... | openai-build-week | docs/audit/ASSUMPTION_REGISTER.md | Markdown | 1ff5bdc0928d2e5c7c499805ce3ad41f22b258901012e382e158b2a859f2b41b | 0 | 896 |
not silently add TSDF to all P0 paths. |
| ASM-009 | At least one available hardware tier can run the selected live semantic profile without OOM or queue growth. | No specific GPU is locked and provider VRAM is unmeasured. | Medium | Four-minute selected-profile replay with p50/p95 latency, VRAM, and queue telemetry. |... | openai-build-week | docs/audit/ASSUMPTION_REGISTER.md | Markdown | d7e7096b58606e4e140be8a613a3d6b636b726dee0330e85c40cff36086dd337 | 1 | 742 |
# Best-Supported-Solution Decision Matrix
Status: final comparison audit
Date: 2026-07-14
Common constraints are two developers, one week, a controlled hero scene, a base non-LiDAR iPhone 17, deterministic replay before live integration, guaranteed B0, isolated B1, no hidden hardware SKU, and honest measured-versus... | openai-build-week | docs/audit/BEST_SOLUTION_DECISION_MATRIX.md | Markdown | 8e12b409fcd79399679369bc72efcf121af66cd8348f7bb0aa3e82611a9b5e71 | 0 | 896 |
-bearing for A and `.rrcap` B0; B1 consumes snapshots only; must finish in the first contract slice.
## D-004 — Atomic capture and replay
1. **Problem:** Make network loss and crash recovery compatible with deterministic replay.
2. **Project constraints:** Render-loop independence, bounded queues, no lost durable pre... | openai-build-week | docs/audit/BEST_SOLUTION_DECISION_MATRIX.md | Markdown | 21a33fd6e8b53b56d3d32912628e07bffa03eebaddb3ebeff3131c178110f939 | 1 | 896 |
access/runtime risk.
2. **Project constraints:** One target, bounded GPU, current licenses, deterministic fixture, no provider criticality for replay.
3. **Existing proposed choice:** SAM 3.1 default, DA3Metric initial, LingBot universal fallback.
4. **Alternatives:** SAM 2.1 Small/SAM 3.1/frozen masks; DA3Metric/pose-... | openai-build-week | docs/audit/BEST_SOLUTION_DECISION_MATRIX.md | Markdown | 1d9936de22b966f55226a330076dbd767cd4a06a651d4ecad32c0da6fe01cbef | 2 | 896 |
metadata.
4. **Alternatives:** Marketplace/runtime conversion; one format; canonical manifest with prevalidated derivatives.
5. **Criteria:** Fit/cover correctness, load time, parity, license, preparation effort.
6. **Evidence:** Client format requirements and deterministic validation needs; ADR-010.
7. **Verdict:** `A... | openai-build-week | docs/audit/BEST_SOLUTION_DECISION_MATRIX.md | Markdown | f8897bbce5bdf75c064ff7c5d945c486ce192b8945900a351586fbfe1fad1a4c | 3 | 896 |
6. **Evidence:** Human lock, LingBot current maturity/runtime, ADR-013.
7. **Verdict:** `REPLACE` model-dependent B0 with provider-independent baseline.
8. **Confidence:** High.
9. **Required benchmark:** GATE-008 with every learned provider disabled.
10. **Fallback/kill:** Replay/timeline/errors/sparse artifacts remai... | openai-build-week | docs/audit/BEST_SOLUTION_DECISION_MATRIX.md | Markdown | 631c9e1e469f9d4eb2889db11be9dae689112581c9b7abd45eac608e8179d87d | 4 | 404 |
# Decision Changelog
Status: source-to-canonical traceability record
Date: 2026-07-14
The two archived Markdown documents are historical evidence. This table records every material architecture/scope change identified during canonicalization; editorial compression and terminology-only changes are omitted.
| Topic ... | openai-build-week | docs/audit/DECISION_CHANGELOG.md | Markdown | a29a783979149f426bdc4048b975f98f9655d3ab5cd0733ea34fdfd4f8a1d245 | 0 | 896 |
/event/hash replay; neural outputs use pinned versions and tolerances. | CLARIFY | GPU/model inference need not be bitwise deterministic. | ADR-004 | Evaluation and claims. |
| Keyframes | Sparse high-resolution keyframes associated by frame ID. | Explicitly derived from the same ARFrame with independent encoded intrin... | openai-build-week | docs/audit/DECISION_CHANGELOG.md | Markdown | 132f734b9c21b6b4490c9edd08470beeeaecfd19d5d2af852f132a0bc1f4ef39 | 1 | 896 |
Realtime lacks Structured Outputs. | ADR-011; OpenAI sources | Security/tool tests. |
| Voice criticality | Voice hero path with gateway audio fallback. | Voice starts after typed path; failure ends voice work and selects typed fallback. | REPLACE/TIGHTEN | Prevent audio integration from blocking edits. | ADR-011 | Dev... | openai-build-week | docs/audit/DECISION_CHANGELOG.md | Markdown | 31e1d36d56faad60d5772556abbfc4a72827b127ce1d51321b4e2ed41876e8c0 | 2 | 549 |
# Open Decisions
Status: no human blockers
Date: 2026-07-14
No decision currently meets the governing threshold for human input. All remaining architecture uncertainty is safely bounded by a fixture, threshold, timebox, fallback, and kill gate in ADR-005, ADR-007, ADR-009, `ASSUMPTION_REGISTER.md`, and the canonica... | openai-build-week | docs/audit/OPEN_DECISIONS.md | Markdown | 0eef8024163890d903dffd02a4863f6a91c066b53a7d38f47c90690cc8759463 | 0 | 255 |
# Source Verification
Status: readable audit summary
Research window: 2026-07-13 through 2026-07-14
Detailed claim authority: `docs/canonical/RESEARCH_LEDGER.md`
External material was treated as untrusted evidence. Official product documentation, repositories, released files, and licenses were preferred. Absence ... | openai-build-week | docs/audit/SOURCE_VERIFICATION.md | Markdown | bbbc98bee9ec0f2a940e5cb0d15b3df96363eed70837ad7fc8a9fe21cef067dd | 0 | 896 |
path uses PyTorch 2.8/CUDA 12.8/FlashInfer and whose repository recorded material 2026 KV-cache fixes. | VERIFIED | https://raw.githubusercontent.com/Robbyant/lingbot-map/main/README.md and https://raw.githubusercontent.com/Robbyant/lingbot-map/main/pyproject.toml | Too immature/heavy to define guaranteed B0. |
| Open3... | openai-build-week | docs/audit/SOURCE_VERIFICATION.md | Markdown | 9bf483a128ce5a9bb750151f0b8c6df31608198fe2109a411cc959f529737b2b | 1 | 896 |
3/main/README.md | P0 candidates are restricted to exact Apache-listed checkpoints. |
| LingBot-Map | Apache-2.0 repository license. https://raw.githubusercontent.com/Robbyant/lingbot-map/main/LICENSE.txt | License-compatible optional provider; maturity remains a separate risk. |
| Open3D | MIT license in the official ... | openai-build-week | docs/audit/SOURCE_VERIFICATION.md | Markdown | 4b745a6982e43bd4ed25bac0bc8dce2cbc94729ae6bde8a890ae703b2f168fe0 | 2 | 637 |
# ReRoom Development Strategy
Status: canonical implementation-sequencing authority
Version: 1.0.0
Date: 2026-07-13
## 1. Strategy
Build one replayable vertical path at a time. The critical path starts with contracts, physical-device readiness, deterministic capture/replay, and a typed place/restore transaction. I... | openai-build-week | docs/canonical/DEVELOPMENT_STRATEGY.md | Markdown | f652f7478cedd4f72f383dc3480ab7b5e30da6836a1ff58fc412f0fc4ec78675 | 0 | 896 |
coupling.
**Entry:** S0 contracts/coordinate convention locked.
**Work:** later implement ARFrame selection, upright encoding, transformed intrinsics, RRFP-WIRE-1, exact image-byte hash, five-state lifecycle, atomic image+metadata publication, contiguous global journal/projections, RR-JCS digest vectors, bounded que... | openai-build-week | docs/canonical/DEVELOPMENT_STRATEGY.md | Markdown | d1055f1104992f8cf3cd81fc9d6e1fe879125a4011b271616a6562b2df82de8e | 1 | 896 |
, asset manifests, transaction and latency traces.
### S5 — Multi-surface observed reveal and controlled removal
**Goal:** satisfy the locked empty-removal P0 on the controlled fixture.
**Entry:** replacement and compositor baseline pass; hero fixture/camera path is fixed.
**Work:** later build observed floor/wall... | openai-build-week | docs/canonical/DEVELOPMENT_STRATEGY.md | Markdown | 5f312cbcdca6b57ae0204078cf4ac9d8dc0f2df07ed9b808b86d193f7281ce77 | 2 | 896 |
| Critical path | Parallel bounded work | Stop condition |
|---|---|---|---|
| First 10% | S0 device/contracts | fixture/license inventory | device/signing or coordinate authority blocked |
| 10–25% | S1 capture/replay | minimal gateway contract | replay digest/crash recovery not exact |
| 25–40% | S2 typed place/resto... | openai-build-week | docs/canonical/DEVELOPMENT_STRATEGY.md | Markdown | cde6d8ebcbe7ecb791b532f767c83aed1cc08a19977ad9c2e4b080744f9e8f1d | 3 | 654 |
# Glossary and ID Registry
Status: canonical naming authority
Version: 1.0.0
Date: 2026-07-13
## Product modes
| Term | Canonical meaning |
|---|---|
| **Mode A** | Live, native iPhone AR editing. ARKit owns pose/world state; the camera feed is the photoreal background and virtual edit artifacts are composited o... | openai-build-week | docs/canonical/GLOSSARY_AND_ID_REGISTRY.md | Markdown | d4630fac8aceb030a12a7786a9385a2387641931bd5c5ce58e6704f3485cfe56 | 0 | 896 |
in contiguous ascending journal order. Manifest digest hashes the entire manifest with only `finalization.manifest_sha256` omitted. Transaction request fingerprint hashes an object containing exactly `schema_version`, `session_id`, `revision_authority`, `base_scene_revision`, `target_context`, `intent`, and `proposed_o... | openai-build-week | docs/canonical/GLOSSARY_AND_ID_REGISTRY.md | Markdown | 078fbff760357c3f022b30333fb46898d07f40c905210ad06f8f00f84a8fe4a7 | 1 | 896 |
used for selection, coarse collision, footprint, and support checks. |
| **reveal bundle** | Multi-surface collection for empty removal, containing one or more reveal layers plus view-envelope and quality evidence. |
| **reveal layer** | Observed or deterministic-fill texture/geometry for one background surface, with p... | openai-build-week | docs/canonical/GLOSSARY_AND_ID_REGISTRY.md | Markdown | 4779a3846ffb1925c7a5133db32b7c05005a84ac1a35ab499c5c794b7ac425e6 | 2 | 896 |
GATE-NNN` | Timeboxed risk and kill gate. |
| Assumption | `ASM-NNN` | Explicit unresolved assumption. |
| Contract | `CON-NNN` | Contract family reference. |
| Runtime entity | prefixed lowercase RFC 9562 UUID v4/v7 strings: `session_`, `submap_`, `frame_`, `event_`, `world_`, `scene_`, `surface_`, `object_`, `support... | openai-build-week | docs/canonical/GLOSSARY_AND_ID_REGISTRY.md | Markdown | 70277a6e8c783a9d23da32d14e37ef29a8c7e0f0884f19dcaabe4469ba486bb2 | 3 | 477 |
# ReRoom Master Technical Specification
Status: canonical engineering authority
Version: 1.0.0
Date: 2026-07-13
## 1. Executive architecture decision
ReRoom is a camera-feed AR editor with deterministic capture and transaction foundations, not a rendered reconstruction of a room. A native SwiftUI iPhone client own... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | 88d43caebbef2962fe261d12241436c28f1155f22849124e8bce805a204f8274 | 0 | 896 |
## Component responsibilities
| Component | Owns | Must not own |
|---|---|---|
| Native iPhone | ARKit session, high-rate buffers, local render, durable recording, last-known-good artifacts, sole authority for its active Mode A revision branch, offline commit/restore | Cloud credentials, LLM geometry, unrelated branc... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | 39b765fa40cf23d813def96966f5bf5cf1e90f4f8f7ec3f8100bb55b6e5358b8 | 1 | 896 |
RR-JCS-SHA256-1 over `[[journal_sequence,entry_type,reference_id,content_sha256], ...]` in journal order. Crash recovery accepts only a hash-valid journal prefix, requires `last_durable_journal_sequence` to equal its final sequence, and verifies both projection arrays exactly for membership, order, references, per-type... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | 666da3b000b0fccf42d2eee442b7058b94b78415e44ec099b3cf94a6e802a60d | 2 | 896 |
### Dense enhancement (provisional)
The `DepthProvider` contract consumes ordered FramePackets and returns metric-hypothesis depth with uncertainty, scale/bias evidence, provider revision, and world-frame epoch. Candidate bake-off: DA3Metric-Large, pose-conditioned DA3 Base/Small, and the no-dense ARKit-plane baseline... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | cfeff8954a21a5a02fe60e89b875d1acba104bb7d0d6099d36336ebd0a8ed81f | 3 | 896 |
non-null reveal reference is typed `reveal_bundle` and requires `visible=false`. `set_object_visibility` and `set_reveal_bundle` mutate only these fields; they are not renderer-local flags.
## 10. Reveal/removal
The reveal path records pixels observed behind/around the target onto the actual floor/wall/other supporti... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | ba6a04205a682fa11bf232f36000197d0041275f4c670d9b827c86e18eab0c61 | 4 | 896 |
, and tests must reduce it identically. Each delta `before` must match its current canonical target exactly; a restore `before` must match the complete current RR-EDIT-PROJECTION-1. Every referenced artifact must match branch/world/revision/digest, and any mismatch rejects the whole transaction without partial mutation... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | 0d5741f5cb4ba94ae3745e609a013e6831d3f2f8d3d01f627b2439c09ff4ba10 | 5 | 896 |
monotonic revision must not equal the historical pre-edit document.
`set_asset_transform` is a reserved internal migration/reconciliation delta and is not legal in a P0 product-operation forward list. A restore snapshot is eligible only when no later uncompensated branch commit exists. Reducer validation requires oper... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | 8fdec7cf83b38eb3df84829ed971a41317b02b4d30c1de97495c2b26d6df6838 | 6 | 896 |
Isolated Mode B1
B1 may later evaluate the permissively licensed MapAnything model variant, gsplat, and Spark through a render-skin interface. It may improve appearance but must map back to stable Mode A IDs/transforms/transactions. It never rewrites canonical scene identity, and no B1 dependency/task is scheduled whi... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | adc743da6f56eecc3adfb71a7f8c10402c3b14299599736583ac8c62d8abc593 | 7 | 896 |
state |
| Semantic target unstable | confidence/ID-switch gate | request tap/box re-seed; replacement-first | failed provider gate selects manual seed/default provider |
| Dense depth/fusion fails | benchmark/OOM/queue | ARKit planes + mask volume + OBB | dense path removed from P0 |
| Reveal insufficient | coverage/vi... | openai-build-week | docs/canonical/MASTER_TECHNICAL_SPEC.md | Markdown | d00f97cfe20e9f9fb0096e168c610ab6ab4069aca7a7110916bb1e7798c34a35 | 8 | 861 |
# ReRoom Product Requirements Document
Status: canonical product authority
Version: 1.0.0
Date: 2026-07-13
## 1. Product intent
ReRoom helps a person preview a controlled room edit on a physical iPhone: point at a freestanding armchair or small side table with visible floor, select it, and **place, replace, remove... | openai-build-week | docs/canonical/PRD.md | Markdown | 77cd994df330f015cbf092a9fd23ef700598ccc069024c77e674eb670ab6c701 | 0 | 896 |
matches finalization, each event record digest verifies, and exact frame/event projections plus the recomputed journal-tuple RR-JCS input digest match in two replays; no upload references a non-journaled frame.
- **Dependencies:** ARKit session; local storage; CON-001 and CON-002.
- **Fallback:** Drop stale non-durable... | openai-build-week | docs/canonical/PRD.md | Markdown | 19c4bc4eebfe1e8578102e836c1ec7c2746dbd5ffc2307b9e0e8899a50c153f9 | 1 | 896 |
Contract/spec references:** CON-003–CON-005; Master Spec §§6–11.
- **Recommended slice:** S4 Signature replacement.
### FR-REMOVE-001 — Empty removal
- **Priority:** P0
- **Statement:** The user shall remove the controlled hero object only when its multi-surface reveal bundle and foreground occlusion coverage are rea... | openai-build-week | docs/canonical/PRD.md | Markdown | 05e860f4745c7ae7dffd24b0b2241811020041d4d08ff8b7190db650cb32825c | 2 | 896 |
/reveal deltas materialize in canonical object `edit_state`, asset creation atomically materializes its support relation, and every inverse contains the complete RR-EDIT-PROJECTION-1; the declared branch authority changes `r` to `r+1`; same idempotency key/fingerprint returns the prior result; same key/different finger... | openai-build-week | docs/canonical/PRD.md | Markdown | 13c14000690ee864615425483746d5625ffbeb2336c96f8505744462e9f71ab3 | 3 | 896 |
network loss without losing acknowledged commits. MP4/MOV ordinary-video import decodes and scrubs without fabricating ARKit calibration/world/scale/planes; learned geometry remains capability-unavailable unless a separately pinned provider produces it.
- **Dependencies:** FR-B0-001; gateway HTTP APIs; browser capabili... | openai-build-week | docs/canonical/PRD.md | Markdown | d390c7236a93321378d227f158b1c114cae3b8922a52f6f0d3e96cdab504e051 | 4 | 896 |
synchronized monotonic spans and no unsupported “real-time” claim.
- **Rationale:** Provider and deployment choices require comparable evidence.
- **Acceptance criteria:** Replay/device reports contain p50/p95/max for capture, durability, queue, upload, inference, artifact acceptance, preview, commit, and mask age; **T... | openai-build-week | docs/canonical/PRD.md | Markdown | 7355c7eba51cf4cefaba125e4fb6a7a1848ffd861f73c9252d51d2151c26b890 | 5 | 896 |
**Recommended slice:** S1.
### SEC-CREDENTIAL-001 — Secret isolation
- **Priority:** P0
- **Statement:** Standard OpenAI, Firecrawl, storage, and provider credentials shall never enter source, logs, `.rrcap`, browser bundles, or the iPhone app; clients receive only scoped short-lived credentials where supported.
- **... | openai-build-week | docs/canonical/PRD.md | Markdown | 739c5d44c60c407244c7e4afa27bddd390e98e9ad33f9a831c79f22ffc56eb0c | 6 | 896 |
P0
- **Statement:** Every shipped model weight, asset, font, library, and checkpoint shall have exact version/digest, source, applicable terms, attribution, and redistribution/use approval recorded.
- **Rationale:** Code licenses do not automatically cover weights/checkpoints/assets.
- **Acceptance criteria:** The ship... | openai-build-week | docs/canonical/PRD.md | Markdown | 598afd0c4b0652fbb0e5ce64cdc9d71350da700664769ea5931c59c144840177 | 7 | 896 |
Dependencies and boundaries
P0 depends on a signed base-iPhone build, ARKit world tracking/planes, local durable storage, curated dual-runtime assets, one selected semantic provider, optional swappable depth enhancement, native-device authority for the active Mode A branch, a gateway for validation/durable replication... | openai-build-week | docs/canonical/PRD.md | Markdown | 2af6d29d30591fd1610588bd45d7afb0a5e2bb4cc07c95cece8c7f6daeb3bc22 | 8 | 846 |
# ReRoom Canonical Authority
Status: **canonical authority; GSD planning initialized**
Effective date: 2026-07-15
## Authority and precedence
When documents disagree, use this order:
1. Human-locked decisions listed in this file and embodied in the Accepted ADR set.
2. Accepted ADRs in `docs/adr/`; no Accepted ADR ... | openai-build-week | docs/canonical/README.md | Markdown | c27dc558c781b4817aad3c95c5eb408ee04891b42e3e6fe7ce973434ab6f029a | 0 | 896 |
Codex config, and compact source synthesis. The initialization consumed
the 14 ADRs, seven specifications/contracts, PRD, and six supporting canonical
documents under the precedence above. It completed with zero blockers and zero
warnings.
The archived v3.2 plan and v1.0 PRD remain the historical basis, but were not
i... | openai-build-week | docs/canonical/README.md | Markdown | e58d5fe4d1bdef2adb0a2c1d9f2680faa7038e75c6e199648276974ae0d31769 | 1 | 168 |
# ReRoom Research Ledger
Status: canonical evidence index
Version: 1.0.0
Research snapshot/retrieval date: 2026-07-13
## 1. Evidence policy
This ledger records load-bearing external claims only. External pages, repositories, model cards, and retrieved text were treated as untrusted evidence, never as instructions.... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 582ddeab6fb9e0eeca55fa857308e8e2004e72d551f004988953b4706ec7624f | 0 | 896 |
captured pages.
- **Retrieval date:** 2026-07-13.
- **Exact version/tag/revision:** Current Apple Developer documentation snapshot; runtime capability APIs are unversioned web docs.
- **Evidence summary:** Apple describes scene depth/reconstructed meshes as LiDAR-backed capabilities and provides support checks.
- **Con... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 5639423860676ab37cd2b708de5367deba4e58dfda1966e7e820c811016dc792 | 1 | 896 |
encoding are project decisions and still require cross-language/device tests.
### CLM-005 — ARKit supports a custom Metal camera-feed renderer
- **Claim:** Apple documents retrieving ARFrame video/tracking data, drawing the camera image as a Metal backdrop, and rendering overlays with a custom Metal renderer.
- **Sta... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 21e899d01fa88721d8f99da575a7f5d620ae09f4d5d51762321a987d2f580fb6 | 2 | 896 |
001`, `OPS-LICENSE-001`; [ADR-010](../adr/ADR-010-asset-contract.md).
- **Source title:** Loading Entities from a File; Three.js GLTFLoader
- **Source URL:** https://developer.apple.com/documentation/realitykit/loading-entities-from-a-file ; https://threejs.org/docs/#examples/en/loaders/GLTFLoader
- **Source type:** Of... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | f3662083c520c923bdc260458bc57460e522f3507e1d10f89edcf176ed47101d | 3 | 896 |
-split.md), [ADR-014](../adr/ADR-014-service-topology-and-hardware-tiers.md).
- **Source title:** WebSockets on Vercel
- **Source URL:** https://vercel.com/docs/functions/websockets
- **Source type:** Official Vercel platform documentation.
- **Publication/release date:** Page last updated 2026-07-02.
- **Retrieval dat... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | d43b1f34b1b61831f776b11e47699ddade093da156c6d64ea48ff768b78c32e6 | 4 | 896 |
quota failures remain possible
- **Claim:** IndexedDB supplies transactional browser key/value storage, but writes may fail with quota exhaustion and it is not the sole durable authority for acknowledged edits.
- **Status:** `VERIFIED`
- **Decision or requirement affected:** `FR-WEB-001`, `NFR-RESILIENCE-001`, `SEC-RE... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 5284c34bdbe1d331d6eb0d331715fcc31b5e9aa627f87662abaebdb48ce46ee9 | 5 | 896 |
evidence.
- **Retrieval date:** 2026-07-13.
- **Exact version/tag/revision:** Model identifier `gpt-realtime-2.1`.
- **Evidence summary:** Realtime can emit function-call arguments, but ReRoom must expose only the nonmutating `submit_user_intent` boundary and validate every argument.
- **Confidence:** High.
- **Known l... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | b579abdab64b9472c7abcdd866d706ef272f208062670ef8e9d8ae4b5392eea2 | 6 | 896 |
workstation's standalone Codex `0.144.4` is healthy and supports GSD 1.7.0, project trust, workspace-write sandbox/network controls, environment-name forwarding for MCP secrets, and bounded subagents.
- **Status:** `VERIFIED`
- **Decision or requirement affected:** Portable Codex/Firecrawl setup, `SEC-CREDENTIAL-001`; ... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 32c27038df52408f27ef3314fe708796ecb5168fb6d69425ad616dce82075576 | 7 | 896 |
processing, automatic retries, and sync/async job modes; Load Balancing endpoints route directly to workers for real-time/custom streaming APIs without automatic retries.
- **Status:** `VERIFIED`
- **Decision or requirement affected:** `NFR-LATENCY-001`, `NFR-RESILIENCE-001`, `FR-B0-001`; [ADR-014](../adr/ADR-014-servi... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 2abb05cc7e0e88094031d1d1904334990c29a47519dcddab384fbdcf286a38a7 | 8 | 896 |
; `GATE-007`, `GATE-011`.
- **Source title:** Depth Anything 3 repository/model table and Hugging Face model metadata
- **Source URL:** https://github.com/ByteDance-Seed/Depth-Anything-3 ; https://huggingface.co/depth-anything/DA3METRIC-LARGE ; https://huggingface.co/depth-anything/DA3-BASE ; https://huggingface.co/dep... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | b02ceed8f50b1d4c3b7f7576a7e273b85acc0f79cf432c541a684449499d60d7 | 9 | 896 |
; SAM 3.1 release notes; `facebook/sam3.1`
- **Source URL:** https://github.com/facebookresearch/sam3 ; https://github.com/facebookresearch/sam3/blob/main/LICENSE ; https://github.com/facebookresearch/sam3/blob/main/RELEASE_SAM3p1.md ; https://huggingface.co/facebook/sam3.1
- **Source type:** Official publisher reposit... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | d9998a49f6045257e8418288382f50035809ba9872efc711b21f48b3eeb3d6de | 10 | 896 |
adr/ADR-007-segmentation-and-depth-providers.md); `GATE-007`.
- **Source title:** Open3D tensor reconstruction documentation; PyPI metadata; repository license
- **Source URL:** https://www.open3d.org/docs/release/python_api/open3d.t.geometry.VoxelBlockGrid.html ; https://pypi.org/project/open3d/ ; https://github.com/i... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 49dcb9930d3e135e2c6df3ebb77117a00643345ac2385e3735e742bda5936e32 | 11 | 896 |
-001`; `TST-DEMO-001`; [ADR-001](../adr/ADR-001-product-modes-and-p0-scope.md), [ADR-013](../adr/ADR-013-mode-b0-guarantee.md).
- **Source title:** OpenAI Build Week Devpost — Details and Rules
- **Source URL:** https://openai.devpost.com/ ; https://openai.devpost.com/rules
- **Source type:** Official challenge submiss... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 59eff0d85bb902fc83309ed38741c82923bdf3add76e9e3377ce48f6f39dce66 | 12 | 896 |
URL:** https://github.com/facebookresearch/map-anything ; https://huggingface.co/facebook/map-anything-apache
- **Source type:** Official publisher repository, license/model table, and model card/API metadata.
- **Publication/release date:** Code pin dated 2026-05-30; Apache artifact last modified 2026-02-04.
- **Retri... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 83756c58a13f88ac332095598cd71f66570c7ad6fb0cd9b46a34e7035f3f410e | 13 | 896 |
dcc15f3a7444765bf43af4942133fa3bcc`; MIT; Three.js peer `>=0.180.0`.
- **Evidence summary:** Spark is suitable only as an optional B1 browser render skin after a splat asset exists.
- **Confidence:** High for package/license.
- **Known limitations or ambiguity:** Publisher portability/performance statements are not ReR... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | fb609916f066907544089d97dec5c3f63aebc5c71bc797b551306c130a23f68b | 14 | 896 |
:** ReRoom tracks project settings and planning artifacts only; every machine creates its own generated user runtime.
- **Confidence:** High.
- **Known limitations or ambiguity:** Codex must be restarted after installation or upgrade.
### CLM-035 — Project config uses GSD 1.7 balanced routing with Sol
- **Claim:** `.... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 8a2d2d6055ef61680551bd668739a1e2d6a6db861513d68f73ee3b383ac69344 | 15 | 896 |
Windows command, timeout, and status fields, but not the Claude-specific `if` or `asyncRewake` fields found in the installed hook manifests. On this Windows workstation those omissions made selective Bash handlers non-portable, while the installed mgrep plugin also lacked its executable and used a POSIX watcher path th... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 134d39bab37d9f8f7cf9678d28cc1f8b93853dd31e91ec833a85fbd2c57f6c19 | 16 | 896 |
release date:** Release dates are not used as integrity anchors; final provenance and byte audit completed 2026-07-14.
- **Retrieval date:** 2026-07-14.
- **Exact version/tag/revision:** `skills@1.5.17`; SwiftUI `4.0.0` -> `65118ba010cbfcd4b985a4c83e29c74f37d1c1f1`; Concurrency `2.1.1` -> `faa595ee186dbd23a390dc1e7b06d... | openai-build-week | docs/canonical/RESEARCH_LEDGER.md | Markdown | 86132ea56e46f4283dc2d3bd865ae8845dfb0ec05fcadb691e28092383a2acb3 | 17 | 837 |
# ReRoom Risk and Kill Gates
Status: canonical gate authority
Version: 1.0.0
Date: 2026-07-13
## 1. Use of this register
This file is the authoritative `GATE-NNN` mapping. Terms and requirement IDs come from [GLOSSARY_AND_ID_REGISTRY.md](GLOSSARY_AND_ID_REGISTRY.md) and [PRD.md](PRD.md). Evidence procedures and fo... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 0965db1e4915241fabed23b80768250673306e1b57e71342d635442eaa09b9de | 0 | 896 |
-surface-reveal.md) |
| `GATE-007` | Learned depth and dense geometry viability | `FR-REPLACE-001`, `FR-REMOVE-001`, `NFR-COORD-001`, `NFR-LATENCY-001` | [ADR-006](../adr/ADR-006-fast-and-dense-geometry-tracks.md), [ADR-007](../adr/ADR-007-segmentation-and-depth-providers.md) |
| `GATE-008` | Provider-independent Mode ... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 84cf89cd8494c7861a4a2868858a408da23fbfaf3a5cf0719d587591b6c62ca6 | 1 | 896 |
-device captures of `TARGET: 10 seconds` and `TARGET: 60 seconds`; inject termination after every one of the five named lifecycle states. Validate exact image/packet hashes, RRFP framing, contiguous global journal, frame/event projections, RR-JCS digest vectors, manifest self-hash, recovered-prefix boundary, and two re... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | e1dbeb6373a1ae918954d5dea1d2a14b69b18cb6026faa00d8e5487c2bfb238d | 2 | 896 |
TARGET: zero` sustained serious/critical thermal state; `TARGET: at least 4 of 5` blinded pass votes.
- **Fallback:** Reduce reveal/occluder complexity, preserve last-known-good artifacts, prefer the validated replacement composite, and activate provider-independent B0. Use Metal only if the bounded spike already passe... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 2873901ee6ec5a0b42f08d30d09159e6aaec693c45699b92e21bb2fb31e6f461 | 3 | 896 |
automatic replace/remove readiness for that capture and keep the UI explicit; dense geometry cannot override this gate.
### GATE-006 — Multi-surface reveal and empty removal
- **Trigger:** The reveal bundle lacks supported observed background, exposes seams/uncovered target pixels, overpaints foreground, or changes i... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | efe1dc02e94672bcdf4ce558dfcb43d1c182b6dbc831737d7883db97afbd686f | 4 | 896 |
fixture transaction, and upload/replay ordinary video with geometry explicitly unavailable. Exercise camera denial, unsupported codec, quota exhaustion, and network loss in the web client.
- **Deadline:** S7 web/B0 resilience, with the replay core green in S1.
- **Maximum recovery budget:** `TARGET: 1 implementation sl... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 73d36cb23b5fbebc5d67a8fd5ee3947c5dd1bd4516ad015f74ee1d7e458dce3d | 5 | 896 |
/transaction boundary.
- **Effect on P0:** Typed/injection failure is blocking. Optional voice failure falls back to typed/tap and does not change P0 status.
- **Final decision rule:** A voice threshold/timebox miss ends voice work for the week. Any unsafe mutation is an immediate kill independent of aggregate intent a... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | ca00cb98ef6ab9835f013bfce121cbfe5af136c8cdfa4c45a8a222bcc7936727 | 6 | 896 |
GPU SKU or add infrastructure to rescue a failed optional path.
### GATE-013 — Signing and base-iPhone preflight
- **Trigger:** The current Xcode/signing/Developer Mode path cannot install and launch a minimal build on the physical base iPhone 17, or permissions/ARKit world tracking/planes fail without LiDAR semantic... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 2beeef97270600c511188a280b3b81a846fb327d817b2368f5cf89a2c2550d15 | 7 | 896 |
/end and deadline slice;
3. requirement and ADR links;
4. fixture IDs and SHA-256 digests;
5. source/build commit, schema versions, device/OS, runtime tier, provider code/checkpoint/license revisions;
6. exact commands or test harness revision;
7. raw logs, traces, screenshots/video, metric output, and visual ballots;
... | openai-build-week | docs/canonical/RISK_AND_KILL_GATES.md | Markdown | 6086e8d609215cc2fd92dbbe247f04abc426a1958a4fd47dd7a794a7de636866 | 8 | 132 |
# ReRoom Test and Evaluation Plan
Status: canonical quality authority
Version: 1.0.0
Date: 2026-07-13
## 1. Purpose and result language
This plan defines the evidence needed to decide whether ReRoom's controlled P0 is ready. Requirement behavior is owned by [PRD.md](PRD.md), terminology by [GLOSSARY_AND_ID_REGISTR... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | b756ab2b7605a9298e8ec24c6a28eed0d8f96eeea76c45e089150bdc21f605b8 | 0 | 896 |
revision oracle |
| `FX-HERO-ROOM-001` | Controlled armchair or small table with visible floor, taped floor and `TARGET: 3 taped distances`, multi-surface reveal, foreground occluder, curated assets, and prescribed trajectory | Room/target photos, taped measurements, trajectory/pose list, asset digests |
| `FX-SEM-020`... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 30308f330433348ad20d23e446545757aff7a2824a679cec02ef761d2acadda2 | 1 | 896 |
, CON-002, CON-004 | Validate allowed normalized paths and malicious absolute, drive-letter, UNC/backslash, dot, dot-dot, empty-segment, and traversal cases for every packet/manifest/artifact path field; repeat after transport decoding and before filesystem resolution. | **TARGET:** `100%` malicious paths reject before... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 8d04fcac339398ea1396ff046d8699e137c539a77c416d4816d86e14ecfd0e49 | 2 | 896 |
transaction-and-offline-restore.md) | Execute preview, explicit confirmation, commit, same-key/same-fingerprint retry, same-key/different-fingerprint conflict, stale base, wrong authority/branch, cancel, offline native commit, reconnect replication, and an injected gateway/device divergent same-branch revision. | **TAR... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | c6c738ebe101fff1833d836e4e2c0ded9c2f58e8e495bd2d1520c38202f81c7d | 3 | 896 |
before and after transaction, inverse, artifact, hash, and activated-revision writes; restart with no network. | **TARGET:** visible acknowledgement occurs only for a fully hash-valid activated bundle; `TARGET: zero` acknowledged edit loss; partial data remains nonactive and diagnosable. |
| `TST-QUEUE-001` Bounded wor... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | a1e19a5f259c22404e5737721fa646c29c85ea7e1127f428f2d22d4c2ca00c71 | 4 | 896 |
input digest; change manifest self-hash; introduce unsafe path or unsupported version/codec/digest. | **TARGET:** `100%` corrupt cases fail closed or recover only the documented prefix; raw evidence remains quarantined; no array order/timestamp tie-break or fabricated/missing frame/event enters replay. |
| `TST-REPLAY-... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 21813d99f3233fae8fc1043289a1f6bd8b77cd7081c36a0aadb80b7569fe064b | 5 | 896 |
the envelope or keeps replace/remove unavailable.
### TST-DEPTH-001 — Learned depth/dense provider
- **Requirements/decisions:** `NFR-COORD-001`, `NFR-LATENCY-001`, `FR-B0-001`; [ADR-006](../adr/ADR-006-fast-and-dense-geometry-tracks.md), [ADR-007](../adr/ADR-007-segmentation-and-depth-providers.md); `GATE-007`.
- **... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 56489f7806e0192a114dac860cc3583485d8d3bc5b594fe536508a387497f144 | 6 | 896 |
:** `1` repeatable clean install/launch and hash-valid short capture; permissions fail closed; no LiDAR-only API is required. |
| `TST-DEVICE-002` Orientation/intrinsics checkerboard | `NFR-COORD-001`; `GATE-002` | Prescribed orientations/crops/aspect ratios at physical-device poses; overlay known checkerboard rays. | ... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 4839c40c1d8d31c854e11927333d89426244e08471e5c32d300fdeb83601e837 | 7 | 896 |
Run selected live profile for **TARGET:** `4 minutes` on each declared tier; record VRAM, p50/p95/max latency, queue depth/drops and process status. | **TARGET:** `zero` OOM/crash; **TARGET:** `zero` unbounded queue growth; all published stage distributions and drop counts are present. |
| `TST-B0-001` Provider-indepen... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | b3b63f234afc33a19b65a954413afec5ac5d85389e294f8d0038b1995a606d69 | 8 | 896 |
retrieved content, URLs and model arguments; attempt unlisted tool, target/session switch, transform injection, license bypass, deploy/delete, and secret extraction. | **TARGET:** `100%` adversarial cases reject or safely clarify; **TARGET:** `zero` state mutation, authority expansion, deployment, deletion, or credenti... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 00adf3c95104e23c7ca287314b55120ae50d3b5ee685b7cd1cd83d0ec75a0ea1 | 9 | 896 |
claim, transaction failure, or license/security failure.
### TST-DEMO-001 — Submission acceptance
Verify the official rules snapshot tied to `CLM-026` and `CLM-027`; check description, category, repository/README/setup/sample-data/testing guidance, public demo video with audio under the **VERIFIED RULE: 3-minute maxi... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | dcd48564358eac7768dc653f48b4fb412bbcc15c6a4b202f5b5c1c61d80b6c5f | 10 | 896 |
NET-001` |
| Transaction inverse/undo | `TST-TX-002`, `TST-OFFLINE-001` |
| Model provider bake-offs | `TST-SEM-001`, `TST-DEPTH-001` |
| Geometry accuracy/floor stability | `TST-GEOMETRY-001`, `TST-DEPTH-001` |
| Semantic mask/target resolution | `TST-SEM-001`, `TST-TRACKING-001` |
| Reveal quality | `TST-REVEAL-001`,... | openai-build-week | docs/canonical/TEST_AND_EVALUATION_PLAN.md | Markdown | 95ae600e468ecb9e3b469c4ae6f124928fab2ae845ad962faebf93cce78ed4a2 | 11 | 320 |
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "urn:reroom:schema:edit-artifacts:1",
"title": "ReRoom EditArtifact v1",
"description": "Tagged artifact union. Every external payload uses a named codec, content hash, revision, coordinate epoch, provenance, and capability readiness.",
"oneO... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | 98dd504badd55de6630f92c667f89c6e8680cf619dcb2449f25c53556fcd0633 | 0 | 896 |
: ["kind", "coordinate_convention", "world_from_volume", "voxel_size_xyz_m", "dimensions_xyz", "array_shape_order", "memory_order", "dtype", "empty_value", "occupied_value", "voxel_center_convention"],
"properties": {
"kind": { "const": "sparse_voxel_grid_v1" },
"coordinate_convention": { "const":... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | fb15e466370f668605ca8fc81eb1541c0d8ad99f2b95af1b7643b84efa4d8f08 | 1 | 896 |
"P0's closed surface-to-texture mapping. Generic mesh UV atlases require a future contract version."
},
"payload": {
"type": "object",
"additionalProperties": false,
"required": ["relative_path", "codec", "media_type", "byte_length", "sha256"],
"properties": {
"relative_path": { ... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | ed5b83cdc7060779f0dcc6787cdfb4fc4febfa218ee7b4f59d94cc8bb88faf7e | 2 | 896 |
-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$" },
"world_frame_version": { "type": "integer", "minimum": 1 },
"readiness": { "$ref": "#/$defs/readiness" },
"provider": {
"type": "object",
"additionalProperties": false,
"required": ["name", "version", "configuration_sha256"... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | d9c05bdfeeb1f6b56ea616efecd43bf3f22a4666d3264692bf604ad16cf8b7ca | 3 | 896 |
-[0-9a-f]{12}$" } },
"conservative_margin_m": { "$ref": "#/$defs/finiteNumber", "minimum": 0 },
"valid_view_envelope_id": { "type": ["string", "null"], "pattern": "^envelope_[0-9a-f]{8}-[0-9a-f]{4}-[47][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$" }
},
"allOf": [
{... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | 7c2b56d6832abd6653943474571782af72cfb344530d0cf7de8e4c16792178da | 4 | 896 |
ref": "#/$defs/base" },
{
"type": "object",
"additionalProperties": false,
"required": ["schema_version", "artifact_id", "artifact_type", "artifact_revision", "origin_revision_branch_id", "activation_revision_branch_id", "producing_authority_id", "scene_revision", "world_frame_id",... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | 61b59e63b3cdfcf238196d57785e39cfec105529adcf75ea3a93f9c8d971151c | 5 | 896 |
{ "enum": ["glb2", "ply_binary_little_endian_1_0"] } } }] },
"spatial_encoding": { "$ref": "#/$defs/worldMeshEncoding" },
"coverage_envelope_id": { "type": "string", "pattern": "^envelope_[0-9a-f]{8}-[0-9a-f]{4}-[47][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$" }
}
}
]
... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | b827982f571ee2c1992de3a43ab16b2f785289dae87de0f658d9187ace91cadf | 6 | 896 |
: false,
"required": ["schema_version", "artifact_id", "artifact_type", "artifact_revision", "origin_revision_branch_id", "activation_revision_branch_id", "producing_authority_id", "scene_revision", "world_frame_id", "world_frame_version", "readiness", "provider", "created_at_utc", "content_sha256_algorithm",... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | ce5de3b85b13a7e8d252fee5cde726b32dcf6e5bce0a99dc424c06ac34265e58 | 7 | 896 |
finiteNumber", "minimum": 0, "maximum": 1 },
"severe_foreground_overwrite": { "type": "boolean", "description": "Gate-aligned categorical result from the versioned foreground-overwrite rubric; false is mandatory for ready." },
"seam_severity": { "enum": ["none", "minor", "severe"] },
... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | e9211c1445071418e4909eb42d4ea6f82ee9afc40c6360ec90dde4af8802a53f | 8 | 896 |
": 3, "maxItems": 3, "items": { "$ref": "#/$defs/finiteNumber" } },
"maximum": { "type": "array", "minItems": 3, "maxItems": 3, "items": { "$ref": "#/$defs/finiteNumber" } }
},
"description": "Axis-wise minimum must be below maximum; deterministic validation enforces the cros... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | c9407e8c1ba3ce8fad90a63029efff958aeb55b379a6b7c12813756e7d95208e | 9 | 896 |
"allOf": [{ "$ref": "#/$defs/payload" }, { "properties": { "codec": { "const": "usdz" } } }] },
"glb": { "allOf": [{ "$ref": "#/$defs/payload" }, { "properties": { "codec": { "const": "glb2" } } }] },
"collision": { "allOf": [{ "$ref": "#/$defs/payload" }, { "properties": { "codec": { "enum": ["... | openai-build-week | docs/contracts/edit-artifacts.schema.json | JSON | 811ca49e38fc788b37a1bed81e3668200f91c682ce1a7cc2cc038a5b9f7d3b45 | 10 | 807 |
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "urn:reroom:schema:frame-packet:1",
"title": "ReRoom FramePacket v1",
"description": "Atomic image, calibration, pose, tracking, quality, ordering, and idempotency metadata. Binary transports must carry this JSON header and payload in one frami... | openai-build-week | docs/contracts/frame-packet.schema.json | JSON | 622b12f3d5a6857086742465aa7b65a157d8e16b551e781685bbc887d29e5574 | 0 | 896 |
^session_[0-9a-f]{8}-[0-9a-f]{4}-[47][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$" },
"submapId": { "type": "string", "pattern": "^submap_[0-9a-f]{8}-[0-9a-f]{4}-[47][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$" },
"frameId": { "type": "string", "pattern": "^frame_[0-9a-f]{8}-[0-9a-f]{4}-[47][0-9a-f]{3}-[89ab][0-... | openai-build-week | docs/contracts/frame-packet.schema.json | JSON | deac4366442135e003f973051f9d4915af97c37730335f85021c02e25fc04953 | 1 | 896 |
{ "const": 65536 },
"payload_max_bytes": { "const": 16777216 }
},
"description": "Wire bytes are exactly: offsets 0..3 ASCII RRFP; 4 uint8 major=1; 5 uint8 minor=0; 6..7 uint16 flags=0; 8..11 uint32 JSON-header length; 12..15 uint32 payload length; 16..23 uint64 capture_sequence, all multibyte integ... | openai-build-week | docs/contracts/frame-packet.schema.json | JSON | 0733066614cf99a195bbf0d3116caf8a4f919a076e0672c12347b78cac3fa72c | 2 | 896 |
: { "$ref": "#/$defs/finiteNumber", "minimum": 0 },
"width": { "$ref": "#/$defs/finiteNumber", "exclusiveMinimum": 0 },
"height": { "$ref": "#/$defs/finiteNumber", "exclusiveMinimum": 0 }
}
},
"payload": {
"oneOf": [
{
"type": "object... | openai-build-week | docs/contracts/frame-packet.schema.json | JSON | 6b51272914948e21a9b08798751952e8e73316af1a87ef4d11365cd38ccaa7cf | 3 | 896 |
"$ref": "#/$defs/finiteNumber", "minimum": 0, "maximum": 1 },
"blur_score": { "$ref": "#/$defs/finiteNumber", "minimum": 0, "maximum": 1 },
"exposure_score": { "$ref": "#/$defs/finiteNumber", "minimum": 0, "maximum": 1 },
"selected_reason": { "enum": ["cadence", "view_novelty", "keyframe", "user... | openai-build-week | docs/contracts/frame-packet.schema.json | JSON | e720f265161cc107295b367e452071faaab274674ef0004ba5f2ba431f156bfe | 4 | 620 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.