Dated source reference. Current release status: handoff state. Full original relative paths are in the developer archive.
Developer integration work64
Current work70: private API70e/API55 passes634 Windows tests/126 schemas; unchanged native70d retains470-check original/college adapter evidence. Original70liveo passes natural9–6 and college70liveq passes6–3, including copied native metadata, selected IFF reads, classification, independent immutable results and repeated ACK with zero faults. The college startup decoder fix is now game-verified. Both runs restore98 files/DLL50/INI; final preservation passes and custody is released for arena r11. Changed FT-count live verification remains limited to the recorded observations; no public/stable promotion. Earlier checkpoints retain their source scope. Current ledger.
Work70 updates only the frozen-rule schemas/tests and generated SDK types over runtime64f. Current API70b evidence records629 Windows API tests,126 canonical schemas/50 operations and freshly compiled strict TypeScript/25 runtime cases. Broker55 runtime logic and the eight previously accepted simulated scenarios are unchanged; those scenarios retain64f scope. The source identities below describe the earlier developer-kit checkpoint.
The user authorized three headless additions on October19,2026: a unified compatibility report, a typed TypeScript client, and game-free integration scenarios. These consume the public HTTP contract and keep the adapter independent of any first-party application. No new GUI, public package/deployment or game launch is part of this work.
Work63's API54/native63i candidate and original/college live checks remain pending. The user will explicitly say when the Dell is safe; do not install a hook, register a game task or start OpenHoops until then. MVP and protected saves, binaries, hosting and historical evidence remain untouched. Builds/tests run on the Dell; the Mac edits and stores sources. No agents are authorized.
Implemented outcomes:
- Combine request, ROS, participant/reference, metadata, package and hook checks into one bounded, read-only report with stable check IDs, actionable fixes and explicit unknowns. A static pass does not certify rendering or live readiness. Never reserve a request ID, change a game or register inputs during preflight.
- Generate strict TypeScript types and typed operations from the maintained OpenAPI schema; implement authenticated JSON/binary/SSE transport, cancellation, structured failures and safe durable-result consumption. Preserve nullable uncertain outcomes. Do not retry mutating operations automatically or change IDs after an ambiguous response. Respect existing loopback authentication and origin restrictions.
- Supply deterministic scenarios through the production broker using explicitly simulated game inputs, separate private state and bounded loopback servers. Cover success, startup refusal/uncertainty, stale commands, disconnection/recovery, abandoned/review-required matches and independent result ACKs. Mark simulation in API/session/result provenance. No fabricated engine verification.
Acceptance requires exact staged source identities, Windows Python regression, strict TypeScript compilation and runtime tests, actual HTTP scenario tests, schema/type drift checks, documentation and preservation. Keep every failed stage and distinguish simulation, headless instruction checks and live game evidence.
Accepted headless runtime
Private broker0.2.0.dev55/HTTPv1, mainABI26/event4/profile4/court3. Runtime source devkit64f, archive95adee1517752258ee23f73e9034f3260bd264b14419cfd5649ffe0e4228fd84; manifest. No native adapter is rebuilt or installed. The88 native code inputs and eight native/request protocol inputs are identical to native63i/DLL56e66f24. The current broker adds a read-only route, developer entry points/capability and proper decoding of encoded path IDs such as season:game; mutable game handlers are unchanged.
| Evidence | Accepted scope |
|---|---|
| API regression | 628 Windows tests, zero skips;33 new compatibility/scenario cases |
| Source and schemas | 125 canonical schemas/50 operations, all staged hashes,88 native/eight protocol matches |
| TypeScript | TypeScript7.0.2 strict compilation, eight negative type cases and25 runtime cases on Windows/Node24.22.0 |
| Compiler identity | Six compilation inputs and generated artifacts match accepted64c; scenario-only fixes do not alter the client |
| Independent client HTTP | Eight cases/63 checks/18 schema responses: actual Node client, broker HTTP, file uploads, SSE/cancellation, SQLite inbox, lost command/ACK responses and independent consumers |
| Preservation | 95 originals in both copies,98 isolated files, protected binaries/packages/four hosts and1,425 compacted historical files verify;6,062,755,840bytes free at closeout |
Preflight uses temporary scratch intake/authoring rather than real state registration. Reports retain static scope and unknowns, even when all available checks pass. Game-free sources use integration_scenario; result metadata explicitly marks simulation. The production server has no simulator controls. Scenario resume refuses unrelated broker directories. Native game verification is not inferred from any case in this table.
Preserved failures:64a omitted the required pre-tip roster/name binding guards and initially tried to pass raw JSON metadata bytes through an object-only TypeScript signature.64c's first actual HTTP schema run exposed incomplete synthetic clock parameters;64d then caught raw native numeric launch codes being used in public string fields. The corrected64e fixture passes both standalone schema tests and independent HTTP validation. Earlier failed snapshots/logs remain separate;625 passing64c API tests were not sufficient to establish its scenario schema acceptance.
The public editor/Windows downloads, stable35/restored50,95 originals in both games,98 isolated files and all hosts remain protected. Work63 is still waiting for the user's explicit Dell-safe reply. These developer tools can be used headlessly now; later native integration must intentionally stage the matching current broker and adapter rather than silently selecting the installed older API.
The final64f review keeps stored roster membership separate from unverified native instantiation order. Differing planned slot order is unknown, and repeated names without optional metadata are unknown source-row associations rather than invalid external identities. A companion still requires explicit rows when names are ambiguous. Two added regressions bring the full suite628; actual pre-play name/readback guards remain mandatory.
Aggregate acceptance records all three developer tools as complete within this scope. Storage notes document transparent compression of closed historical captures with before/after hashes, preserved paths and timestamp journals. No MVP file/process/task, live save, installed adapter or cloud resource was modified. Shared free space is volatile; future tests must recheck the3GiB floor. No public package was created.