Add safe BLAKE3 service host/model prerequisites

[Th0rgal]

Summary

• freeze transport-independent, epoch-bound SERVICE_REQUIRED / SERVICE_RESPONSE schemas
• add host compression and executable model adapter prerequisites with bounded infrastructure-only retry
• test malformed messages, wrong binding/digest conflicts, stale/duplicate responses, ABORT/reset, stalls/backpressure, and RETIRE-only atomic commit
• compare 67 deterministic cases against the pinned official blake3_guts low-level compression API
• preserve scatter/gather as a documented executable design contract without changing production SystemVerilog or ASIC/FPGA transport

Scope boundary

Base: exact main commit 2fa58d4d00e007dbd806bf459db77ee643557cf5.

No production SystemVerilog, FPGA/ASIC transport, UART/JTAG, hardware, or publication changes. Compiler/runtime instruction integration remains explicitly unsupported until the logical protocol stabilizes.

Test receipts

• PYTHONDONTWRITEBYTECODE=1 python3 -m unittest -v sim.test_blake3_service — 10 tests, OK
• PYTHONDONTWRITEBYTECODE=1 python3 -m unittest discover -s sim -v — 277 tests, OK
• make check — PASS (includes 277 sim tests and 269 FPGA harness tests, checksum/consistency/smoke and recorded upstream gates)
• make sim — PASS (stream ALU, UART bridge, packet UART bridge, packet frontend adversarial simulation, mutation checks, and 4 RTL differential tests)
• git diff --cached --check before commit — PASS

The official oracle is pinned by tools/blake3_reference/Cargo.lock to blake3_guts 0.0.0, crates.io checksum dda08308cd5393f749c853f8eb805d73e30b8bff2055856dfc24d7a9370a3ae7. Oracle build/execution errors use a distinct infrastructure exception; byte disagreement is a semantic test failure.

---

Note

Low Risk
Host-side protocol scaffolding and tests only; no RTL, transport, or host runtime BLAKE3 instruction path changes. Crypto correctness is backed by differential tests against a pinned reference oracle.

Overview
Adds transport-independent BLAKE3 compression prerequisites at the host/model boundary without wiring compiler/runtime execution or production RTL.

host/blake3_service.py defines 131-byte SERVICE_REQUIRED and 53-byte SERVICE_RESPONSE codecs, v1 payload bridging with a host session_epoch, a Python compress implementation, and ModelServiceAdapter for single-outstanding service keys, operand-identical retries, bounded infrastructure-only retry, and mapping back to v1 service responses.

sim/test_blake3_service.py exercises schema validation, adapter epoch/abort/reset/replay rules, stall/backpressure through the executable endpoint, wrong-binding and digest conflicts, and RETIRE-only commit; compression is differential-tested against a pinned blake3_guts CLI oracle in tools/blake3_reference/.

docs/BLAKE3_EXTERNAL_SERVICE.md freezes the software boundary and future scatter/gather RTL direction; docs/STATUS.md and host/README.md now state BLAKE3 prerequisites are tested while full integration remains unsupported. .gitignore ignores the Cargo target/ tree; SHA256SUMS is refreshed.

^{Reviewed by Cursor Bugbot for commit d4e76aa. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_8ba75a05-b5a4-4b3a-8144-2a2da4e46bfb)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cfbeed1b5c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_c9edfb3a-8160-4be7-a0ef-257d4092c769)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: de996066bc

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_7cffe1da-8323-4d46-a6b0-df333e75e453)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f9aa0b88f8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_a5abc2b9-f08e-4c2f-99f8-efd898a3ceea)