[bong-water-water-bong]

third_party/laya: 6f0496f797dc -> 0a05a45052e2 (NandhaKishorM/laya main). config/laya.json revision: 7b928d828b7b -> 7b928d828b7b (convaiinnovations/laya).

Before merging: scripts/fetch-laya.sh

(verifies every file) and the Laya tests.

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit dd3dcd4)

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 1 🔵⚪⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ No major issues detected

[github-actions]

Persistent review updated to latest commit d059187

[github-actions]

Persistent review updated to latest commit ad3a5df

[github-actions]

Persistent review updated to latest commit dd3dcd4

[bong-water-water-bong]

Merged after reading the delta, because nothing in the engine consumes third_party/laya:

• grep -rn "third_party/laya" over cmake/, CMakeLists.txt, scripts/, app/, tools/: no hits. The router's
  scorer is built from third_party/ggmlc (cmake/laya-ggml.cmake) and the laya checkpoint comes from
  ~/models/laya-pinned via config/laya.json; the tree builds with third_party/laya empty.
• So the range 6f0496f7..0a05a450 (96 files, +9058/−233 — releases 0.4.14, the Go/JS ports, the
  llamaindex/langchain/crewai integrations, laya/lang.py) moves a tracked reference (the upstream Python/Go SDK),
  not a build input, and cannot change any artifact.
• The pin that does matter, the model revision in config/laya.json (7b928d8), is unchanged by this PR:
  scripts/fetch-laya.sh verified all 21 files of convaiinnovations/laya@7b928d828b7b plus the GGUF.
• Engine side: tests/laya_route_e2e.sh on the candidate build passes — code→hrx (X-1bit-Route: code 0.41 hrx),
  prose→cpu, short→cpu, decision reuse on a second turn, re-decision on a new conversation, clean failure without a
  checkpoint.

CI green (build, CodeQL, pr_agent); hw_gate skipped for a pin-only change.