Skip to content

AGENTS.md

Latest commit

 

History

History
75 lines (53 loc) · 2.24 KB

AGENTS.md

File metadata and controls

75 lines (53 loc) · 2.24 KB

AGENTS.md

Guidance for coding agents working in llama-web-bridge.

Scope and Ownership

  • This repository owns WebGPU bridge source/runtime build for llama.cpp web.
  • It publishes versioned assets to llama-web-bridge-assets via workflow.
  • llamadart consumes those published assets.

Related Repositories

Common maintainer sibling layout:

../llamadart
../llamadart-native
../llama-web-bridge
../llama-web-bridge-assets

Build Commands

./scripts/build_bridge.sh

Useful environment overrides:

  • LLAMA_CPP_DIR
  • BUILD_DIR
  • OUT_DIR
  • CMAKE_BUILD_TYPE

Local Verification Notes

When validating bridge runtime changes locally, keep build/cache output outside the repo so generated wasm artifacts and toolchain caches do not dirty the checkout or hit sandboxed Homebrew/cache paths:

export CCACHE_DIR=/private/tmp/llama_web_bridge_ccache
export EM_CACHE=/private/tmp/llama_web_bridge_emcache
BUILD_DIR=/private/tmp/llama_web_bridge_build MEM64_BUILD_DIR=/private/tmp/llama_web_bridge_build_mem64 OUT_DIR=/private/tmp/llama_web_bridge_dist WEBGPU_BRIDGE_BUILD_MEM64=1 ./scripts/build_bridge.sh

CI / Release

  • CI build gate: .github/workflows/ci.yml
  • Publish workflow: .github/workflows/publish_assets.yml
    • Requires WEBGPU_BRIDGE_ASSETS_PAT
    • Pushes assets + tag to llama-web-bridge-assets

Change Boundaries

  • Keep runtime bridge code in js/ and src/.
  • Keep publishing logic in workflow only.
  • Do not edit assets repository files from here outside publish flow.

Cross-Repo Handoff to llamadart

After publishing assets tag:

  1. Update/fetch pinned bridge assets in llamadart: WEBGPU_BRIDGE_ASSETS_TAG=<tag> ./scripts/fetch_webgpu_bridge_assets.sh
  2. Update docs/changelog in llamadart if behavior changed.

Regression Smoke Guidance

  • For pthread/runtime changes, test a BERT-class embedding model in Chromium with cross-origin isolation enabled. The regression shape is: loadModelFromUrl, tokenize, embed, and embedBatch on a host where navigator.hardwareConcurrency is greater than the bridge pthread pool size.
  • Run the smoke through both direct runtime (disableWorker: true) and the bridge worker path; both should report n_threads capped to the pool size.