Peter Steinberger
7c5eaede1e
* feat(codex): pass explicit skill selections as structured turn input Thread OpenClaw-resolved explicit $skill selections through the reply pipeline to the Codex app-server harness and emit them as structured UserInput::Skill turn items when the Codex catalog (skills/list) knows the same SKILL.md path. Codex then injects the skill natively and blocks its text scanner from double-selecting the same name, while unmatched names stay scannable for Codex-native-only skills and the prompt instruction block keeps owning skills Codex cannot see. Catalog misses and RPC errors fail open to the instruction path; the turn never fails on skill lookup. Part of #123367 * fix(plugins): defer llama-cpp ssrf-runtime load out of the setup closure The doctor-contract closure guard forbids the SSRF barrel in the doctor/legacy-setup closure (cold-loads DNS, proxy state, logging); the static import landed with the managed llama-server switch (#123105) and first fails when a PR selects the guard lane. Defer to a dynamic import at download time per the guard's stated remedy.
@openclaw/llama-cpp-provider
Official managed llama.cpp provider for OpenClaw GGUF chat and embeddings.
The plugin installs a pinned, integrity-verified llama-server and configures
OpenClaw's existing localService supervisor. Model traffic uses the normal
OpenAI-compatible chat and embedding transports.
Install
openclaw plugins install @openclaw/llama-cpp-provider
Restart the Gateway after installing or updating the plugin, then choose llama.cpp once during interactive onboarding or configuration.
Configure text inference
After explicit consent, OpenClaw installs the matching server build and downloads Gemma 4 E4B IT Q4_K_M (approximately 5.0 GB) plus EmbeddingGemma (approximately 0.3 GB). The default chat download is offered only on machines with at least 16 GiB of RAM.
Custom GGUF models remain supported through params.modelPath. Rerun llama.cpp
setup after changing the model so OpenClaw can verify the file and regenerate
the managed router preset.
See the llama.cpp provider guide for platform requirements, custom GGUF configuration, diagnostics, and repair.
Configure embeddings
Set memory.search.provider to local. The plugin preserves the historical
local embedding provider and index identity while serving requests through
the managed server's /v1/embeddings endpoint.
Package
- Plugin id:
llama-cpp - Package:
@openclaw/llama-cpp-provider - Minimum OpenClaw host:
2026.6.2