Peter Steinberger
f65a6f81de
The managed llama-server default ctx-size was 8192, but the full OpenClaw agent system prompt alone is ~31K tokens, so the first agent turn overflowed the context window and forced immediate compaction (observed live on the Mac app local-model onboarding). Raise the default to 65536 so a fresh local-model install can run a real agent turn out of the box. The default-download 16 GiB RAM floor already bounds weaker machines, and Gemma 4 supports far more than 64K, so this only changes headroom, not the offer gate. Docs updated to match.
@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