SPEAR

SPEAR Next

SPEAR Next is the Rust/async implementation of SPEAR’s core services:

• SMS: the metadata/control-plane server.
• SPEARlet: the node-side agent/runtime.

Chinese README: [README.zh.md](./README.zh.md)

Repository layout

• src/apps/sms: SMS binary entrypoint
• src/apps/spearlet: SPEARlet binary entrypoint
• web-admin/: Web Admin frontend source
• assets/admin/: built Web Admin static assets embedded/served by SMS
• samples/wasm-c/: C-based WASM samples (WASI)
• docs/: design notes and usage guides

Architecture

Quick start

Prerequisites

• Rust toolchain (latest stable recommended)
• Docker (Docker Desktop on macOS/Windows, or Docker Engine on Linux)
• Docker Compose v2 (docker compose)

This repo uses protoc-bin-vendored, so you typically don’t need to install protoc manually.

Run locally with Docker Compose (SMS + SPEARlet)

This is the recommended cross-platform local setup (no Kubernetes required). It runs:

• SMS (gRPC + HTTP gateway + optional Web Admin)
• SPEARlet (agent/runtime) connecting to SMS via the Compose network

Use the provided Compose file:

• deploy/docker/compose.local.yaml

Start:

docker compose -f deploy/docker/compose.local.yaml up -d --build

Useful endpoints (default host ports):

• SMS health: http://127.0.0.1:18080/health
• SMS Swagger: http://127.0.0.1:18080/swagger-ui/
• SPEAR Console (served by SMS): http://127.0.0.1:18080/console
• SMS Web Admin: http://127.0.0.1:18082/
• SPEARlet health: http://127.0.0.1:18081/health

Stop:

docker compose -f deploy/docker/compose.local.yaml down

Remove local data (volumes):

docker compose -f deploy/docker/compose.local.yaml down -v

Common build/network notes:

• If Docker Hub is not reachable, override base images via environment variables (example mirrors):

export NODE_IMAGE=docker.m.daocloud.io/library/node:20-bookworm-slim
export RUST_IMAGE=docker.m.daocloud.io/library/rust:1.91-bookworm
export DEBIAN_IMAGE=docker.m.daocloud.io/library/debian:trixie-slim
docker compose -f deploy/docker/compose.local.yaml up -d --build

• If you use Local AI Models (llama.cpp), SPEARlet needs llama-server. The Compose file defaults to a build target that includes it. You can override:

SPEARLET_BUILD_TARGET=runtime_with_node_and_llama docker compose -f deploy/docker/compose.local.yaml up -d --build spearlet

Build

make build

# release
make build-release

# build with Rust features (e.g. sled / rocksdb)
make FEATURES=sled build

# enable local microphone capture implementation (optional)
make FEATURES=mic-device build

# macOS shortcut (equivalent to FEATURES+=mic-device)
make mac-build

Run SMS

./target/debug/sms

# enable Web Admin
./target/debug/sms --enable-web-admin --web-admin-addr 127.0.0.1:8081

Useful endpoints:

• HTTP gateway: http://127.0.0.1:8080
• Swagger UI: http://127.0.0.1:8080/swagger-ui/
• OpenAPI spec: http://127.0.0.1:8080/api/openapi.json
• gRPC: 127.0.0.1:50051
• Web Admin (when enabled): http://127.0.0.1:8081/admin

Run SPEARlet

SPEARlet connects to SMS once you provide --sms-grpc-addr (then it auto-registers by default).

./target/debug/spearlet --sms-grpc-addr 127.0.0.1:50051

Configuration

Config file locations

• SMS: ~/.sms/config.toml (or --config )
• SPEARlet: ~/.spear/config.toml (or --config )

Repo-shipped examples:

• SMS: config/sms/config.toml
• SPEARlet: config/spearlet/config.toml

Priority

1. CLI --config file
2. Home config (~/.sms/config.toml or ~/.spear/config.toml)
3. Environment variables (SMS_*, SPEARLET_*)
4. Built-in defaults

Secrets

Do not put secrets into config files. Use spearlet.ai.credentials[].api_key_env to reference environment variables and bind them from backends via credential_ref.

AI backend notes:

• [[spearlet.ai.backends]] hosting is required and must be local or remote.
• credential_ref is optional. If set, the referenced env var must exist (otherwise the backend is filtered). If not set, the backend is treated as “no-auth” (useful for self-hosted proxies).

Ollama discovery

SPEARlet can import models from a local Ollama on startup and materialize them as AI backends.

• Docs: docs/ollama-discovery-en.md

Routing and debugging

• Route by model: if some backends are configured with model = "...", requests can be routed by setting only model (no explicit backend required).
• Observe the selected backend:
• cchat_recv JSON includes a top-level _spear.backend / _spear.model.
• Router emits a router selected backend debug log after selection.

Web Admin

Web Admin provides Nodes/Tasks/Files/AI Models pages.

• AI Models provides an aggregated view across nodes, split into Local/Remote.
• Local AI Models supports creating/deleting model deployments on a node.

Local model provisioning (llamacpp):

• model is just a display key; actual download uses params.model_url when the model file is missing.
• Supported params:
• model_url: http/https URL to a .gguf file (large files are supported).
• download_timeout_s: total download budget in seconds (default: 3600).
• model_path: absolute path, or relative to spearlet.local_models_dir.
• skip_download=1: fail if the model file is missing (no download).

Docs:

• docs/web-admin-overview-en.md
• docs/web-admin-ui-guide-en.md

WASM samples

make samples

Artifacts are written to samples/build/ (WASM-C), samples/build/js/ (WASM-JS), and samples/build/rust/ (WASM-Rust).

Docs:

• docs/samples-build-guide-en.md
• samples/README-en.md
• samples/wasm-js/README-en.md
• samples/wasm-rust/README-en.md
• sdk/rust/crates/spear-boa/README.md
• sdk/rust/crates/spear-wasm-helper/README.md
• docs/spear-console-voice-input-design-en.md

Development

make help
make dev
make ci

UI tests (Playwright):

make test-ui

Documentation

• docs/INDEX.md

License

Apache-2.0. See LICENSE.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: lfedgeai
• Source: lfedgeai/SPEAR
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.