LocalClaw

LocalClaw is a Pinokio package that adapts OpenClaw for strictly local model usage with:

• Ollama (http://127.0.0.1:11434/v1)
• LM Studio (http://127.0.0.1:1234/v1)

No cloud endpoints are configured by default.

What this package adds

• A local-only config renderer (scripts/render_local_config.py) that:
  • rejects non-local hosts,
  • seeds OpenClaw with OpenAI-compatible local model definitions,
  • writes config to .localclaw/openclaw.local.json.
• Auto backend detection (scripts/start_localclaw.sh) for Ollama vs LM Studio.
• A reproducible smoke test (scripts/smoke_test_local.sh) using a mock OpenAI-compatible server (scripts/mock_openai_server.py) that auto-responds with LOCAL_TEST_PHRASE.
• Pinokio menu entries for Install / Start / Test / Update / Uninstall.

Pinokio usage

1. Click Install.
2. Start either:
   • Ollama with an OpenAI-compatible endpoint (/v1), or
   • LM Studio local server.
3. Click Start (Auto detect Ollama/LM Studio).
4. Optionally run Local API smoke test.

Runtime overrides

You can change behavior with environment variables:

• LOCALCLAW_BACKEND=auto|ollama|lmstudio
• LOCALCLAW_BASE_URL=http://127.0.0.1:11434/v1
• LOCALCLAW_MODEL=llama3.2:latest
• LOCALCLAW_GATEWAY_PORT=18789

Failure modes & mitigations

1) OpenClaw reports unknown model

OpenClaw validates model IDs against configured provider models.

Mitigation: set LOCALCLAW_MODEL=<your-local-model-id> and rerun Start. The config renderer injects that model ID into provider definitions.

2) Local server not detected in auto mode

Auto mode checks 11434 then 1234.

Mitigation: set LOCALCLAW_BACKEND explicitly or set LOCALCLAW_BASE_URL directly.

3) Accidental remote endpoint usage

Strict local mode blocks non-local hosts.

Mitigation: only 127.0.0.1, localhost, and ::1 are allowed by config generation.

4) Endpoint compatibility differences

Some OpenAI-compatible servers support /v1/chat/completions only.

Mitigation: LocalClaw config uses openai-completions API mode for compatibility.

Community dashboard visibility (Pinokio/GitHub)

To maximize visibility when published:

• Keep pinokio.js, pinokio.json, and icon.svg in the repository root.
• Use a clear title + description in pinokio.json.
• Include a concise README and screenshots if/when UI behavior changes.
• Tag the GitHub repo clearly (pinokio, openclaw, ollama, lmstudio, local-llm) and provide release notes so community users understand what is local-only.

Notes on OpenClaw usage patterns and added gaps

OpenClaw users commonly run multi-channel automations and cloud providers by default. This package fills local-first gaps by adding:

• hardened localhost-only guardrails,
• backend auto-detection for common local inference runtimes,
• deterministic API emulation tests for safer packaging/sharing.