temple-bridge

Temple Bridge binds two repositories into a sovereign AI stack that runs entirely locally. The Action Layer (BTB) handles commands, file I/O, and test execution, while the Memory/Governance Layer (Threshold & Spiral Protocols) provides persistent memory, governance rules, and recursive reflection to ensure safe, governed execution. Running Hermes-3-Llama-3.1-8B on your machine, the system can perform 8 built-in tools and leverage 3 resources to reason, plan, and act without cloud dependence. In LM Studio, enable MCP to connect to temple-bridge and start a spiral session that moves through phases of perception, analysis, and action with structured tool calls and memory access.

To use the tools, load Hermes-3-Llama-3.1-8B, create or load a System Prompt, and start a new spiral session. The agent can access the BTB repository to execute commands, read/write files, and run tests, while the Threshold memory provides persistent state and governance rules. Expect to see tool availability (8 tools) and resources (3) reflected in the logs when connected. You can initialize as a Spiral Observer, prompting the agent to read its cognitive protocols, list directories (e.g., bt b_list_directory), reflect, and progress through Spiral phases. The workflow emphasizes governance and recursion to avoid naive, unbounded actions and to keep operations local and private.

Prerequisites:

• Mac with Apple Silicon (M1/M2/M3 or later)
• LM Studio v0.3.17+ (or newer)
• Python 3.9+
• uv (Python package manager)

Setup steps:

1. Clone required repositories:

cd ~/Desktop  # or your preferred location
git clone https://github.com/templetwo/back-to-the-basics.git
git clone https://github.com/templetwo/threshold-protocols.git

2. Clone and install temple-bridge:

git clone https://github.com/templetwo/temple-bridge.git
cd temple-bridge
~/.local/bin/uv sync

3. Configure LM Studio (mcp.json is typically created automatically):

# Open LM Studio and verify the config, example shown in README:
cat ~/.lmstudio/mcp.json

4. Download the model in LM Studio:

• In LM Studio: Discover tab -> search for mlx-community/Hermes-3-Llama-3.1-8B-4bit -> download

5. Start the server:

# From temple-bridge directory you can run the server; LM Studio will spawn the MCP via uv

Notes:

• The example mcp.json in the README shows the exact uv-based invocation and environment paths. Adjust paths to your machine if needed.
• Ensure Hermes-3-Llama-3.1-8B-4bit is available in LM Studio before starting.

Tips and notes:

• The system is designed to be fully local with no cloud dependency, leveraging the 8 tools and 3 resources exposed by Temple Bridge.
• Environment variables TEMPLE_BASICS_PATH and TEMPLE_THRESHOLD_PATH should point to the respective local repos (back-to-the-basics and threshold-protocols).
• If LM Studio reports tool or resource counts mismatched, recheck the temple-bridge configuration and ensure the MCP server is running correctly.
• For governance, Spiral Protocols introduce memory and reflective steps before actions; ensure your model and prompt configuration align with the governance expectations.
• If you encounter permission or path issues, verify Python and uv are installed and the paths in mcp.json match your environment. You may need to regenerate the mcp.json if you move the repository.
• Maintaining a clean environment (virtualenv or conda) can help avoid conflicting Python packages when working with uv and the MCP stack.