std-context-lab

This repository is an isolated sandbox to battle-test the context-pipe and semantic-sift tools.

🚀 Agnostic Setup Instructions

To initialize this lab on your local machine, follow these steps. This setup follows the Sovereign Dual-Repo Pattern.

1. Prerequisites

• uv (Python package manager)
• Rust / Cargo (For native sifting)
• Node.js / npm (For shadow MCPs)

2. Environment Initialization

Create a shared virtual environment and install the target repositories in editable mode:

# Create venv
uv venv

# Install target repos in editable mode with neural extras
uv pip install -e "target_repos/context-pipe"
uv pip install -e "target_repos/semantic-sift[neural]"

# Install common lab utilities
uv pip install yq markitdown

3. Compile Native Sidecar

The heuristic refinery requires the compiled Rust core:

cd target_repos/semantic-sift/crates/sift-core
cargo build --release

4. Install Shadow MCPs

cd scenarios/shared_mcps
npm install

5. IDE Configuration

This lab is environment-agnostic. To use the tools:

• MCP Registration: Point your AI assistant (Gemini CLI, Cursor, VS Code) to the context_pipe.server and semantic_sift.server modules using the Python interpreter in .venv/Scripts/python.
• Hooks: For transparent sifting, register target_repos/context-pipe/pipe_hook.py as a post-tool execution hook in your IDE settings.

🖥️ IDE/Environment Onboarding SOP

When opening this lab in a new IDE (Cursor, VSCode, etc.), follow this protocol to ensure full feature parity:

1. Set the Interpreter: Select the .venv at the repository root as your IDE's Python interpreter.
2. Register MCPs: Register the context-pipe and semantic-sift servers in your IDE settings using the venv python path and the -m <module>.server arguments.
3. Run Onboarding: Ask your AI assistant to run pipe_onboard(environment="NAME"). This automates hook injection for your specific platform.
4. Verify Health: Run pipe_verify() to confirm core tools are connected and the refinery is responding.
5. Update Matrix: Open LAB_STATUS.md and mark Scenario 01 as verified for your current IDE once the basic pipe run succeeds.

🧪 Testing Scenarios

All 17 technical claims and battle-tested use cases are consolidated in SCENARIOS.md.

Detailed setup scripts and artifacts for each individual experiment are located in the /scenarios directory. Each scenario folder also includes its own README.md containing deep technical execution steps and localized findings.

🔄 Regression Testing

When a new version of the upstream tools is released, we follow a strict re-verification protocol. See the Regression Testing SOP for details on how to test updates against the "Core Four" scenarios to maintain the LAB_STATUS.md baseline.

🐞 Bug Reporting

Found a bug? Do not patch the target_repos/. Instead, create a standardized report in the /bugs directory following the protocol in AGENTS.md.