Skip to content

Releases: JunHwan-Kwon/deepbom-gemini

DEEPBOM for Gemini CLI 1.0.0

Choose a tag to compare

@JunHwan-Kwon JunHwan-Kwon released this 18 Sep 08:25

Install DEEPBOM in Gemini CLI to inspect local AI deployment artifacts without executing model code.

gemini extensions install https://github.com/JunHwan-Kwon/deepbom-gemini

Start gemini in your model directory, check /mcp, then use /deepbom:inspect ./model.onnx.

  • Bundled DEEPBOM 1.103.0; no analyzer download at startup.
  • Commands for inspection, comparison, and capabilities.
  • Full SHA-256, serialized structure, quantization evidence, defects, cautions, and evidence gaps through the existing MCP tools.
  • Workspace-scoped access, including symlink checks. Host approval settings remain unchanged.
  • No extension telemetry; local model bytes stay with the analyzer, while returned evidence enters the Gemini conversation.

The ZIP contains gemini-extension.json at its root and includes the engine. Extract it and install the resulting directory for a local installation. Verify the download with SHA256SUMS. Node.js 20+ is required for DEEPBOM, along with a supported Gemini CLI installation and sign-in for conversations.

Validation: real MCP analysis of ONNX and SafeTensors, hash mismatch and outside-workspace rejection, deterministic archive rebuild, and Gemini CLI installation/MCP connection checks. See the cross-platform workflow and validation notes. Signed-in natural-language model responses were not tested.

This is a community Gemini CLI extension. Gemini web/mobile apps and Google endorsement are not claimed. Official gallery discovery depends on Google's crawl and validation.