deepbom 1.96.11 → 1.96.13

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -3,6 +3,16 @@
3
3
  Local deployment-artifact analysis for TFLite, ONNX, GGUF, SafeTensors, Core ML,
4
4
  and ExecuTorch.
5
5
 
6
+ For a local Codex or Claude Code project, preview and install the bundled Agent
7
+ Skill without operating an analysis server:
8
+
9
+ ```console
10
+ npx -y deepbom@1.96.13 integrate codex
11
+ npx -y deepbom@1.96.13 integrate codex --apply
12
+ npx -y deepbom@1.96.13 integrate claude-code
13
+ npx -y deepbom@1.96.13 integrate claude-code --apply
14
+ ```
15
+
6
16
  ```console
7
17
  npx deepbom audit model.onnx --format cyclonedx
8
18
  npx deepbom audit model.onnx --format sarif --output deepbom.sarif --fail-on high
@@ -14,8 +24,13 @@ npx deepbom verify model.tflite --contract production-interface.json
14
24
  npx deepbom diff baseline.tflite candidate.tflite
15
25
  npx deepbom explore model.tflite --target-profile target-profile.json
16
26
  npx deepbom audit model.pte --executorch-build deepbom.executorch-build.json --compact
27
+ npx deepbom capabilities --format agent-json
28
+ npx -y deepbom@1.96.13 mcp
17
29
  ```
18
30
 
31
+ Claude Desktop can install the version-matched `deepbom-1.96.13.mcpb` asset
32
+ from the corresponding GitHub Release as a local desktop extension.
33
+
19
34
  The default output is a bounded human-readable summary. Use `--json` or
20
35
  `--compact` for the complete analysis document, `--format envelope` for the
21
36
  canonical cross-format contract, `--format cyclonedx` for CycloneDX 1.7, or
@@ -28,7 +43,9 @@ WASM Pareto candidates. These commands do not add a second analysis engine.
28
43
 
29
44
  The package contains one generated JavaScript analysis bundle and the canonical
30
45
  TFLite WebAssembly module. It does not send model bytes or results over the
31
- network. TensorRT parser observations and build profiles can be imported with
46
+ network, and it does not depend on a hosted DEEPBOM analysis endpoint. A plain
47
+ chat client without local shell or MCP access cannot execute an audit.
48
+ TensorRT parser observations and build profiles can be imported with
32
49
  `--tensorrt-parser-evidence` and `--tensorrt-profile`.
33
50
 
34
51
  ONNX external data next to the model is discovered only from safe serialized