deepbom 1.109.0 → 2.1.0
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 +41 -8
- package/bin/deepbom.mjs +103 -103
- package/package.json +11 -1
- package/pkg/release-manifest.json +4 -4
- package/pkg/tflite_wasm_audit_bg.wasm +0 -0
- package/sdk/index.d.mts +62 -0
- package/sdk/index.mjs +172 -0
- package/skills/deepbom/SKILL.md +1 -1
- package/skills/deepbom/scripts/verify-deepbom.mjs +1 -1
package/README.md
CHANGED
|
@@ -3,14 +3,47 @@
|
|
|
3
3
|
Local deployment-artifact analysis for TFLite, ONNX, GGUF, SafeTensors, Core ML,
|
|
4
4
|
and ExecuTorch.
|
|
5
5
|
|
|
6
|
+
## Programmatic integration
|
|
7
|
+
|
|
8
|
+
The Node/TypeScript SDK is available from version **2.1.0**:
|
|
9
|
+
|
|
10
|
+
```sh
|
|
11
|
+
npm install deepbom@2.1.0
|
|
12
|
+
```
|
|
13
|
+
|
|
14
|
+
Follow the [integration walkthrough](https://github.com/JunHwan-Kwon/deepbom/blob/main/examples/integrations/README.md)
|
|
15
|
+
for runnable exporter, MLflow and release-review examples.
|
|
16
|
+
|
|
17
|
+
```js
|
|
18
|
+
import { inspect, audit, DeepBomPolicyBlocked } from "deepbom";
|
|
19
|
+
|
|
20
|
+
const summary = await inspect("model.onnx");
|
|
21
|
+
console.log(summary.artifact.sha256, summary.coverage);
|
|
22
|
+
try {
|
|
23
|
+
const evidence = await audit("model.onnx", { gate: "defects" });
|
|
24
|
+
// Store this native evidence; apply deployment policy separately.
|
|
25
|
+
} catch (error) {
|
|
26
|
+
if (!(error instanceof DeepBomPolicyBlocked)) throw error;
|
|
27
|
+
console.log(error.code, error.document);
|
|
28
|
+
}
|
|
29
|
+
```
|
|
30
|
+
|
|
31
|
+
This is an asynchronous ESM API for Node >=20 that invokes the bundled CLI
|
|
32
|
+
without a shell or engine download. It reuses the common analysis engine and
|
|
33
|
+
ships TypeScript declarations. Python uses the same engine in its wheel.
|
|
34
|
+
See the [SDK contract](https://github.com/JunHwan-Kwon/deepbom/blob/main/docs/SDK_CONTRACT.md)
|
|
35
|
+
for supported options, structured errors and independent IR versioning.
|
|
36
|
+
|
|
37
|
+
## Command line
|
|
38
|
+
|
|
6
39
|
For a local Codex or Claude Code project, preview and install the bundled Agent
|
|
7
40
|
Skill without operating an analysis server:
|
|
8
41
|
|
|
9
42
|
```console
|
|
10
|
-
npx -y deepbom@1.
|
|
11
|
-
npx -y deepbom@1.
|
|
12
|
-
npx -y deepbom@1.
|
|
13
|
-
npx -y deepbom@1.
|
|
43
|
+
npx -y deepbom@2.1.0 integrate codex
|
|
44
|
+
npx -y deepbom@2.1.0 integrate codex --apply
|
|
45
|
+
npx -y deepbom@2.1.0 integrate claude-code
|
|
46
|
+
npx -y deepbom@2.1.0 integrate claude-code --apply
|
|
14
47
|
```
|
|
15
48
|
|
|
16
49
|
```console
|
|
@@ -28,10 +61,10 @@ npx deepbom explore model.tflite --target-profile target-profile.json
|
|
|
28
61
|
npx deepbom audit model.pte --executorch-build deepbom.executorch-build.json --compact
|
|
29
62
|
npx deepbom capabilities --format agent-json
|
|
30
63
|
npx deepbom capabilities --format agent-text
|
|
31
|
-
npx -y deepbom@1.
|
|
64
|
+
npx -y deepbom@2.1.0 mcp
|
|
32
65
|
```
|
|
33
66
|
|
|
34
|
-
Claude Desktop can install the version-matched `deepbom-1.
|
|
67
|
+
Claude Desktop can install the version-matched `deepbom-2.1.0.mcpb` asset
|
|
35
68
|
from the corresponding GitHub Release as a local desktop extension.
|
|
36
69
|
|
|
37
70
|
ChatGPT developer-mode users can connect `https://deepbom.org/mcp` for one
|
|
@@ -90,6 +123,6 @@ and private rulepack-generation sources are not included.
|
|
|
90
123
|
|
|
91
124
|
## Optional metadata and OMOP connections
|
|
92
125
|
|
|
93
|
-
After inspecting a model, create a model-bound input with `deepbom audit model.onnx --metadata-template omop -o metadata.json`. Fill the selected OMOP 5.4/5.5 `CDM_SOURCE` row and institution/release identifiers, then use `--metadata metadata.json --evidence-files ./evidence --section
|
|
126
|
+
After inspecting a model, create a model-bound input with `deepbom audit model.onnx --metadata-template omop -o metadata.json`. Fill the selected OMOP 5.4/5.5 `CDM_SOURCE` row and institution/release identifiers, then use `--metadata metadata.json --evidence-files ./evidence --section provenance_ir --json`. A generic template connects non-OMOP datasets, cohorts, code, runs and reports through the same IR.
|
|
94
127
|
|
|
95
|
-
The Web **
|
|
128
|
+
The Web **Provenance** workspace provides import, record creation, an interactive relationship diagram and JSON/CycloneDX downloads. Local MCP `deepbom_audit` accepts `metadata_template`, `metadata` and `evidence_files`; ChatGPT/Claude widgets share only explicitly requested counts and digests. Connections remain declarations: file hashes and reference consistency do not authenticate lineage or clinical suitability. See the [OMOP guide](https://deepbom.org/guides/omop-metadata/) and [common IR contract](https://github.com/JunHwan-Kwon/deepbom/blob/main/docs/PROVENANCE_IR.md).
|