Adding Local AI Generation

.dockerignore

@@ -6,5 +6,5 @@ dist
 .gitignore
 .npmrc
 coverage
-.env
+.env*
 *.log

.github/workflows/update-codejson.yml

@@ -22,12 +22,12 @@ jobs:
 
       - name: Update code.json
         id: generator
-        uses: DSACMS/automated-codejson-generator@main
+        uses: DSACMS/automated-codejson-generator@sachin/local-ai-feature
         with:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          ADMIN_TOKEN: ${{ secrets.ADMIN_PAT }}
           BRANCH: "main"
           SKIP_PR: "false"
+          ENABLE_AI: "true"
 
       - name: Post update information
         run: |

.gitignore

@@ -69,7 +69,7 @@ typings/
 .yarn-integrity
 
 # dotenv environment variables file
-.env
+.env*
 .env.test
 
 # parcel-bundler cache (https://parceljs.org/)

README.md

@@ -155,6 +155,11 @@ SKIP_PR:
 ADMIN_TOKEN:
   description: "Personal Access Token with admin/write privileges for direct push. Required when SKIP_PR is true."
   required: false
+
+ENABLE_AI:
+  description: "Use a local AI model to draft missing fields from the README. Generated fields always go through a pull request."
+  required: false
+  default: "false"
 ```
 
 ### Outputs

action.yml

@@ -1,4 +1,4 @@
-name: Code.json Generator
+name: automated-codejson-generator
 description: "Automatically generates and updates code.json metadata files for CMS repositories"
 author: DSACMS
 
@@ -26,6 +26,11 @@ inputs:
     required: false
     default: "false"
 
+  ENABLE_AI:
+    description: "Use a local AI model to draft missing fields from the README. Generated fields always go through a pull request."
+    required: false
+    default: "false"
+
 outputs:
   updated:
     description: "Boolean indicating whether code.json was updated"
@@ -38,7 +43,7 @@ outputs:
 
 runs:
   using: docker
-  image: Dockerfile
+  image: docker://ghcr.io/dsacms/automated-codejson-generator:v2.0.2
 
 branding:
   icon: "file-text"

code.json

@@ -2,7 +2,7 @@
   "name": "automated-codejson-generator",
   "version": "1.2.1",
   "description": "Tooling that allows for code.json creation through automation",
-  "longDescription": "This project provides a GitHub Action that helps federal agencies maintain their code.json files, which are required for compliance with the Federal Source Code Policy. The action automatically calculates and updates various metadata fields including labor hours, programming languages used, repository information, and timestamps. It creates pull requests with these updates, making it easier to keep code.json files accurate and up-to-date.",
+  "longDescription": "",
   "status": "Production",
   "permissions": {
     "licenses": [
@@ -28,9 +28,9 @@
     "forks": 4,
     "clones": 0
   },
-  "platforms": ["web", "linux"],
-  "categories": ["developer-tools", "automation"],
-  "softwareType": "standalone/backend",
+  "platforms": [],
+  "categories": [],
+  "softwareType": "",
   "languages": ["TypeScript", "JavaScript"],
   "maintenance": "internal",
   "contractNumber": [],
@@ -43,23 +43,15 @@
     "lastModified": "2025-10-21T21:52:17Z",
     "metadataLastUpdated": "2025-10-30T22:53:09.463Z"
   },
-  "tags": [
-    "cmsoss-tier2",
-    "metadata",
-    "codejson",
-    "ci",
-    "github-action",
-    "workflow",
-    "featured"
-  ],
+  "tags": [],
   "contact": {
     "email": "opensource@cms.hhs.gov",
     "name": "CMS Open Source Team"
   },
   "feedbackMechanism": "https://github.com/DSACMS/automated-codejson-generator/issues",
   "AIUseCaseID": "0",
   "localisation": false,
-  "repositoryType": "tools",
+  "repositoryType": "",
   "userInput": false,
   "fismaLevel": "low",
   "group": "CMS/OA/DSAC",

dockerfile

@@ -1,18 +1,73 @@
-FROM node:20-alpine
+ARG MODEL_REPO=unsloth/gemma-4-E2B-it-GGUF
+ARG MODEL_FILE=gemma-4-E2B-it-Q4_K_M.gguf
+ARG MODEL_REVISION=0314792d7f1f7e229411f620751375812bb9faf2
+ARG MODEL_SHA256=740185b21d22ceb83a11c3aa62ad5842ef32c70f6096d756bbee85a1e4ec34b8
 
-RUN apk add --no-cache go git
+# stage 1: builder
+# install deps and build TS
+FROM node:20-bookworm AS builder
 
-RUN go install github.com/boyter/scc/v3@v3.4.0
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        git \
+    && rm -rf /var/lib/apt/lists/*
 
-WORKDIR /action
+ENV NODE_LLAMA_CPP_SKIP_DOWNLOAD=true
 
-COPY package*.json ./
+WORKDIR /app
+
+COPY package.json package-lock.json ./
 RUN npm ci
 
 COPY . .
+RUN npm run package
+
+# stage 2: model 
+# download the model GGUF and cache
+FROM debian:bookworm-slim AS model
+ARG MODEL_REPO
+ARG MODEL_FILE
+ARG MODEL_REVISION
+ARG MODEL_SHA256
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ca-certificates \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN mkdir -p /opt/models \
+    && curl -fSL --retry 3 \
+        "https://huggingface.co/${MODEL_REPO}/resolve/${MODEL_REVISION}/${MODEL_FILE}?download=true" \
+        -o /opt/models/model.gguf \
+    && echo "${MODEL_SHA256}  /opt/models/model.gguf" | sha256sum -c -
+
+# stage 3: runtime 
+# final image the action actually runs 
+FROM node:20-bookworm-slim AS runtime
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        git \
+        ca-certificates \
+        curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# scc used for the laborHours estimate
+RUN curl -fSL --retry 3 \
+        "https://github.com/boyter/scc/releases/download/v3.4.0/scc_Linux_x86_64.tar.gz" \
+        -o /tmp/scc.tar.gz \
+    && echo "8f812d25c237112ea74a4285f0699b7a2c448d70c5e63e370d2d199d38b13eda  /tmp/scc.tar.gz" | sha256sum -c - \
+    && tar -xzf /tmp/scc.tar.gz -C /usr/local/bin scc \
+    && rm /tmp/scc.tar.gz \
+    && chmod +x /usr/local/bin/scc \
+    && scc --version
+
+WORKDIR /app
+
+COPY --from=builder /app/node_modules ./node_modules
+COPY --from=builder /app/dist ./dist
+COPY --from=builder /app/package.json ./package.json
 
-RUN npm run bundle
+COPY --from=model /opt/models/model.gguf /opt/models/model.gguf
 
-ENV PATH="/root/go/bin:${PATH}"
+ENV ACG_MODEL_PATH=/opt/models/model.gguf
 
-ENTRYPOINT ["node", "/action/dist/index.js"]
+ENTRYPOINT ["node", "/app/dist/index.js"]