feat: GitOps workflow for WebSearch ADK Agent

.github/actions/run-adk-agent/README.md

@@ -0,0 +1,47 @@
+# Run ADK Agent
+
+A reusable GitHub Action that runs `gradient agent run` to validate an ADK agent starts and is available at http://localhost:8080.
+
+## Usage
+
+```yaml
+- uses: actions/checkout@v4
+- uses: ./.github/actions/run-adk-agent
+  with:
+    working-directory: WebSearch
+    digitalocean-inference-key: ${{ secrets.DIGITALOCEAN_INFERENCE_KEY }}
+```
+
+## Inputs
+
+| Input | Required | Default | Description |
+|-------|----------|---------|-------------|
+| `working-directory` | Yes | - | Path to agent directory (e.g., WebSearch, StateGraph, RAG) |
+| `digitalocean-inference-key` | Yes | - | DigitalOcean GenAI inference key for LLM calls |
+| `python-version` | No | `3.10` | Python version |
+| `openai-api-key` | No | - | OpenAI API key (for RAG embeddings; not needed for WebSearch) |
+
+## Secrets Setup
+
+Add the required secret in your GitHub repo before the workflow can run:
+
+1. Go to your repo on GitHub → **Settings** → **Secrets and variables** → **Actions**
+2. Click **New repository secret**
+3. Name: `DIGITALOCEAN_INFERENCE_KEY`
+4. Value: Your DigitalOcean GenAI inference key (from [GenAI Settings](https://cloud.digitalocean.com/gen-ai))
+5. Click **Add secret**
+
+The secret is stored encrypted by GitHub and never shown in logs or to contributors.
+
+## Secrets by Template
+
+| Template | Required | Optional |
+|----------|----------|----------|
+| WebSearch | `DIGITALOCEAN_INFERENCE_KEY` | - |
+| StateGraph | `DIGITALOCEAN_INFERENCE_KEY` | - |
+| RAG | `DIGITALOCEAN_INFERENCE_KEY` | `OPENAI_API_KEY` |
+| Others | See template `.env.example` | - |
+
+## GitHub Enterprise
+
+Works on GitHub.com and GitHub Enterprise Server. Uses standard `actions/checkout` and `actions/setup-python`.

.github/actions/run-adk-agent/action.yml

@@ -0,0 +1,39 @@
+name: 'Run ADK Agent'
+description: 'Run gradient agent run to validate agent starts and is available at http://localhost:8080'
+inputs:
+  working-directory:
+    required: true
+    description: 'Path to agent directory (e.g., WebSearch, StateGraph)'
+  digitalocean-inference-key:
+    required: true
+    description: 'DigitalOcean inference key (for LLM)'
+  python-version:
+    required: false
+    default: '3.10'
+  openai-api-key:
+    required: false
+    description: 'OpenAI API key (for RAG embeddings; not needed for WebSearch)'
+runs:
+  using: composite
+  steps:
+    - uses: actions/setup-python@v5
+      with:
+        python-version: ${{ inputs.python-version }}
+    - run: pip install gradient-adk
+      shell: bash
+    - run: pip install -r requirements.txt
+      working-directory: ${{ inputs.working-directory }}
+      shell: bash
+    - run: |
+        gradient agent run &
+        for i in $(seq 1 30); do
+          curl -s -o /dev/null --connect-timeout 2 http://127.0.0.1:8080/ && { echo "Port 8080 is listening"; exit 0; }
+          sleep 2
+        done
+        echo "Agent failed to start within 60 seconds"
+        exit 1
+      working-directory: ${{ inputs.working-directory }}
+      shell: bash
+      env:
+        DIGITALOCEAN_INFERENCE_KEY: ${{ inputs.digitalocean-inference-key }}
+        OPENAI_API_KEY: ${{ inputs.openai-api-key }}

.github/workflows/validate-websearch.yml

@@ -0,0 +1,13 @@
+on:
+  pull_request:
+    paths: ['WebSearch/**']
+  workflow_dispatch:
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ./.github/actions/run-adk-agent
+        with:
+          working-directory: WebSearch
+          digitalocean-inference-key: ${{ secrets.DIGITALOCEAN_INFERENCE_KEY }}

WebSearch/main.py

@@ -1,9 +1,10 @@
 import os
 from gradient_adk import entrypoint
 from langchain_core.tools import tool
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_gradient import ChatGradient
 from langchain_core.messages import HumanMessage
-from langchain.agents import create_agent
+from langgraph.prebuilt import create_react_agent
 from pydantic import BaseModel
 
 from langchain_community.tools import DuckDuckGoSearchRun
@@ -25,9 +26,12 @@ def web_search(query: str) -> str:
     model="openai-gpt-oss-120b",
 )
 
-agent = create_agent(
-    llm, tools=[web_search], system_prompt=SYSTEM_PROMPT
-)
+prompt = ChatPromptTemplate.from_messages([
+    ("system", SYSTEM_PROMPT),
+    ("placeholder", "{messages}"),
+])
+
+agent = create_react_agent(model=llm, tools=[web_search], prompt=prompt)
 
 
 class Message(BaseModel):

WebSearch/requirements.txt

@@ -1,8 +1,8 @@
 gradient-adk
-langchain==1.0.0
-langchain-core==1.2.11
-langchain-community
-langgraph==1.0.0
-langchain-gradient>=0.1.0
+langchain>=0.3.0,<0.4.0
+langchain-core>=0.3.15,<0.4.0
+langchain-community>=0.3.0
+langgraph
+langchain-gradient
 pydantic>=2.6,<3
 ddgs