August 21 Bug Fixes and Updates

Author: jjfrost

.devcontainer/devcontainer.json

@@ -9,9 +9,8 @@
         "ghcr.io/devcontainers/features/node:1": {
             "version": "22.14"
         },
-        "ghcr.io/devcontainers/features/azure-cli:1": {
-            "installBicep": "true",
-            "extensions": "rdbms-connect"
+         "ghcr.io/devcontainers/features/azure-cli:1": {
+            "installBicep": "true"
         },
         "ghcr.io/azure/azure-dev/azd:latest": {},
         "ghcr.io/rchaganti/vsc-devcontainer-features/azurebicep:1": {},
@@ -78,6 +77,5 @@
             ]
         }
     },
-    "postCreateCommand": "bash ./.devcontainer/entrypoint.sh",
-    "postStopCommand": "docker stop arize-phoenix || true"
+    "postCreateCommand": "bash ./.devcontainer/entrypoint.sh"
 }

.devcontainer/docker-compose.yml

@@ -11,25 +11,7 @@ services:
     networks:
       - agentic-net
 
-  arize-phoenix:
-    container_name: arize-phoenix
-    profiles: ["tracing"]
-    build:
-      context: ../arize-phoenix
-      dockerfile: Dockerfile
-    ports:
-      - "6006:6006"
-      - "4317:4317"
-    volumes:
-      - phoenix_data:/root/.phoenix/
-    networks:
-      - agentic-net
-
 networks:
   agentic-net:
     name: agentic-net
     driver: bridge
-
-volumes:
-  phoenix_data:
-    name: phoenix_data

.vscode/scripts/start-arize.sh

@@ -2,13 +2,26 @@
   "version": "2.0.0",
   "tasks": [
     {
-        "label": "start-arize",
-        "type": "shell",
-        "command": "bash",
-        "args": [
-            ".vscode/scripts/start-arize.sh"
+      "label": "start-arize",
+      "type": "shell",
+      "command": "${workspaceFolder}/backend/.venv/bin/python",
+      "args": ["-m", "phoenix.server.main", "serve"],
+      "isBackground": true,
+      "problemMatcher": {
+        "pattern": [
+          {
+            "regexp": ".*",
+            "file": 0,
+            "message": 0,
+            "line": 0
+          }
         ],
-        "problemMatcher": []
+        "background": {
+          "activeOnStart": true,
+          "beginsPattern": ".",
+          "endsPattern": "."
+        }
+      }
     }
   ]
 }

.vscode/tasks.json

@@ -14,7 +14,6 @@ Explore how intelligent agents handle search, sentiment, memory, and personaliza
 - [Tool-Enabled Query Routing](#-smart-query-routing-via-tool-enabled-agent)
 - [Understanding What Queries Work](#-understanding-what-queries-work-and-why)
 
-
 ## App Scenario: Personalized E-Commerce with GenAI Agents
 
 **AgenticShop** is a GenAI-powered e-commerce demo that showcases how multi-agent systems and in-database intelligence can deliver personalized shopping experiences.
@@ -71,6 +70,7 @@ The following image shows the high-level architecture of the solution, highlight
 ### 🏗️ Infrastructure Summary
 
 The solution is deployed entirely within a single **Azure Resource Group** and uses the following core infrastructure:
+
 - **Azure Container Apps Environment**
 - **Azure OpenAI**
 - **Azure Flexible PostgreSQL Server**
@@ -80,83 +80,101 @@ The solution is deployed entirely within a single **Azure Resource Group** and u
 ## 🚀 Solution Accelerator Deployment
 
 ### 🧰 Prerequisites
+
 The following are prerequisites for deploying this solution:
-1. [Azure Developer CLI](https://learn.microsoft.com/en-us/azure/developer/azure-developer-cli/install-azd?tabs=winget-windows%2Cbrew-mac%2Cscript-linux&pivots=os-linux)
-2. [Azure CLI](https://learn.microsoft.com/en-us/cli/azure/install-azure-cli)
-3. [Azure CLI extension](https://learn.microsoft.com/en-us/cli/azure/azure-cli-extensions-overview) `rdbms-connect`
-4. An Azure account with an active subscription.
-5. [Powershell Core](https://learn.microsoft.com/en-us/powershell/scripting/install/installing-powershell?view=powershell-7.5)
-6. Appropriate roles attached to user for solution deployment (`Contributor` role and `Role Based Admin Control Administrator` role for the subscription)
-7. [Git](https://git-scm.com/)
+
+1. [Azure Developer CLI](https://learn.microsoft.com/azure/developer/azure-developer-cli/install-azd?tabs=winget-windows%2Cbrew-mac%2Cscript-linux&pivots=os-linux)
+2. [Azure CLI](https://learn.microsoft.com/cli/azure/install-azure-cli)
+3. An Azure account with an active subscription.
+4. [Powershell Core](https://learn.microsoft.com/powershell/scripting/install/installing-powershell?view=powershell-7.5)
+5. Appropriate roles attached to user for solution deployment (`Contributor` role and `Role Based Admin Control Administrator` role for the subscription)
+6. [Git](https://git-scm.com/)
 
 ### 🛠️ Deployment Steps
 
 #### Clone the Repository
+
 Clone the repository. Once done, navigate to the repository:
+
 ```sh
 git clone https://github.com/Azure-Samples/postgres-agentic-shop.git
 cd postgres-agentic-shop
 ```
 
 #### Log in to your Azure account
+
 To log in to Azure CLI, use the following command. You can use the `--use-device-code` flag if the command fails.
+
 ```sh
 az login
 ```
+
 To log in to Azure Developer CLI, use this command. You can use the `--use-device-code` flag if the command fails.
+
 ```sh
 azd auth login
 ```
 
 #### Create a new Azure Developer environment
+
 In the root of the project, execute the following command to create a new `azd` environment. Provide a name for your `azd` environment:
+
 ```sh
 azd env new
 ```
 
-#### Grant permissions to azd hook scripts
-If you are deploying the solution from **Windows**, run the following command to grant permissions to the current session to execute `pwsh` scripts located in the `azd-hooks` directory:
-```sh
+#### **Windows Users Only – Grant permissions to azd hook scripts**
+
+> **⚠️ IMPORTANT:** This step is **only** required if you are deploying from **Windows**.
+> **Mac** and **Linux** users can skip this — nothing needs to be done.
+
+If you are on **Windows**, run the following command in your current terminal session to allow execution of `pwsh` scripts located in the `azd-hooks` directory:
+
+```powershell
 Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
-```
 
-If you are deploying the solution from **Unix-like environment on Windows OS** (for instance Cygwin, MinGW), grant the following permissions for execution of scripts in the `azd-hooks` directory:
-```sh
-pwsh -NoProfile -Command "Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy Bypass"
 ```
 
 #### Solution Deployment
-**NOTE** that this solution uses postgres authentication for creating connection to database. It is recommended to change the `ADMINISTRATOR_LOGIN_USER` and `ADMINISTRATOR_LOGIN_PASSWORD` parameters in `main.parameters.json` file in `infra` directory before deployment as best security practice.
 
-1. Run the following command to provision the resources.
-```sh
-azd up
-```
-Once the above command is executed, the `azd` workflow prompts user to select the subscription for deployment, two locations (one for the solution accelerator resources and another for Azure OpenAI models), and the resource group to create.
+> **NOTE:** This solution uses PostgreSQL username/password authentication. During deployment, the admin login and password are **autogenerated** and written to your project’s `.env` file.
+> Keep your `.env` file out of version control.
+
+1. Run the following command to provision the resources:
+
+    ```sh
+    azd up
+    ```
+
+    Once the above command is executed, the `azd` workflow prompts user to select the subscription for deployment, location and the resource group to create.
 
 2. Make sure that you have enough Azure OpenAI model quota in the region of deployment. **The `azd` workflow automatically filters and shows the region where the Azure OpenAI quota is available.** The Azure OpenAI quota required for `GlobalStandard` **deployment type** for this solution is listed below. This configuration can be changed from the `main.parameters.json` file in the `infra` directory using the following parameters:
+
     - **`GlobalStandard` GPT-4o:** 50K TPM - `AZURE_OPENAI_CHAT_DEPLOYMENT_CAPACITY`
     - **`GlobalStandard` text-embedding-3-small:** 70K TPM - `AZURE_OPENAI_EMBED_DEPLOYMENT_CAPACITY`
-If you have changed the above parameters from the `main.parameters.json` file, you **must change** the following configuration in `main.bicep` file so that the changes are reflected in automatic Azure OpenAI region filtering as well:
-```sh
-@metadata({
-  azd: {
-    type: 'location'
-    usageName : [
-      'OpenAI.GlobalStandard.gpt-4o, 50'
-      'OpenAI.GlobalStandard.text-embedding-3-small, 70'
-    ]
-  }
-})
-param openAILocation string
-```
+
+    If you have changed the above parameters from the `main.parameters.json` file, you **must change** the following configuration in `main.bicep` file so that the changes are reflected in automatic Azure OpenAI region filtering as well:
+
+    ```sh
+    @metadata({
+      azd: {
+        type: 'location'
+        usageName : [
+          'OpenAI.GlobalStandard.gpt-4o, 50'
+          'OpenAI.GlobalStandard.text-embedding-3-small, 70'
+        ]
+      }
+    })
+
+    ```
 
 3. Before the `azd` workflow proceeds, checks are performed in the selected infra region and recommendations are generated on failure for following cases to ensure that the deployment is successful.
-    - Azure Flexible Server for PostgreSQL SKU
+    - Azure Flexible Server for PostgreSQL SKU [we recommend avoiding burstable instances as they might result in "Illegal Instruction" error in certain regions for vector queries]
     - Azure Container Apps quota
     - azd env name
 
-4. After creating the resource group, the workflow prompts the user for the Azure Container Apps deployment. Input `yes` as shown below. The deployment might take some time and will provide progress updates in the terminal as well as on the Azure Portal.
+4. After creating the resource group, the workflow prompts the user for the Azure Container Apps deployment. Input `yes` as shown below -- this will build the docker images and deploy to Azure Container Apps. The docker image build can take 15+ mins.
+
 ```sh
 Do you want to deploy Azure Container Apps? (y/n): yes
 ```
@@ -168,23 +186,27 @@ Once the deployment is complete, `azd` will output the **application URLs** for
 > **Note:** Make sure to **copy the frontend URL** displayed in the output and open it in your browser to access the application.
 
 ### 🧹 Tear Down
+
 To destroy all the resources that have been created in the steps above, as well as remove any accounts deployed by the solution accelerator, use the following command:
+
 ```sh
 azd down --purge
 ```
+
 The `--purge` flag deletes all the accounts permanently.
 
 ### 🛟 Troubleshooting
-1. The troubleshooting guide for `azd cli` is [here](https://learn.microsoft.com/en-us/azure/developer/azure-developer-cli/troubleshoot?tabs=Browser).
-2. A validation error occurs when unsupported characters, such as `_`, `#` etc. are used while initializing or creating a new environment or resources. Refer to [rules and restrictions](https://learn.microsoft.com/en-us/azure/azure-resource-manager/management/resource-name-rules) for naming conventions.
+
+1. The [troubleshooting guide](https://learn.microsoft.com/azure/developer/azure-developer-cli/troubleshoot?tabs=Browser) for `azd cli`.
+2. A validation error occurs when unsupported characters, such as `_`, `#` etc. are used while initializing or creating a new environment or resources. Refer to [rules and restrictions](https://learn.microsoft.com/azure/azure-resource-manager/management/resource-name-rules) for naming conventions.
 3. A scope error occurs when the user does not have appropriate permissions when deploying resources through `azd` workflow. Attach `Contributor` role and `Role Based Access Control Administrator` role to user permissions before deploying the solution accelerator.
 4. When `The resource entity provisioning state is not terminal. Please wait for the provisioning state to become terminal and then retry the request` error occurs, restart the deployment using the `azd up` command.
 
 ### 📝 Additional Notes
 
 - Ensure all services are running and accessible at their respective ports.
 - If you encounter issues, check the logs for each service and verify your environment variables.
-- For troubleshooting Azure deployments, refer to the [Azure Developer CLI troubleshooting guide](https://learn.microsoft.com/en-us/azure/developer/azure-developer-cli/troubleshoot).
+- For troubleshooting Azure deployments, refer to the [Azure Developer CLI troubleshooting guide](https://learn.microsoft.com/azure/developer/azure-developer-cli/troubleshoot).
 - Make sure Docker has sufficient resources allocated for smooth operation.
 
 ## 🤖 Multi-Agent Personalization Workflow
@@ -205,8 +227,9 @@ These agents are orchestrated via LlamaIndex’s `Workflow` module, providing tr
 ### 🧬 Personalization & Memory
 
 The system uses **mem0** to persist user preferences across sessions. This memory influences agent behavior in real time:
+
 - Past interactions are retrieved at workflow start
-- New preferences (e.g., “I care about portability”) are stored dynamically
+- New preferences (e.g., “I want always see critical reviews about durability”) are stored dynamically
 - All memory is stored in the `mem0_chatstore` table in PostgreSQL
 
 > Agent prompts can be modified in [`backend/src/agents/prompts.py`](backend/src/agents/prompts.py) to customize tone, structure, or output logic.
@@ -240,39 +263,31 @@ Experience how AgenticShop delivers tailored product suggestions through multi-a
 
   ![product-details](./assets/images/product-details.png)
 
-
 - On the top right is the **Agentic Flow** drawer which provides an under-the-hood look at the multi-agent flow. Open the **Agentic Flow** panel to inspect how each agent participates in this multi-agent flow. Note that all agents are not always triggered. For example, in this case the review agent is not triggered.
 
   ![personalization](./assets/images/personalization.png)
 
-
 - Now lets try a command to update our personalization. In the command bar input new preferences (e.g., "Show critical reviews about durability").
 
   ![reviews_summary_query](./assets/images/reviews_summary_query.png)
 
-
 - This will re-trigger the personalization workflow and update the product page with this new information.
 
   ![updated_personalization](./assets/images/updated_personalization.png)
 
-
 - Open the **Agentic Flow** panel to see confirm that the **Review Agent** has now been triggered.
 
   ![personalization_agents](./assets/images/personalization_agents.png)
 
-
-- To see persistent personalization in action, try entering a preference like:
   **"Always show me critical reviews about durability."**
   In case the memory is updated a "Memory updated" message is shown. This new information be stored in memory using **mem0**.
 
   ![memory-updated](./assets/images/memory-updated.png)
 
-
 - Now, open a different product whose personalization hasn’t been generated yet. You should see a new section highlighting **critical reviews about durability** as part of the personalized output.
 
   ![updated_preferences](./assets/images/updated_preferences.png)
 
-
   > ⚠️ Note: Personalizations are cached for performance. If a product’s personalization was generated *before* the new memory was saved, it may not reflect the updated preference. Personalization workflow typically takes **~10 seconds** to complete.
 
 This walkthrough showcases how AgenticShop delivers a responsive, transparent, and memory-driven shopping experience.
@@ -303,7 +318,6 @@ This agent is equipped with **three tools**, each designed to handle a distinct
   Query (on a product detail page): `"Always show if red color is available in stock"`
   → The agent invokes the `query_about_product` tool, launching the multi-agent personalization workflow.
 
-
 ### 🛰️ Inspect Tool Execution
 
 - Open the **Agentic Flow** panel to see how the system:
@@ -314,7 +328,6 @@ This agent is equipped with **three tools**, each designed to handle a distinct
 
 ![command-routing-agent](./assets/images/command-routing-agent.png)
 
-
 > ⚠️ If a sentiment-based query fails to map to a known feature or lacks reviews, the system gracefully falls back to a general vector search and displays a notice.
 
 ## 🔍 Understanding What Queries Work (and Why)
@@ -412,7 +425,8 @@ If no reviews discuss a feature for a product, it won’t affect the ranking—e
 Use the examples below to test how different tools behave:
 
 ### 🧍 Product-Specific Query Handling
-_(Used when you're on a product page)_
+
+(*Used when you're on a product page*)
 
 - "Show critical reviews about durability"
 - "Always show a summary of critical reviews"
@@ -421,7 +435,8 @@ _(Used when you're on a product page)_
 - "Always highlight the product’s availability in red"
 
 ### 🧭 Standard Vector Search
-_(General product discovery with semantic matching)_
+
+(*General product discovery with semantic matching*)
 
 - "Wireless headphones"
 - "Waterproof smartwatches"
@@ -430,7 +445,8 @@ _(General product discovery with semantic matching)_
 - "Smartwatches with step tracking"
 
 ### 💬 Sentiment-Aware Search
-_(Feature-based preference queries)_
+
+(*Feature-based preference queries*)
 
 - "Headphones with great noise cancellation"
 - "Smartwatches with accurate heart rate monitoring"

README.md

@@ -1,4 +1,4 @@
-FROM arizephoenix/phoenix:version-8.31.0
+FROM arizephoenix/phoenix:version-11.24.1
 
 # Expose the ports Phoenix uses
 EXPOSE 6006