Port from Chat Completions API to Responses API (#292)

* Port to Responses API

* Revert temperature change

* Update async credentials, rm GitHub Models

* Port frontend to be more Responses like, remove unused session state and follow-up Qs

* Hopefully fix the CI and tests

* Use async cred for pg in prod

* Fix CI and mypy

* Exclude 3.10 from macos arm64, they dont work

Author: pamelafox

.devcontainer/Dockerfile

@@ -4,5 +4,5 @@ FROM mcr.microsoft.com/devcontainers/${IMAGE}
 ENV PYTHONUNBUFFERED 1
 
 RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
-    && apt-get -y install --no-install-recommends postgresql-client \
-     && apt-get clean -y && rm -rf /var/lib/apt/lists/*
+    && apt-get -y install --no-install-recommends postgresql-client zstd \
+     && apt-get clean -y && rm -rf /var/lib/apt/lists/*

.devcontainer/devcontainer.json

@@ -29,16 +29,13 @@
 			"extensions": [
 				"ms-python.python",
 				"ms-python.vscode-pylance",
-				"ms-python.vscode-python-envs",
 				"charliermarsh.ruff",
 				"mtxr.sqltools",
 				"mtxr.sqltools-driver-pg",
 				"esbenp.prettier-vscode",
 				"mechatroner.rainbow-csv",
 				"ms-vscode.vscode-node-azure-pack",
-				"esbenp.prettier-vscode",
 				"twixes.pypi-assistant",
-				"ms-python.vscode-python-envs",
 				"teamsdevapp.vscode-ai-foundry",
 				"ms-windows-ai-studio.windows-ai-studio"
 			],

.env.sample

@@ -5,16 +5,15 @@ POSTGRES_PASSWORD=postgres
 POSTGRES_DATABASE=postgres
 POSTGRES_SSL=disable
 
-# OPENAI_CHAT_HOST can be either azure, openai, ollama, or github:
+# OPENAI_CHAT_HOST can be either azure, openai, or ollama:
 OPENAI_CHAT_HOST=azure
-# OPENAI_EMBED_HOST can be either azure, openai, ollama, or github:
+# OPENAI_EMBED_HOST can be either azure, openai, or ollama:
 OPENAI_EMBED_HOST=azure
 # Needed for Azure:
 # You also need to `azd auth login` if running this locally
 AZURE_OPENAI_ENDPOINT=https://YOUR-AZURE-OPENAI-SERVICE-NAME.openai.azure.com
-AZURE_OPENAI_VERSION=2024-03-01-preview
-AZURE_OPENAI_CHAT_DEPLOYMENT=gpt-4o-mini
-AZURE_OPENAI_CHAT_MODEL=gpt-4o-mini
+AZURE_OPENAI_CHAT_DEPLOYMENT=gpt-5.4
+AZURE_OPENAI_CHAT_MODEL=gpt-5.4
 AZURE_OPENAI_EMBED_DEPLOYMENT=text-embedding-3-large
 AZURE_OPENAI_EMBED_MODEL=text-embedding-3-large
 AZURE_OPENAI_EMBED_DIMENSIONS=1024
@@ -35,9 +34,3 @@ OLLAMA_ENDPOINT=http://host.docker.internal:11434/v1
 OLLAMA_CHAT_MODEL=llama3.1
 OLLAMA_EMBED_MODEL=nomic-embed-text
 OLLAMA_EMBEDDING_COLUMN=embedding_nomic
-# Needed for GitHub Models:
-GITHUB_TOKEN=YOUR-GITHUB-TOKEN
-GITHUB_MODEL=openai/gpt-4o
-GITHUB_EMBED_MODEL=openai/text-embedding-3-large
-GITHUB_EMBED_DIMENSIONS=1024
-GITHUB_EMBEDDING_COLUMN=embedding_3l

.github/copilot-instructions.md

@@ -11,7 +11,7 @@ RAG on PostgreSQL is a Python FastAPI backend with React TypeScript frontend tha
 Install the following tools before beginning development:
 
 - **Python 3.10+** (3.12 recommended)
-- **Node.js 18+** for frontend development  
+- **Node.js 18+** for frontend development
 - **PostgreSQL 14+** with pgvector extension
 - **Azure Developer CLI (azd)** for deployment
 - **Docker Desktop** for dev containers (optional)
@@ -37,7 +37,7 @@ Run these commands in sequence. NEVER CANCEL any long-running commands:
    ```bash
    # Ubuntu/Debian:
    sudo apt update && sudo apt install -y postgresql-16-pgvector
-   
+
    # Start PostgreSQL and set password
    sudo service postgresql start
    sudo -u postgres psql -c "ALTER USER postgres PASSWORD 'postgres'"
@@ -97,7 +97,7 @@ Use "Frontend & Backend" configuration in the VS Code Run & Debug menu.
 ### Linting and Formatting (ALWAYS run before committing)
 ```bash
 ruff check .          # Lint code (takes <1 second)
-ruff format .          # Format code (takes <1 second)  
+ruff format .          # Format code (takes <1 second)
 mypy . --python-version 3.12  # Type check (takes ~42 seconds)
 ```
 
@@ -121,7 +121,7 @@ pytest tests/e2e.py --tracing=retain-on-failure
 **CRITICAL TIMING INFORMATION** - Set these timeout values and NEVER CANCEL:
 
 - **Dependencies install**: 90 seconds (use 180+ second timeout)
-- **Frontend npm install**: 22 seconds (use 60+ second timeout)  
+- **Frontend npm install**: 22 seconds (use 60+ second timeout)
 - **Frontend build**: 12 seconds (use 30+ second timeout)
 - **MyPy type checking**: 42 seconds (use 90+ second timeout)
 - **Full test suite**: 25 seconds (use 60+ second timeout)
@@ -156,7 +156,7 @@ pytest tests/e2e.py --tracing=retain-on-failure
    # Test API endpoints
    curl http://localhost:8000/items/1
    # Should return JSON with item data
-   
+
    # Test frontend
    curl http://localhost:8000/ | head -n 5
    # Should return HTML with "RAG on PostgreSQL" title
@@ -226,9 +226,6 @@ The application supports multiple OpenAI providers:
 3. **Ollama** (local):
    Set `OPENAI_CHAT_HOST=ollama`
 
-4. **GitHub Models**:
-   Set `OPENAI_CHAT_HOST=github`
-
 ## Common Issues and Solutions
 
 ### Database Connection Issues
@@ -267,7 +264,7 @@ Open `http://localhost:8089/` and point to your running application.
 The application provides these REST API endpoints (view full docs at `http://localhost:8000/docs`):
 
 - `GET /items/{id}` - Get specific item by ID
-- `GET /search` - Search items with text query 
+- `GET /search` - Search items with text query
 - `GET /similar` - Find similar items using vector search
 - `POST /chat` - Chat with RAG system (requires OpenAI configuration)
 - `POST /chat/stream` - Streaming chat responses
@@ -286,7 +283,7 @@ curl "http://localhost:8000/search?query=tent&limit=5"
 **Quick ls -la output for repository root:**
 ```
 .devcontainer/          # Dev container configuration
-.env.sample            # Environment variables template  
+.env.sample            # Environment variables template
 .github/               # GitHub Actions workflows
 .gitignore            # Git ignore patterns
 .pre-commit-config.yaml # Pre-commit hook configuration
@@ -309,8 +306,8 @@ tests/        # Test suite
 - **Always build and test locally before committing**
 - **Use pre-commit hooks** - they run ruff automatically
 - **Check the GitHub Actions** in `.github/workflows/` for CI requirements
-- **Reference the full README.md** for deployment and Azure-specific details  
+- **Reference the full README.md** for deployment and Azure-specific details
 - **Use VS Code with the Python and Ruff extensions** for the best development experience
 - **Never skip the frontend build** - the backend serves static files from `src/backend/static/`
 
-This project follows modern Python and TypeScript development practices with comprehensive tooling for code quality, testing, and deployment.
+This project follows modern Python and TypeScript development practices with comprehensive tooling for code quality, testing, and deployment.

.github/workflows/app-tests.yaml

@@ -27,10 +27,10 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: ["ubuntu-latest", "macos-latest-xlarge", "macos-13", "windows-latest"]
+        os: ["ubuntu-latest", "macos-26", "windows-latest"]
         python_version: ["3.10", "3.11", "3.12"]
         exclude:
-          - os: macos-latest-xlarge
+          - os: macos-26
             python_version: "3.10"
     env:
       UV_SYSTEM_PYTHON: 1
@@ -43,12 +43,12 @@ jobs:
         - uses: actions/checkout@v4
 
         - name: (MacOS) Install postgreSQL and pgvector using brew
-          if: matrix.os == 'macos-13' || matrix.os == 'macos-latest-xlarge'
+          if: matrix.os == 'macos-26'
           run: |
-            brew install postgresql@14
-            brew link --overwrite postgresql@14
+            brew install postgresql@18
+            brew link --overwrite postgresql@18
             brew install pgvector
-            brew services start postgresql@14 && sleep 1
+            brew services start postgresql@18 && sleep 1
             createuser -s ${{ env.POSTGRES_USERNAME }}
             psql -d postgres -c "ALTER USER ${{ env.POSTGRES_USERNAME }} WITH PASSWORD '${{ env.POSTGRES_PASSWORD }}'"
             psql -d postgres -c 'CREATE EXTENSION vector'

.github/workflows/evaluate.yaml

@@ -34,7 +34,6 @@ jobs:
       OPENAI_CHAT_HOST: ${{ vars.OPENAI_CHAT_HOST }}
       OPENAI_EMBED_HOST: ${{ vars.OPENAI_EMBED_HOST }}
       AZURE_OPENAI_ENDPOINT: ${{ vars.AZURE_OPENAI_ENDPOINT }}
-      AZURE_OPENAI_VERSION: ${{ vars.AZURE_OPENAI_VERSION }}
       AZURE_OPENAI_CHAT_DEPLOYMENT: ${{ vars.AZURE_OPENAI_CHAT_DEPLOYMENT }}
       AZURE_OPENAI_CHAT_MODEL: ${{ vars.AZURE_OPENAI_CHAT_MODEL }}
       AZURE_OPENAI_EMBED_DEPLOYMENT: ${{ vars.AZURE_OPENAI_EMBED_DEPLOYMENT }}

.vscode/settings.json

@@ -36,5 +36,6 @@
         "htmlcov": true,
         ".mypy_cache": true,
         ".coverage": true
-    }
+    },
+    "python-envs.defaultEnvManager": "ms-python.python:system"
 }

AGENTS.md

@@ -8,3 +8,19 @@ When adding new azd environment variables, update:
 1. infra/main.parameters.json : Add the new parameter with a Bicep-friendly variable name and map to the new environment variable
 1. infra/main.bicep: Add the new Bicep parameter at the top, and add it to the `webAppEnv` object
 1. .github/workflows/azure-dev.yml: Add the new environment variable under `env` section. If it's a @secure variable in main.bicep, it should come from `secrets`, otherwise from `vars`.
+
+## Upgrading Python packages
+
+1. Update the version constraint in src/backend/pyproject.toml
+
+2. Re-compile src/backend/requirements.txt from the src folder:
+
+    ```shell
+    uv pip compile pyproject.toml -o requirements.txt --python-version 3.10
+    ```
+
+3. Reinstall with:
+
+    ```shell
+    python -m pip install -r src/backend/requirements.txt
+    ```

azure.yaml

@@ -42,7 +42,6 @@ pipeline:
     - OPENAI_CHAT_HOST
     - OPENAI_EMBED_HOST
     - AZURE_OPENAI_ENDPOINT
-    - AZURE_OPENAI_VERSION
     - AZURE_OPENAI_CHAT_DEPLOYMENT
     - AZURE_OPENAI_CHAT_MODEL
     - AZURE_OPENAI_CHAT_DEPLOYMENT_VERSION

evals/eval_config.json

@@ -8,8 +8,7 @@
             "use_advanced_flow": true,
             "top": 3,
             "retrieval_mode": "hybrid",
-            "temperature": 0.3,
-            "seed": 42
+            "temperature": 0.3
         }
     },
     "target_response_answer_jmespath": "message.content",