Upgrade HackGPT: Improve codebase, multi-provider support and more

.devcontainer/devcontainer.json

@@ -0,0 +1,24 @@
+// Based on the official Microsoft-maintained Python Dev Container template:
+// https://github.com/devcontainers/templates/tree/main/src/python
+{
+  "name": "HackGPT Python",
+  "image": "mcr.microsoft.com/devcontainers/python:3-3.14-trixie",
+  "forwardPorts": [8501],
+  "portsAttributes": {
+    "8501": {
+      "label": "HackGPT (Streamlit)",
+      "onAutoForward": "openPreview"
+    }
+  },
+  "postCreateCommand": "python -m pip install -r requirements-dev.txt",
+  "customizations": {
+    "vscode": {
+      "settings": {
+        "task.allowAutomaticTasks": "on"
+      },
+      "extensions": [
+        "ms-python.python"
+      ]
+    }
+  }
+}

.env.example

@@ -0,0 +1,12 @@
+# Copy to .env or export these variables in the process environment.
+# Add only credentials for providers you intend to use.
+OPENAI_API_KEY=
+OPENAI_BASE_URL=
+
+ANTHROPIC_API_KEY=
+ANTHROPIC_BASE_URL=
+
+# Omit OLLAMA_HOST to use the Ollama client's local default.
+OLLAMA_HOST=http://localhost:11434
+# Optional when connecting to Ollama Cloud or another authenticated endpoint.
+OLLAMA_API_KEY=

.gitignore

@@ -1 +1,7 @@
-app/.streamlit/secrets.toml
+src/.streamlit/secrets.toml
+.env
+.env.*
+!.env.example
+__pycache__/
+*.py[cod]
+.venv/

.vscode/tasks.json

@@ -0,0 +1,36 @@
+{
+  "version": "2.0.0",
+  "tasks": [
+    {
+      "label": "HackGPT: Start Streamlit",
+      "type": "process",
+      "command": "python",
+      "args": [
+        "-m",
+        "streamlit",
+        "run",
+        "src/app.py",
+        "--server.address=0.0.0.0",
+        "--server.port=8501"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "isBackground": true,
+      "problemMatcher": [],
+      "presentation": {
+        "reveal": "always",
+        "revealProblems": "never",
+        "focus": false,
+        "panel": "dedicated",
+        "clear": true,
+        "showReuseMessage": false
+      },
+      "runOptions": {
+        "runOn": "folderOpen",
+        "instanceLimit": 1,
+        "instancePolicy": "silent"
+      }
+    }
+  ]
+}

README.md

@@ -1,73 +1,120 @@
 # HackGPT
 
-Welcome to HackGPT, a powerful and customizable AI chatbot interface designed specifically for developers. With a ChatGPT-like experience, HackGPT brings a host of key features to enhance your hacking capabilities and empower your development process.
+HackGPT is a Streamlit chat interface with provider-neutral model discovery,
+streaming, setup prompts, and schema-driven generation controls.
 
-[![HackGPT Demo](images/demo.gif)](./images/demo.gif)
+[![HackGPT Demo](static/images/demo.gif)](./static/images/demo.gif)
 
-## Current Features
+## Features
 
-- **ChatGPT-like Interface**: Immerse yourself in a chat-like environment with streaming output and a typing effect. Enable or disable the typing effect based on your preference for quick responses.
+- **Multiple providers**: Use OpenAI, Anthropic Claude, or Ollama without changing
+  the chat interface.
+- **Dynamic model discovery**: The model picker reflects the models available to
+  the configured account or Ollama host. Models can be refreshed from the sidebar.
+- **Capability-aware configuration**: Every model supplies its own typed option
+  schema, including defaults, ranges, choices, and conditional fields. Unsupported
+  controls are never sent to a provider.
+- **Provider-neutral chat**: Text history and system prompts remain usable when
+  switching between compatible providers and models.
+- **Streaming responses**: All provider adapters stream response text into the
+  existing chat interface.
+- **Setup prompts and historical context**: Select a bundled or custom setup prompt
+  and optionally add a validated historical knowledge cutoff.
 
-- **Fine-tuning**: Tailor your HackGPT experience with the sidebar's range of options. Choose from different models like GPT-3, GPT-4, or specific models such as 'gpt-3.5-turbo'. Fine-tune model response parameters and configure API settings.
+The capability model already represents streaming, system prompts, tools,
+structured output, image input, and reasoning. Adding UI for future capabilities
+does not require provider-specific branches in the generic option renderer.
 
-- **Set-up Prompt Selection**: Unlock more specific responses, results, and knowledge by selecting from a variety of preset set-up prompts. Additionally, craft your own custom set-up prompt for personalized interactions.
+## Provider configuration
 
-- **Knowledge Cut-off**: Utilize GPT-4's time-travel capabilities by setting a "knowledge cut-off" date. Limit the AI's knowledge to a specific time in the past and explore historical insights. Gain a greater understanding of historical events, cultural nuances, and prevailing trends.
+Credentials and endpoints are read only from environment variables. They are not
+entered or displayed in the Streamlit interface.
 
-- **On-demand Switching**: Seamlessly switch between models (GPT-3, GPT-4, etc.) and set-up prompts to unlock diverse and more dynamic AI-generated responses. Benefit from the advanced contextual understanding and improved transfer learning capabilities of different models.
+| Provider | Required | Optional |
+| --- | --- | --- |
+| OpenAI | `OPENAI_API_KEY` | `OPENAI_BASE_URL` |
+| Anthropic | `ANTHROPIC_API_KEY` | `ANTHROPIC_BASE_URL` |
+| Ollama | A reachable Ollama service | `OLLAMA_HOST`, `OLLAMA_API_KEY` |
 
-## Getting Started
+Copy the example file and fill in only the providers you use:
 
-HackGPT is built using Python 3.9 and Streamlit. It can be easily deployed in a containerized environment using the included Docker Compose set-up.
+```bash
+cp .env.example .env
+```
+
+Docker Compose reads the root `.env` file automatically. For a local Python run,
+export the variables into the process environment before starting Streamlit. If
+`OLLAMA_HOST` is omitted, the Ollama client uses its local default. When Streamlit
+runs in a container and Ollama runs on the host, set `OLLAMA_HOST` to an address
+reachable from that container.
+
+Provider availability is account- and deployment-specific. A configured provider
+can still report a clear discovery error when credentials are invalid, its endpoint
+is unreachable, or it has no compatible chat models.
 
-To get started, follow the steps below:
+## Getting started
 
-### Docker Compose (recommended)
+HackGPT requires Python 3.12 or newer.
 
-The easiest way to run HackGPT is to use Docker Compose. To do so, run:
+### Docker Compose
 
 ```bash
-docker-compose up # -d to run in background
+docker compose -f docker/docker-compose.dev.yml up
 ```
 
-To stop the container, press `Ctrl+C` in the terminal window, or run:
+Stop or remove the development service with:
 
 ```bash
-docker-compose stop
+docker compose -f docker/docker-compose.dev.yml stop
+docker compose -f docker/docker-compose.dev.yml down
 ```
 
-... in case you used the `-d` flag to run the container in the background.
-
-To remove the container, run:
+### Local development
 
 ```bash
-docker-compose down
+python -m pip install -r requirements-dev.txt
+streamlit run src/app.py
 ```
 
-### Docker (manually)
+### VS Code Dev Container
 
-If you don't want to use Docker Compose, you can also run the container manually. To do so, run:
+Choose **Dev Containers: Reopen in Container** from the Command Palette. The
+container installs the development requirements, forwards port 8501, and runs the
+`HackGPT: Start Streamlit` workspace task when the folder opens. Its dedicated
+integrated terminal is revealed automatically so startup messages, request logs,
+and errors remain visible. The forwarded app also opens in a VS Code preview.
 
-```bash
-docker run --rm -p 8501:8501 -v $(pwd)/app:/app -w /app python:3.9-slim bash -c "pip install -r requirements.txt && streamlit run app.py"
-```
+Use `Ctrl+C` in that terminal to stop Streamlit. To start it again, run
+**Tasks: Run Task** and select `HackGPT: Start Streamlit`. After changing
+`.devcontainer/devcontainer.json`, use **Dev Containers: Rebuild and Reopen in
+Container** so the updated container settings take effect.
 
-### Local Python environment
+## Quality checks
 
-If you don't want to use Docker, you can also run HackGPT in a local Python environment. To do so, run:
+Run the complete local verification suite from the repository root:
 
 ```bash
-cd ./app
-pip install -r requirements.txt
-streamlit run app.py
+ruff format --check src tests
+ruff check src tests
+mypy src tests
+pytest
 ```
 
-## Feedback
-
-We hope this initial release of HackGPT serves as a valuable tool for developers, encouraging exploration and experimentation with GPT-based models. Your feedback on HackGPT is highly appreciated as it will help us shape future versions of the project.
+## Architecture
 
-We're thrilled to have you join us in this journey of unleashing new possibilities and customizability for developers with HackGPT!
+- `src/app.py` — Streamlit entry point and provider-registry lifetime.
+- `src/hackgpt/providers/types.py` — provider protocol, capabilities, option schemas,
+  normalised messages, validation, and user-facing provider errors.
+- `src/hackgpt/providers/` — isolated OpenAI, Anthropic, and Ollama adapters plus the
+  provider registry.
+- `src/hackgpt/ui/options.py` — generic schema-to-Streamlit widget rendering.
+- `src/hackgpt/ui/sidebar.py` — provider/model discovery and application controls.
+- `src/hackgpt/ui/chat.py` — provider-neutral history and streamed chat rendering.
+- `src/hackgpt/prompts.py` — bundled setup-prompt loading and historical context.
+- `src/hackgpt/state.py` — session-state initialisation and history normalisation.
+- `tests/` — schema, selection, adapter translation, state, and Streamlit smoke tests.
 
 ## License
 
-This project is licensed under the terms of the ISC license. See [LICENSE](LICENSE) for more details.
+This project is licensed under the terms of the ISC license. See [LICENSE](LICENSE)
+for details.