Skip to content

AppCoder v1.0.8_alpha

Pre-release
Pre-release

Choose a tag to compare

@ChristophGeske ChristophGeske released this 11 Aug 00:04
· 49 commits to dev since this release

📱 AppCoder v1.0.8_alpha

What's New in This Release?

  • Option to use GPT-5:
    The model picker was simplified. Now only the important OpenAI models GPT-5, GPT-5-mini and GPT-5-nano as well as the Gemini models 2.5 Pro and Flash are listed. GPT-5-nano is the fastest but not really useful for more than tiny changes to an app. GPT-5-mini is similar to Gemini 2.5 Pro.
  • Gemini 2.5 Pro is now the Free Default:
    Gemini 2.5 Pro API is now free for 100 requests a day. This is one of the best coding models and I now made it the default. I highly recommend sticking to it as long as you can. GPT-5 Mini is not that expensive a few cents per app and can be an alternative if 2.5 Pro fails. GPT-5 is so expensive that I did not test it much. My limited testing seemed to result in builds with less retries so you might expect some improvements using it. However the resulting apps were not noticeably more advanced so use it only if 2.5 pro and GPT-5 Mini struggle to build the feature you are working on. GPT-5 Mini costs around 0.04 cent per small app while GPT-5 often costs around 0.4$ for a small app like Tetris.
  • Main Menu allows Opening of Existing Projects:
    Now the main menu allows you to access all your existing projects. For example when you want to look into the code or change a small detail you can use this new access to the apps to change the files manually.
  • **"Less Errors by using Structural Outputs Right:"
    Improved Structural Outputs by implementing best practices. Should help ensure the LLM only delivers code in a correct format reducing errors.

✨ Features:

  • Generate basic Android app code from text descriptions using the Gemini 2.5 Pro LLM model.
  • Modify existing AppCoder-generated projects by providing new descriptions or feature requests.
  • 🔑 Requires a free API key from Google AI Studio (https://aistudio.google.com/apikey). This key must be entered into the AppCoder app to enable AI features.
  • 🛠️ Build, install and modify your AI-generated app directly on your device.

🚀 How to Install and Test AppCoder:

  1. ⬇️ Download the AppCoder-v1.0.8_alpha.apk file.
  2. 📲 Transfer the .apk file to your Android phone.
  3. 📂 On your phone, open a file manager, find the downloaded .apk file, and tap it to install. You may need to enable Install from unknown sources in your settings.
  4. ✅ Once AppCoder is installed, grant necessary permissions and wait for the initial setup to complete. Then:
    a. Tap the neon button in the bottom left to get to the AI menu.
    b1. App Name: To create a new app, type a unique name.
    b2. To modify an existing project, type the existing app name in the text field.
    c1. Provide a simple App Description (e.g., "A to-do list app")
    c2. Describe what to modify on your existing app (e.g., "change text color to red") .
    d. Paste your Gemini API Key which you can get from Google AI Studio: https://aistudio.google.com/apikey.
    e. Either keep the default Gemini Model ("2.5 Pro" because it is free and very good) or choose one of the other models from the dropdown.
    f. Set the checkmark on "Enable Fully Automated Build" (recommended) to let the system send errors to the LLM automatically and retry the build a second time if it fails.
    g. Tap "Generate New App" or "Modify Existing App" in case you haven't enabled fully automated build.
  5. ⏳Wait for the AI to process your request and apply the changes. You can monitor the progress in the "Generation Log" and see a summary in the "AI Conclusion" field.
  6. 📲 If successful, you'll be prompted to build and then install your app. You might need to expand notification details to find and select "Install anyway" if prompted by the Google Play Protect security check.

❗ Troubleshooting & Important Notes:

  • If AI Generation/Modification Fails:
    • If a generated app builds but fails after the installation you can try to modify the app by letting the LLM build it again with a description of the failure. You can also go back to older versions that worked before and try again.
  • Simplicity is Key: For best results with the current alpha, provide clear and concise descriptions and stick to simple app functionalities. When modifying apps try to only ask for one feature change so as not to confuse the llm. The simpler the feature you want to add to an app the higher the probability it works.

⚠️ Known Issues:

  • Inconsistencies when running in Background Not sure what causes this; ideally you leave the app open during the complete generate and build process.
  • Initial AppCoder Setup: The first-time setup installs development tools and can occasionally fail. A reinstall of the AppCoder APK usually fixes this.
  • Navigation During Initial Setup: While AppCoder is performing its first setup, if you close it you can’t get past the initial screen. If stuck on the initial screen, just reinstall AppCoder.
  • Modifying can break apps but you can go back to previous versions, they are now saved with the endings _v1, _v2 and so on. If you want to be 100% sure to keep an app copy the whole project from a file explorer. The file explorer in this app is a bit limited so better use one that has access to all directories instead.

app-arm64-v8a-release.apk is most likely the format you need when using a modern smartphone