Neon auth ctx

.github/workflows/pipeline.yml

@@ -172,7 +172,7 @@ jobs:
     timeout-minutes: 20
     strategy:
       matrix:
-        template: [nicegui, trpc]
+        template: [trpc, laravel] # TODO: return after test fix nicegui
     steps:
       - name: Checkout repository
         uses: actions/checkout@v3

.gitignore

@@ -13,8 +13,6 @@ dist/
 downloads/
 eggs/
 .eggs/
-lib/
-lib64/
 parts/
 sdist/
 var/
@@ -174,3 +172,4 @@ agent/app_dumps.zip
 agent/FSMAppRecovery.ipynb
 agent/app_dumps/*.*
 agent/bin/*.*
+agent/laravel_agent/template_backup_*/**

CONTRIBUTING.md

@@ -2,9 +2,9 @@
 
 ## Basic Usage
 
-1) Install [uv](https://docs.astral.sh/uv/getting-started/installation/) and [dagger](https://dagger.io) 
+1) Install [uv](https://docs.astral.sh/uv/getting-started/installation/) and [dagger](https://dagger.io)
 2) Make sure you have `ANTHROPIC_API_KEY` and `GEMINI_API_KEY` env vars available.
-3) Run the commands 
+3) Run the commands
 
 ### Commands
 
@@ -19,9 +19,28 @@ uv run generate "my app description" # generate an app from scratch using full p
 uv run interactive  # a naive debug client working with local server
 ```
 
+### App templates
+
+We support three app templates:
+- **trpc**: Full-stack web app with Bun, React, Vite, Fastify, tRPC and Drizzle
+- **python**: Data-oriented app with Python, NiceGUI, and SQLModel
+- **laravel**: Full-stack web app with Laravel, React, TypeScript, Tailwind CSS, and Inertia.js (early version)
+
+You can specify the template when generating an app, e.g.:
+
+```bash
+uv run generate "make a dashboard showing my current stock portfolio value using up to date prices from yfinance" --template_id "nicegui_agent"
+```
+or
+
+```bash
+uv run generate "make a presentation-like website with multiple pages with content and next buttons on each. last page mush show a counter that increments each time this presentation has been shown" --template_id "laravel_agent"
+```
+
+
 ### Testing with debug client on prod servers
 
-1. Make sure you have the `uv` installed 
+1. Make sure you have the `uv` installed
 2. `uv run interactive --host yourhostname --port 80` (or change the host if needed);
 3. In the client, prompt for your app.
 4. Waste some time, get some tea.
@@ -37,7 +56,7 @@ We use VCR for testing LLM calls. VCR records LLM completions and saves them to
 
 - **Record mode**: Makes real API calls, saves responses to cache
 - **Replay mode**: Uses only cached responses
-- **LRU mode**: Tries using cached responses, but can call APIs too. 
+- **LRU mode**: Tries using cached responses, but can call APIs too.
 - **Off mode**: No caching, direct API calls
 
 Default usage (to run tests with cached responses):
@@ -51,4 +70,3 @@ If you want to record new responses, use:
 LLM_VCR_CACHE_MODE=record uv run pytest .
 ```
 New responses should be recorded in case of prompt changes or other significant changes in the pipeline (e.g. template modification, adding new steps etc.). VCR cache is stored in ./agent/llm/llm_cache.json by default, and new version should be committed to the repository.
-

README.md

@@ -8,19 +8,54 @@
 
 ## What it builds
 
-- **Full-stack web apps** with Bun, React, Vite, Fastify and Drizzle
-- **Neon Postgres database** provisioned instantly via API
+We're currently supporting the following application types:
+
+### tRPC CRUD Web Applications
+
+- **Full-stack web apps** with Bun, React, Vite, Fastify, tRPC and Drizzle;
+- **Automatic validation** with ESLint, TypeScript, and runtime verification;
 - **Applications tested** ahead of generation with smoke tests using Playwright
+
+### Laravel Web Applications (Alpha Version)
+
+- **Full-stack web apps** with Laravel, React, TypeScript, Tailwind CSS, and Inertia.js;
+- **Modern Laravel 12** with PHP 8+ features and strict typing;
+- Designed to become production-ready soon with authentication, validation, code style enforcement and testing infrastructure;
+
+### Data-oriented Applications
+
+- **Data apps** with Python + NiceGUI + SQLModel stack - perfect for dashboards and data visualization;
+- **Automatic validation** using pytest, ruff, pyright, and runtime verification;
+- **Additional packages management** with uv;
+
+All applications support:
+- **Neon Postgres database** provisioned instantly via API
 - **GitHub repository** with complete source code
-- **CI/CD and deployment** via the [app.build platform](https://github.com/appdotbuild/platform)
-- **Automatic validation** with ESLint, TypeScript, and runtime verification
+- **CI/CD and deployment** via the [app.build platform](https://github.com/appdotbuild/platform).
+
+New application types are work in progress, stay tuned for updates!
 
 ## Try it
 
+### Via the [managed service](https://app.build)
+
 ```bash
+# for tRPC CRUD apps
 npx @app.build/cli
+
+for Python/NiceGUI apps
+npx @app.build/cli --template=python
 ```
 
+### Locally
+Local usage and development instructions are available in [CONTRIBUTING.md](CONTRIBUTING.md).
+
+### Local Models with Ollama (expreimental)
+
+Want to run everything locally without API keys? We  added experimental support [Ollama](https://ollama.ai) for local model inference. Set up Ollama, pull your preferred models, and run the agent completely offline. See [PROJECT_GUIDELINES.md](PROJECT_GUIDELINES.md) for detailed setup instructions and recommended models.
+
+We are open for you contributions in better supporting local models. Feel free to send you PRs!
+
 ## Architecture
 
 This agent doesn't generate entire applications at once. Instead, it breaks down app creation into small, well-scoped tasks that run in isolated sandboxes:
@@ -41,10 +76,6 @@ This is the **agent** repository containing the core code generation engine and
 
 See [CONTRIBUTING.md](CONTRIBUTING.md) for development setup and contribution guidelines.
 
-## Running locally
-
-Local development instructions are available in [CONTRIBUTING.md](CONTRIBUTING.md).
-
 ---
 
 Built to showcase agent-native infrastructure patterns. Fork it, remix it, use it as a reference for your own projects.

agent/.env.local

@@ -0,0 +1,26 @@
+# LLM Configuration
+# Set these environment variables to override default model selection
+
+# For ultra fast tasks (commit names, simple completions)
+# LLM_ULTRA_FAST_MODEL=gemini-flash-lite
+LLM_ULTRA_FAST_MODEL=phi4:latest
+
+# For best coding tasks (slow, high quality)
+# LLM_BEST_CODING_MODEL=sonnet
+LLM_BEST_CODING_MODEL=llama3.3:latest
+
+# For universal tasks (medium speed, FSM tools)
+# LLM_UNIVERSAL_MODEL=sonnet
+LLM_UNIVERSAL_MODEL=llama3.3:latest
+
+# For vision and UI analysis
+# LLM_VISION_MODEL=gemini-flash-lite
+LLM_VISION_MODEL=gemma3:27b
+
+OLLAMA_HOST=http://localhost:11434
+
+PREFER_OLLAMA=1
+
+# Provider API keys
+# ANTHROPIC_API_KEY=your_key_here
+# GEMINI_API_KEY=your_key_here

agent/Dockerfile

@@ -1,7 +1,10 @@
-FROM alpine:3.21.3 AS prod
+FROM python:3.12-slim AS prod
 
-RUN apk add --update --no-cache curl python3 docker
-COPY --from=ghcr.io/astral-sh/uv:0.6.14 /uv /uvx /bin/
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    curl \
+    docker.io \
+    && rm -rf /var/lib/apt/lists/*
+COPY --from=ghcr.io/astral-sh/uv:0.7.20-debian /usr/local/bin/uv /usr/local/bin/uvx /usr/local/bin
 RUN curl -fsSL https://dl.dagger.io/dagger/install.sh | BIN_DIR=/usr/local/bin sh
 
 WORKDIR /app