AI PHP Code Generation Agent
Budget: $10 – $30 USD
I’m building a workflow where an autonomous agent writes the first pass of a web-application in PHP every time I feed it a short functional brief. Your job is to set up that agent: choose or fine-tune an existing LLM, wrap it with the right prompt-engineering logic, and return clean, runnable PHP that follows PSR-12, uses modern OOP patterns, and respects basic security practices (input sanitisation, prepared statements, CSRF tokens, etc.).
The core focus is code generation—bug detection and performance tuning can stay out of scope for now, although I’d like the architecture to leave room for those extensions later. You’re free to integrate OpenAI, Claude, PaLM, or an open-source model; what matters is consistent output quality and minimal hallucination. I’ll provide sample feature descriptions and an empty Git repo; the agent should commit code and a brief markdown summary for each task.
Deliverables
• A reproducible environment (Docker or equivalent) with the agent running via CLI and, ideally, a lightweight web hook
• Prompt templates and any fine-tuning files
• A short README explaining setup, model limitations, and how to extend the agent for future bug-finding or optimisation modules
Acceptance criteria
1. Given the sample briefs, the agent produces namespaced PHP files that run without syntax errors in PHP 8.1+
2. Generated code passes php-cs-fixer on the default ruleset
3. Unit stubs are included with PHPUnit 10
If this sounds like your wheelhouse, tell me about a similar autonomous-coding tool you’ve built and which model stack you’d start with here.
The core focus is code generation—bug detection and performance tuning can stay out of scope for now, although I’d like the architecture to leave room for those extensions later. You’re free to integrate OpenAI, Claude, PaLM, or an open-source model; what matters is consistent output quality and minimal hallucination. I’ll provide sample feature descriptions and an empty Git repo; the agent should commit code and a brief markdown summary for each task.
Deliverables
• A reproducible environment (Docker or equivalent) with the agent running via CLI and, ideally, a lightweight web hook
• Prompt templates and any fine-tuning files
• A short README explaining setup, model limitations, and how to extend the agent for future bug-finding or optimisation modules
Acceptance criteria
1. Given the sample briefs, the agent produces namespaced PHP files that run without syntax errors in PHP 8.1+
2. Generated code passes php-cs-fixer on the default ruleset
3. Unit stubs are included with PHPUnit 10
If this sounds like your wheelhouse, tell me about a similar autonomous-coding tool you’ve built and which model stack you’d start with here.