AI-powered full implementation generation from IntentLang specifications.
This crate translates parsed .intent specs into complete, working
implementations using an LLM. It builds on intent-codegen (skeleton stubs)
and intent-gen (LLM client) to produce contract-aware code.
The generation pipeline:
- Format the spec, generate skeleton code, extract contracts
- Send to LLM with implementation-focused system prompt
- Validate the output (expected names, balanced delimiters, no stubs)
- Retry with error feedback if validation fails