Mentenaz AI — Capabilities

The intelligence layer behind Forge, Helm, and Twin.

🔥 Forge AI

Inside the Forge desktop application.

Reads the LSP error, generates a corrected file, and shows a diff you can accept.

Summarises your staged diff into a conventional commit message.

Schema-aware query suggestions from your connected database.

Explains what a changed file does and why it changed.

Flags dangerous changes (rm -rf, secret leaks, breaking imports) before you commit.

Generates code from a natural language description in your editor.

GitHub intelligence inside the Helm panel.

Summarises what a pull request does in plain English.

Generates changelogs from commits between tags.

Suggests labels and priority for new issues.

Generates a repository description from the codebase.

The conversational AI inside Forge.

General conversation with workspace context. Ask questions about your code.

Structured ideation. Think through architecture decisions and design patterns.

Code generation with open file context. The AI sees what you're working on.

Proactive suggestions about your current file based on LSP diagnostics.

Layer	Technology
Inference	5-provider fallback (Gemini → Groq → HuggingFace → Mistral → OpenRouter)
Embeddings	Voyage AI
Vector search	pgvector on Supabase
Transport	SSE streaming via Supabase Edge Functions
Runtime	Deno on Supabase