LLM tooling
Prompt Perfector

Prompt Perfector

Paste a rough prompt, pick a style, and get a sharper version back.

Sign in

Configuration status

Whether each provider API key is configured on the server. Key values are never sent to the browser.

Reading configuration…

Backend health

Reports the full readiness surface: credentials, verdict ladder, circuit breaker, request metrics and retrieval readiness. The standard check makes no billable request.

Sign-in required to use the enhancer

Enhancements are billable, so you must sign in before running one. Sign in inside the Lovable preview pane — open the preview — then return here.

Sign in
30,000 of 30,000 characters left
Characters: 0Words: 0Tokens: unavailableCounted as typed, including leading and trailing spaces; submission checks the trimmed text.

Long lines wrap automatically — the limit is 30,000 characters.

Limit advisor — get a recommended prompt-length limit

0 / 2,000

Sign-in required — limit advice is a billable model call.

Style
Concise:
Keeps the rewrite tight and direct: filler is removed and every essential constraint is kept.
Detailed:
Expands the prompt with explicit structure, constraints, edge cases and success criteria, inserting a neutral {{PLACEHOLDER}} wherever a specific is missing rather than inventing one.
Structured:
Organises the rewrite into clearly labelled sections such as Role, Task, Context, Constraints and Output format.

Enter a prompt to enhance.

Recent prompts

Prompts you enhance appear here — the last 50, kept in this browser only.

Ask your documents

Answers are grounded in your own documents only. When the passages do not contain the answer, the model says so instead of guessing.

Sign in to add documents. Documents are stored per account, so the fields and the Add document button stay disabled until you are signed in. Sign in

Prepared documents

Loading fills the form only. Nothing is ingested until you choose to add it.

  • Lyapunov/x^x=49 · docs/rag-corpus/lyapunov-xx49.md
  • Qwen3-8B Flash-Next dataflow v2 · docs/rag-corpus/qwen38-flash-next-dataflow-v2.md
  • Role Prompt v1.5 — Gasparyan Verification-First · docs/governance/role-prompt-v1.5.md
  • Technical Retrospective 2026-08-30 · docs/RETROSPECTIVE_2026-08-30.md
  • Claude Code Slash Commands — Practical Reference · docs/rag-corpus/claude-code-slash-commands.md
  • NLP Engineering Data Flow — Retrieval and Attribution · docs/rag-corpus/nlp-engineering-data-flow.md
  • The LLM Pipeline — Executive Memo · docs/rag-corpus/llm-pipeline-executive-memo.md
  • Declarative Attention (arXiv 2609.02737) — Full-Paper Analysis · docs/rag-corpus/declarative-attention-paper-analysis.md
  • State of AI — 100 Trillion Token Study (a16z / OpenRouter, Dec 2025) — Summary · docs/rag-corpus/state-of-ai-openrouter-summary.md
  • Qwen3.8 quickstart — from the vendor's sources · docs/rag-corpus/qwen38-quickstart.md
0 / 100,000 characters
Sign in to add documents. Sign in