UniqAI/bolt.new
1
0
Fork 0
mirror of https://github.com/stackblitz/bolt.new synced 2025-02-06 04:48:04 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

feat: better prompt enhancement

Browse Source
This commit is contained in:
Sujal Shah 2024-11-26 14:17:25 +05:30
parent d41a4acb89
commit e25bb28255
2 changed files with 22 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@ -31,6 +31,7 @@ https://thinktank.ottomator.ai
- ✅ Ability to revert code to earlier version (@wonderwhy-er) - ✅ Ability to revert code to earlier version (@wonderwhy-er)
- ✅ Cohere Integration (@hasanraiyan) - ✅ Cohere Integration (@hasanraiyan)
- ✅ Dynamic model max token length (@hasanraiyan) - ✅ Dynamic model max token length (@hasanraiyan)
- ✅ Better prompt enhancing (@SujalXplores)
- ⬜ **HIGH PRIORITY** - Prevent Bolt from rewriting files as often (file locking and diffs) - ⬜ **HIGH PRIORITY** - Prevent Bolt from rewriting files as often (file locking and diffs)
- ⬜ **HIGH PRIORITY** - Better prompting for smaller LLMs (code window sometimes doesn't start) - ⬜ **HIGH PRIORITY** - Better prompting for smaller LLMs (code window sometimes doesn't start)
- ⬜ **HIGH PRIORITY** - Load local projects into the app - ⬜ **HIGH PRIORITY** - Load local projects into the app
@ -43,7 +44,6 @@ https://thinktank.ottomator.ai
- ⬜ Vertex AI Integration - ⬜ Vertex AI Integration
- ⬜ Deploy directly to Vercel/Netlify/other similar platforms - ⬜ Deploy directly to Vercel/Netlify/other similar platforms
- ⬜ Prompt caching - ⬜ Prompt caching
- ⬜ Better prompt enhancing
- ⬜ Have LLM plan the project in a MD file for better results/transparency - ⬜ Have LLM plan the project in a MD file for better results/transparency
- ⬜ VSCode Integration with git-like confirmations - ⬜ VSCode Integration with git-like confirmations
- ⬜ Upload documents for knowledge - UI design templates, a code base to reference coding style, etc. - ⬜ Upload documents for knowledge - UI design templates, a code base to reference coding style, etc.

23
app/routes/api.enhancer.ts
View File

@ -44,9 +44,28 @@ async function enhancerAction({ context, request }: ActionFunctionArgs) {
content: content:
`[Model: ${model}]\n\n[Provider: ${providerName}]\n\n` + `[Model: ${model}]\n\n[Provider: ${providerName}]\n\n` +
stripIndents` stripIndents`
You are a professional prompt engineer specializing in crafting precise, effective prompts.
Your task is to enhance prompts by making them more specific, actionable, and effective.
I want you to improve the user prompt that is wrapped in \`<original_prompt>\` tags. I want you to improve the user prompt that is wrapped in \`<original_prompt>\` tags.
IMPORTANT: Only respond with the improved prompt and nothing else! For valid prompts:
- Make instructions explicit and unambiguous
- Add relevant context and constraints
- Remove redundant information
- Maintain the core intent
- Ensure the prompt is self-contained
- Use professional language
For invalid or unclear prompts:
- Respond with a clear, professional guidance message
- Keep responses concise and actionable
- Maintain a helpful, constructive tone
- Focus on what the user should provide
- Use a standard template for consistency
IMPORTANT: Your response must ONLY contain the enhanced prompt text.
Do not include any explanations, metadata, or wrapper tags.
<original_prompt> <original_prompt>
${message} ${message}
@ -79,7 +98,7 @@ async function enhancerAction({ context, request }: ActionFunctionArgs) {
}, },
}); });
const transformedStream = result.toAIStream().pipeThrough(transformStream); const transformedStream = result.toDataStream().pipeThrough(transformStream);
return new StreamingTextResponse(transformedStream); return new StreamingTextResponse(transformedStream);
} catch (error: unknown) { } catch (error: unknown) {