Skip to content

Commit

Permalink
feat: better prompt enhancement
Browse files Browse the repository at this point in the history
  • Loading branch information
Sujal Shah committed Nov 26, 2024
1 parent d41a4ac commit e25bb28
Show file tree
Hide file tree
Showing 2 changed files with 22 additions and 3 deletions.
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ https://thinktank.ottomator.ai
- ✅ Ability to revert code to earlier version (@wonderwhy-er)
- ✅ Cohere Integration (@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** - Better prompting for smaller LLMs (code window sometimes doesn't start)
-**HIGH PRIORITY** - Load local projects into the app
Expand All @@ -43,7 +44,6 @@ https://thinktank.ottomator.ai
- ⬜ Vertex AI Integration
- ⬜ Deploy directly to Vercel/Netlify/other similar platforms
- ⬜ Prompt caching
- ⬜ Better prompt enhancing
- ⬜ Have LLM plan the project in a MD file for better results/transparency
- ⬜ VSCode Integration with git-like confirmations
- ⬜ Upload documents for knowledge - UI design templates, a code base to reference coding style, etc.
Expand Down
23 changes: 21 additions & 2 deletions app/routes/api.enhancer.ts
Original file line number Diff line number Diff line change
Expand Up @@ -44,9 +44,28 @@ async function enhancerAction({ context, request }: ActionFunctionArgs) {
content:
`[Model: ${model}]\n\n[Provider: ${providerName}]\n\n` +
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.
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>
${message}
Expand Down Expand Up @@ -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);
} catch (error: unknown) {
Expand Down

0 comments on commit e25bb28

Please sign in to comment.