- type: connection
connection: 🌐 CHAT GPT-4O MINI
switchDuration: command
label: SWITCH TO CHATGPT
silent: true
- type: ask
param: query
message: 'Please provide a search query for web search:'
label: QUERY
- type: gpt
prompt: >-
# Multilingual Search Query Optimizer
You are a search optimization expert who creates effective search queries in
multiple languages. Your task is to generate optimized search queries based
on the user's input topic.
## Configuration:
- Output Language: {{language}} // This language should be the search terms.
- Input Query: {{query}}
## Instructions:
Generate at least four (4) effective search queries in the target language
that will yield comprehensive, accurate, and authoritative results when used
in search engines. Each query should be optimized for different search
intents related to the topic.
## Query Types to Include:
1. Informational query (basic facts and definitions)
2. In-depth research query (detailed analysis and expert content)
3. Source-specific query (authoritative websites or scholarly resources)
4. Specialized/niche aspect query (focusing on a particular angle)
## Output Format:
```
### Optimized Search Queries for: "{{query}}"
Search Term 1:
```
[Generated query in target language]
```
Intent: [Brief explanation of what this query targets]
Search Term 2:
```
[Generated query in target language]
```
Intent: [Brief explanation of what this query targets]
[Continue for all search terms]
```
## Additional Guidelines:
- Use appropriate search operators (site:, intitle:, filetype:, etc.) when
beneficial
- Incorporate synonyms and related terms to expand search coverage
- Consider cultural and linguistic nuances of the target language
- Focus on queries that will appear in high-quality search results
- For specialized topics (academic, technical, medical, legal, religious),
use proper terminology
This structure will help users find optimal search queries for any topic in
their desired language, significantly improving search result quality.
silent: true
label: MULTILINGUAL SEARCH QUERY OPTIMIZER
param: query
- type: gpt
prompt: >-
# Web Search Prompt
You are an expert in conducting web searches and summarizing relevant
information.
LANGUAGE: {{language}} // Use this parameter to specify the language of the
output
TASK: Perform a web search based on the user's input and create a
comprehensive answer to their question, adhering to the specific format
provided below.
CONTEXT: Web search results should be analyzed to provide concise and
accurate answers that address the user's query. The information should be
presented in an organized manner with links to relevant sources.
FORMAT:
## Key takeaway
Provide a single, most important takeaway from the web search results in
{{language}}.
## Detailed answer
Analyze and present detailed information that helps answer the question.
There should be no limit in words or bullet points to the report. Ensure
that all ideas, facts, and relevant information are concisely reported, and
the answer is comprehensive. Incorporate the maximum number of source links
within the text.
METHODOLOGY:
- Utilize information from the web search results provided.
- Extract as much useful information as possible from the web search
results.
- Use Markdown Format to increase readability, such as header etc.
- Integrate insights from multiple web search sources, and ensure inclusion
of relevant links in a markdown fully rendered format: [➊](URL), [➋](URL)
etc, to each bullet point.
- Every URL should correspond to only one symbol: ➊ ➋ ➌ ➍ ➎ ➏ ➐ ➑ ➒ ➓.
- If multiple sources repeat the same information, describe it and cite all
sources at once.
- If you know the answer, add your own knowledge to make it more complete.
- Do not hallucinate facts or information.
- Do not use any other tools.
- Avoid general phrases and be more specific and detailed.
CONSTRAINTS:
- **Strictly adhere to the format above.**
- Do not include any unnecessary introductory text. Directly present the key
takeaway and detailed answer.
- Use clear and concise language.
- Ensure all ideas and facts are relevant and clearly explained.
- Include source links only when they are directly relevant to the
information provided.
Please perform a web search using the following USER INPUT and use as MANY
sources as possible:
{{query}}
param: response
label: WEB SEARCH
silent: true
- type: gpt
prompt: >-
# Response Reformatting with Symbolic Citations
Reformat my previous response using the same content but with the following
specific citation format:
## FORMAT REQUIREMENTS:
1. Keep the "Key takeaway" section as a single, concise paragraph
2. In the "Detailed answer" section, cite all URLs using symbolic notation
3. Assign EXACTLY ONE symbol per unique URL: ➊ ➋ ➌ ➍ ➎ ➏ ➐ ➑ ➒ ➓
4. Format each citation as [➊](URL) where the symbol links to the source
5. When multiple points come from the same source, use the same symbol
6. If multiple sources contain the same information, cite all relevant
symbols together
7. Avoid empty line breaks in the output
8. Do not use line breaks between numbering, listing, bulletpoints etc.
## EXAMPLE:
Instead of:
"Create a Make.com Account: Visit Make.com to sign up.
([harpa.ai](https://harpa.ai/guides/connecting-harpa-to-make-com))"
Use:
"Create a Make.com Account: Visit Make.com to sign up
[➊](https://harpa.ai/guides/connecting-harpa-to-make-com)"
Simply reformat my previous response with these citation rules WITHOUT
adding or removing any information.
label: MINIMAL RESPONSE REFORMATTING
param: finalAnswer
- type: js
args: finalAnswer
code: |-
function formatMarkdownLinks(args) {
const text = args.finalAnswer || "";
if (!text) return "No text provided to extract links from.";
// Regex pattern for links with ➊, ➋, ➌ etc.
const linkPattern = /\[(➊|➋|➌|➍|➎|➏|➐|➑|➒|➓)\]\(([^)]+)\)/g;
// Use a map to store unique links by symbol and URL
const uniqueLinks = new Map();
// Find all matches
let match;
while ((match = linkPattern.exec(text)) !== null) {
try {
const symbol = match[1];
const url = match[2];
// Create a key that combines symbol and URL
const key = `${symbol}:${url}`;
// Only process this link if we haven't seen it before
if (!uniqueLinks.has(key)) {
const urlObj = new URL(url);
const fullPath = urlObj.hostname.replace('www.', '') + urlObj.pathname;
const shortPath = fullPath.length > 42
? fullPath.slice(0, 42) + '...'
: fullPath.replace(/\/$/, '');
uniqueLinks.set(key, `[${symbol}](${url}) [${shortPath}](${url})`);
}
} catch (e) {
// Handle invalid URLs
console.error("Error processing URL:", e);
}
}
// Return formatted output or a message if no links found
return uniqueLinks.size > 0
? Array.from(uniqueLinks.values()).join('\n')
: "No numbered reference links found in the text.";
}
return formatMarkdownLinks(args);
param: sources
timeout: 15000
onFailure: MINIMAL RESPONSE REFORMATTING
silent: true
label: EXTRACT SOURCES
- type: say
message: |-
 Sources
{{sources}}
label: DISPLAY SOURCES
- type: ask
param: custom
message: Do you want to work with the result or return to a new query?
options:
- label: 🔙 RETURN
value: return
- $custom
default: ''
vision:
enabled: false
mode: area
hint: ''
send: true
label: CUSTOM PROMPTING
optionsInvalid: false
- type: gpt
prompt: >-
# Direct Execution Prompt
You are a direct execution AI assistant. Your sole purpose is to perform the
user's requested action on the provided text.
LANGUAGE: {{language}}
TASK: Execute the specific instruction provided by the user directly on the
accompanying text.
CONTEXT: The user requires a direct transformation or processing of their
text according to their specific instruction. No interpretation,
elaboration, or additional information is needed unless explicitly
requested.
FORMAT: The output should be *only* the direct result of executing the
user's instruction on the provided text. Do not add any labels,
explanations, or surrounding text unless the user explicitly asks for it in
their request.
CONSTRAINTS:
- Strictly adhere to the user's instruction.
- Perform *only* the action specified by the user.
- Do not add any introductory phrases, concluding remarks, explanations, or
any content not explicitly requested.
- Output the result directly.
Execute the user's request on the following text:
{{custom}}
TEXT:
{{finalAnswer}}
silent: false
label: CUSTOM FURTHER PROCESSING
condition: '{{custom-option}} = $custom'
param: ''
- type: jump
to: QUERY
condition: '{{custom}} = return'
- type: jump
to: CUSTOM PROMPTING
This automation command is created by a community member. HARPA AI team does not audit community commands.
Please review the command carefully and only install if you trust the creator.
All rights reserved © HARPA AI TECHNOLOGIES LLC, 2021 — 2025
Designed and engineered in Finland 🇫🇮