A multilingual research workflow
Turning a question into a structured research response in the language of the request, using n8n and Perplexity.
- n8n
- Perplexity
- Webhooks
- LLM APIs
My role
I connected the request intake, language handling and research step in n8n. Perplexity supplies the research capability; my work is the workflow and the instructions that shape its response.
The problem
A research request should not require manually moving the question between tools and reformatting every answer. I wanted a consistent response structure while keeping Bulgarian and English requests in their original language.
How it works
A webhook receives the question
The AI step identifies the request’s language
Perplexity produces a research response
The answer is organised into overview, main findings and conclusion
Technical decisions
Use a repeatable intake
A webhook gives the process a consistent entry point. n8n connects that input to the research step, so the process can be invoked from another tool.
Keep language handling inside the workflow
The instructions ask for a response in the detected language. This avoids maintaining separate versions of the workflow for Bulgarian and English requests.
Give the answer a predictable structure
An overview, main findings and conclusion make the response easier to review. This is an output format, not a guarantee that the underlying information is correct.
Outcome and limitations
The workflow connects a research question to a formatted response through one repeatable path. The integration makes the research service accessible from other tools through a consistent entry point.
Generated findings and any cited sources still need checking. Ambiguous or mixed-language requests can lead to an unsuitable response. I review the response against the question and the sources before using it.