Get answers from your files

Analyze workspace files to extract verified findings and generate a grounded answer with clickable citations.

Created by Chris Moen • Version 68 • 13 steps

Stop searching and start finding

Finding specific information across multiple documents and recordings takes time away from actual work. This workflow changes how you interact with your workspace by turning your files into a searchable knowledge base. Instead of skimming through long PDFs or listening to hours of audio, you can ask a direct question and receive a verified answer based only on your uploaded materials.

How the workflow processes your data

The process begins by sorting your uploads. It separates media files from text documents to ensure every file is handled correctly. If you've uploaded videos or audio recordings, the system automatically collects transcripts or flags files that require further processing. Once it confirms at least one file is readable, it prepares the data for a deep analysis.

To ensure accuracy, the flow uses OpenAI and Mistral models to synthesise an answer. It doesn't just summarise, it flattens verified findings and assigns specific reference IDs to every piece of information. The final step involves a grounding audit, where the Large Language Model checks its own work to make sure the response stays true to your source files. Finally, it renders a clean, cited answer that you can trust.

Why your team needs this template

This template is built for researchers, project managers, and analysts who need to maintain high standards of evidence. Because the system includes clickable citations, you can instantly verify where a specific fact came from, which eliminates the guesswork often associated with automated tools.

Key features and benefits:

  • Verified findings: Every claim includes a direct link to the source file, ensuring complete transparency.
  • Multi-format support: Process documents, transcripts, and media files in a single workspace.
  • Dual-model verification: Uses both OpenAI and Mistral to synthesise and audit answers for maximum reliability.
  • Time savings: Skip the manual skimming and get straight to the insights you need for your reports.
  • Reliable grounding: The built-in audit step prevents the system from making up information or pulling from outside sources.

If you are looking for a way to make your internal documentation more accessible, this flow provides a clear, reliable, and efficient solution.

Steps

  1. Split media uploads from analyzable resources (function)
  2. Collect transcript resources from uploaded media (function)
  3. Account for media files without transcripts (function)
  4. Check at least one file can be analyzed (function)
  5. Prepare analysis input (function)
  6. Use original question (function)
  7. Flatten verified findings (function)
  8. Assign reference ids (function)
  9. Format findings for cited answer (function)
  10. Synthesize cited answer (llm)
  11. Audit answer for grounding (llm)
  12. Parse cited answer (function)
  13. Render answer with references (function)

FAQ

What does the "Get answers from your files" workflow do?

This workflow analyses your workspace documents and media to provide specific answers based only on your data. It includes clickable citations, so you can verify exactly where each piece of information originated.

How does the automated analysis process work?

The system extracts text from documents and transcripts from media files. It then uses Large Language Models to identify key findings, assign reference IDs, and audit the final response to ensure every claim is grounded in your uploaded content.

Which services does this research workflow use?

This workflow connects with OpenAI and Mistral to process your queries and synthesise answers. It also uses built-in functions to handle file splitting, transcript collection, and citation formatting.

How do I set up this workflow to extract findings from my data?

You simply need to upload your files to the workspace and provide an original question. The workflow handles the technical steps, including media transcription and finding verification, to deliver a cited response automatically.