Research Upload

How to upload, manage, and process research sources in Praxiom AI.

Overview

Research sources are the foundation of everything in Praxiom AI. They represent the raw user voice — interviews, support tickets, surveys, usage data, and feedback — that the AI synthesizes into structured insights. This guide covers uploading, managing, and monitoring the processing of your research files.

How It Works

When you upload a file, Praxiom stores it in cloud storage (Cloudflare R2), creates a database record, and triggers background processing. Processing extracts text from PDFs, transcribes audio/video, and parses structured data from CSVs — making the content available for AI synthesis.

1

Upload a file

Send a multipart form request to POST /api/research/upload with the following fields:

FieldTypeDescription
fileFileThe research file (max 100 MB)
workspace_idUUIDTarget workspace
source_typeStringOne of: interview, support_ticket, survey, usage_data, feedback, other
titleStringA human-readable title for this source

The response returns a ResearchSourceResponse with the new source's id and a processing_status of pending.

2

Wait for processing

Processing runs in the background. Poll the source status via GET /api/research/sources/{source_id} or list all sources with GET /api/research/sources?workspace_id=....

The processing_status field transitions through:

  • pending — Queued for extraction
  • completed — Text extracted and ready for synthesis
  • failed — Extraction encountered an error
3

Synthesize

Once sources are completed, select them for synthesis. See the Synthesis guide for details.

Supported File Types

Praxiom accepts a wide range of document formats:

  • Documents — PDF, Word (.docx), plain text, Markdown, JSON
  • Spreadsheets — CSV, Excel (.xlsx)
  • Audio/Video — MP3, MP4, WAV, WebM (auto-transcribed)
  • Images — PNG, JPEG, TIFF (OCR-extracted)

Executable files (.exe, .sh, .py, .js, .bat, etc.) are blocked for security. The upload endpoint validates both MIME types and file magic bytes.

API Reference

List Sources

GET /api/research/sources?workspace_id={id}

Optional query parameters:

ParameterDescription
source_typeFilter by type (e.g. interview)
processing_statusFilter by status (e.g. completed)
limitMax results (default 200, max 500)
offsetPagination offset

Returns { sources: [...], total: N }.

Get Single Source

GET /api/research/sources/{source_id}

Returns the full source record including file_url, file_type, file_size_bytes, extracted_text, and transcript.

Delete Source

DELETE /api/research/sources/{source_id}

Removes the source record and its file from cloud storage. Returns 204 No Content.

Reprocess Source

POST /api/research/sources/{source_id}/reprocess

Resets a pending or failed source back to pending and retriggers background extraction. Useful when a source is stuck. Returns the updated source record.

Sources that have already completed successfully cannot be reprocessed. The endpoint returns 422 with code ALREADY_PROCESSED.

Key Concepts

  • Source types categorize your research for better synthesis context. The AI uses this metadata to weight different kinds of evidence appropriately.
  • File size limit is 100 MB per upload. Empty files are rejected with a 400.
  • Usage limits are enforced per workspace based on your plan. If you hit the file upload limit, the API returns 402 with code USAGE_LIMIT_EXCEEDED.
  • Workspace access is checked on every operation. You must be a member of the workspace to upload, view, or delete sources.

What's Next

Once your sources are processed, move on to the Synthesis guide to extract structured insights.

Was this helpful?

NextChat Agent