Document Support in Sparkforge AI Knowledge Bases

By Sparkforge Support Team ·Last reviewed May 3, 2026

Sparkforge Knowledge Bases now ingest DOC, DOCX, and PDF files plus content typed directly in a rich-text editor. Sparkforge auto-parses headings into chunks so AI agents retrieve precise answers from your existing SOPs and product docs.

Sparkforge AI Knowledge Bases now accept document uploads and rich-text entries alongside URLs and FAQs. Upload a Word or PDF file, and Sparkforge extracts text, detects headings, and chunks the content for accurate retrieval inside your AI agents.

Supported formats and behavior

  • File types: DOC, DOCX, PDF.
  • Text-only ingestion. Embedded images are skipped to keep knowledge searchable and lightweight.
  • Heading awareness. H1–H4 hierarchy improves chunking and retrieval relevance.
  • Instant indexing. New content is available within about a minute.

Upload a document

  1. Go to AI Agents from your sub-account.
  2. Click the Knowledge Base tab.
  3. Edit an existing Knowledge Base, or click Create New and give it a name and description.
  4. Click Add Source → File Upload.
  5. Drag and drop or browse for DOC, DOCX, or PDF files.
  6. Click Upload Files. Status shows Processed when parsing finishes.

Use the arrow icon to open an uploaded file in a new tab, or the bin icon to remove it.

Add rich-text content

Click Add Source → Rich-Text to write or paste content directly. Click Save to push it into the Knowledge Base.

How parsing improves answers

  • Extracts text and detects headings to form a logical hierarchy.
  • Splits large files into optimized chunks for faster, more relevant retrieval.
  • Preserves document-level context so related sections stay connected.
  • Indexes immediately so agents use the latest content.

Prep tips for clean ingestion

  • Remove sensitive content before uploading.
  • Make sure headings reflect the actual document structure — clear headings produce better chunks.
  • Combine documents and rich text with URLs and FAQs to strengthen training data.

Frequently Asked Questions

Do images inside my Word or PDF import?
No. Only text is ingested. Images are skipped to keep the knowledge base searchable and lightweight.
How fast do uploads become available?
Most files index within about a minute. Larger files may take slightly longer.
Can I upload Google Docs directly?
Not yet. Export the doc as .DOCX or .PDF and upload the exported file.
Do file uploads overwrite existing entries?
No. Each upload is added as a new source. File-based sources, rich-text entries, URLs, and FAQs all coexist.
Will rich-text edits replace my uploaded files?
No. Files and rich-text entries are stored separately, so both versions stay intact.

Was this article helpful?

Related Articles