Optimized for Hermes

Document ingestion for your Hermes agent

Hermes knows how to remember. BlazeDocs makes sure what it remembers is actually useful. Convert messy PDFs into clean Markdown before they hit your knowledge pipeline — better structure, better retrieval, better answers.

Connect BlazeDocs to Hermes View API Docs
One-click Connect setup
Free tier included
Human-approved, agent-initiated

Why raw PDFs poison knowledge bases

Garbage in, garbage out
Your knowledge agent is only as good as what it ingests
  • ✗Broken reading order corrupts context
  • ✗Scanned PDFs return blank or garbled text
  • ✗Tables become meaningless word soup
Wasted embeddings
Bad structure = bad retrieval = wrong answers
  • ✗Headers/footers pollute every chunk
  • ✗No heading hierarchy for smart splitting
  • ✗Encoding artifacts break tokenizers
Structured Markdown works
Clean input enables better retrieval and reasoning
  • Heading hierarchy for smart chunking
  • Tables become queryable Markdown
  • Scanned docs run through AI OCR first

Connect BlazeDocs to Hermes in minutes

No manual API key generation or dashboard navigation. Agent initiates, you approve.

1

Hermes requests a connection

Your agent creates a short-lived Connect session with BlazeDocs, specifying it's acting on behalf of Hermes.

2

You approve in a hosted page

A browser page opens where you review and approve the connection. Nothing happens until you say yes — you're always in control.

3

Hermes can now convert documents

BlazeDocs mints an API key and returns it to Hermes. From that point, any PDF your agent encounters gets converted to clean Markdown automatically.

How BlazeDocs fits your knowledge workflow

The pipeline

PDF → BlazeDocs → Markdown → Hermes Memory

1. Agent finds a PDF (email, web, local file)
2. Sends it to BlazeDocs /api/v1/convert
3. Gets back clean, structured Markdown
4. Feeds that Markdown into Hermes memory
5. Now it can actually reason about the content

Python integration

from blazedocs import Client

client = Client(api_key=BLAZEDOCS_API_KEY)

# Convert a PDF before ingesting into Hermes
result = client.convert(
    file_path="research_paper.pdf",
    output_format="markdown"
)

markdown = result["data"]["markdown"]
pages = result["data"]["page_count"]

# Feed clean Markdown into Hermes memory
hermes.ingest({
    "source": "research_paper.pdf",
    "content": markdown,
    "metadata": {
        "pages": pages,
        "tokens": result["data"]["token_count"]
    }
})

Why Hermes + BlazeDocs is better than raw PDF ingestion

Better retrieval

Clean Markdown with proper heading structure means smarter chunking and more accurate vector search results. Your agent finds the right information, faster.

Scanned PDFs handled

AI OCR runs on scanned PDFs before conversion. No more blank pages or "no text found" dead ends destroying your agent's workflow.

Token-efficient memory

Remove layout noise, headers, footers, and duplicated text. Store only what matters — your memory database shrinks and retrieval quality improves.

Metadata for context

Page counts, token estimates, and processing metadata come back with every conversion. Use them to build smarter ingestion pipelines.

What your agent gets back

Everything needed: Markdown, page count, token estimate, processing time.

POST /api/v1/convert
{
  "success": true,
  "data": {
    "markdown": "# Research Paper\n\n## Abstract\n...",
    "page_count": 15,
    "token_count": 12340,
    "processing_time_ms": 4200,
    "file_name": "research_paper.pdf"
  }
}

Give your knowledge agent better input

Clean documents in = better reasoning out. Connect BlazeDocs to Hermes today.

Connect BlazeDocs to Hermes Read API Docs