Send PDFs, text files, CSVs, and other documents to any model via a unified API.

Document & File Input

osmAPI enables your applications to send documents and files to AI models through a unified interface. Whether you're processing PDFs, analyzing CSV data, or parsing HTML pages, our gateway handles format conversion and provider-specific transformations automatically.

Supported Content Formats

Two content block formats are supported for sending documents:

1. Document Format (Anthropic-style)

Use the document content type with a base64-encoded source or a URL reference.

curl -X POST "https://api.osmapi.com/v1/chat/completions" \
  -H "Authorization: Bearer $OSM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "Summarize this document."
          },
          {
            "type": "document",
            "source": {
              "type": "base64",
              "media_type": "application/pdf",
              "data": "JVBERi0xLjQKJcfs..."
            },
            "filename": "report.pdf"
          }
        ]
      }
    ]
  }'

URL Source

You can also reference documents via URL:

{
	"type": "document",
	"source": {
		"type": "url",
		"url": "https://example.com/report.pdf"
	},
	"filename": "report.pdf"
}

2. File Format (OpenRouter-style)

Use the file content type with a data URI or URL in file_data.

curl -X POST "https://api.osmapi.com/v1/chat/completions" \
  -H "Authorization: Bearer $OSM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "What does this CSV data show?"
          },
          {
            "type": "file",
            "file": {
              "filename": "sales.csv",
              "file_data": "data:text/csv;base64,bmFtZSxhZ2Usc2FsZXMKSm9obiwzMCwxMDAw..."
            }
          }
        ]
      }
    ]
  }'

Provider Handling

osmAPI automatically routes and transforms document content based on the target provider:

Provider	Handling	Supported Types
Anthropic	Native document blocks	PDF, TXT, CSV, HTML, MD
OpenAI	Native file support	PDF, TXT, CSV, and more
Google (Gemini)	Native inline_data / file_data	PDF, TXT, CSV, HTML
AWS Bedrock	Native Converse API document blocks	PDF, TXT, CSV, HTML
All other providers	Auto-conversion (see below)	Text-based formats only

Automatic Text Conversion

For providers that don't natively support document input (e.g., Groq, Cerebras, Together, DeepSeek, Neysa), osmAPI automatically converts text-based documents to plain text content blocks. This means your request works seamlessly — no code changes needed.

Text-based MIME types (auto-converted):

MIME Type	Extension	Auto-converted
`text/plain`	.txt	Yes
`text/csv`	.csv	Yes
`text/html`	.html	Yes
`text/markdown`	.md	Yes
`text/xml`	.xml	Yes
`text/css`	.css	Yes
`text/javascript`	.js	Yes
`application/json`	.json	Yes
`application/xml`	.xml	Yes
`application/javascript`	.js	Yes

Binary formats (not auto-converted):

MIME Type	Extension	Behavior
`application/pdf`	.pdf	Requires native provider support
`application/vnd.openxmlformats...`	.docx, .xlsx	Requires native provider support

Tip: If you need to send a PDF to a model that doesn't support documents natively, extract the text from the PDF on your side and send it as a text content block instead.

Anthropic API Compatibility

Documents are also supported via the Anthropic-compatible endpoint (/v1/messages):

curl -X POST "https://api.osmapi.com/v1/messages" \
  -H "x-api-key: $OSM_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 1024,
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "Analyze this document."
          },
          {
            "type": "document",
            "source": {
              "type": "base64",
              "media_type": "application/pdf",
              "data": "JVBERi0xLjQK..."
            },
            "filename": "analysis.pdf"
          }
        ]
      }
    ]
  }'

Multi-Document Requests

You can include multiple documents in a single message for comparative analysis:

{
	"role": "user",
	"content": [
		{ "type": "text", "text": "Compare these two reports." },
		{
			"type": "document",
			"source": { "type": "base64", "media_type": "text/csv", "data": "..." },
			"filename": "q1_sales.csv"
		},
		{
			"type": "document",
			"source": { "type": "base64", "media_type": "text/csv", "data": "..." },
			"filename": "q2_sales.csv"
		}
	]
}

Model Support

Models with native document support are indicated by the documents capability flag. You can find document-capable models via the Models page.

For models without native document support, text-based documents (TXT, CSV, HTML, MD, JSON, XML) are automatically converted to text and will work with any model.

Document & File Input

On this page