Use SourceFusion to turn documents, web pages, files, and source text into RAG-ready JSONL, API JSON, extracted tables, metadata, source lineage, and AI-ready records.
Upload a file or paste source text. SourceFusion extracts text from text, HTML, PDF, DOCX, PPTX, and XLSX files in the web app; larger source systems and private media routes are provisioned through business deployment.
DeepWave signal datasets, private connectors, VPC deployment, and customer-specific source domains remain separate product lines.
SourceFusion covers the processing path teams expect for RAG, model evaluation, internal search, data products, and governed API delivery.
Run SourceFusion through the hosted app, cloud marketplaces, or private infrastructure.
Bring source material into one governed processing path.
Use common enterprise and data-platform sources with customer-controlled credentials.
Convert unstructured and semi-structured material into AI-ready records.
Register common enterprise file types for extraction, chunking, enrichment, and delivery.
Split documents into useful units before chunking or extraction.
Prepare retrieval units for RAG and downstream model workflows.
Add context that makes records useful after extraction.
Hand off clean chunks to embedding and retrieval systems.
Connect the processing output to the AI platforms teams already use.
Deliver outputs into files, APIs, data warehouses, or retrieval systems.
Support downstream data and retrieval platforms.
Operate repeatable processing jobs instead of one-off scripts.
Protect customer inputs and processing credentials.
Support enterprise review and procurement.
SourceFusion is designed for customer-controlled credentials, encrypted transport, permissioned access, source-level manifests, and customer-controlled retention. Enterprise deployments can be delivered as hosted SaaS, a dedicated instance, a private marketplace plan, or a customer VPC deployment.
Uploaded customer documents are processed for the customer job and are not used to train DeepWave models.
Privacy mode avoids stored source/chunk previews. Stored previews can be purged from the job record by the account owner.
Business accounts can use DeepWave hosted processing, private AWS/Azure procurement, or customer cloud deployment.
The estimate endpoint is public for pricing previews. Job creation requires an authenticated account so page allowance and usage can be tracked.
POST /api/sourcefusion/estimate
{
"sourceName": "Policy archive",
"sourceType": "PDF / report",
"requestedPages": 18000,
"outputFormat": "RAG JSONL + API JSON",
"privacyMode": true
}
POST /api/sourcefusion/jobs
Authorization: DeepWave account session
{
"sourceName": "Policy archive",
"sourceType": "PDF / report",
"requestedPages": 18000,
"outputFormat": "RAG JSONL + API JSON",
"privacyMode": true,
"sourceText": "Optional source text for immediate chunking"
}