Skip to content
Seahorse Docs

Getting Started

From signup to your first cognitive cycle in under 5 minutes.

1

Create Your Account

Go to seahorse.tools/dashboard/register and enter your email and password (minimum 8 characters).

After registration, you will receive:

  • A JWT token for dashboard authentication (stored automatically)
  • An API key (prefixed with sh_) for programmatic access
Important: Your API key is shown only once after registration. Copy it and store it securely. You can regenerate it later from the Settings modal, but the old key will be invalidated.
2

Add Your Documents

After logging in, choose how you want to bring documents into Seahorse. You can use one, two, or all three methods:

Option A: Connect Your RAG

The onboarding wizard walks you through connecting your vector database. Seahorse supports four providers:

Provider Required Fields Notes
Qdrant URL, Collection name Cloud or self-hosted. Uses query_points() API.
Pinecone API key, Index name Batched fetch (1000 IDs max), paginated listing.
Weaviate URL, Collection name, API key (optional) Cloud and local instances supported. v4 SDK.
ChromaDB URL, Collection name HTTP client mode. Supports remote instances.

Select your provider, fill in the connection details, and click Connect. Seahorse will verify the connection and report the number of documents found.

Option B: Upload PDFs

Go to the Home page and drag-and-drop PDF files into the upload zone (or click to browse). Seahorse auto-chunks and embeds the content. No vector database needed. Your chunk quota depends on your plan (Free: 100 chunks, Starter: 5,000, Pro: 50,000, Enterprise: unlimited).

Use the source type toggle to choose between PDFs and RAG for the upload target. Uploaded PDFs are immediately available for cognitive analysis and chat.

Option C: Document Editor

Go to Documents in the sidebar. Upload a single PDF to open it in the WYSIWYG editor. The cognitive engine runs analysis on the document and surfaces findings as inline annotations. You can edit the text directly and export the corrected version.

Security note: Seahorse validates all RAG connection URLs against private/internal IP ranges (SSRF protection). Only publicly accessible endpoints are accepted. Your connection details are stored securely and used only to read and (optionally) write to your vector database.
3

Run Your First Scan

Navigate to the Intelligence page and click Run Cognitive Cycle. You can choose a tier level for the analysis:

Tier Lenses Used Approx. Cost
Free Orphans + Obsolescence (pure logic, no LLM) ~$0.02
Pro All structural lenses (adds Contradictions + Gaps with LLM) ~$0.07
Enterprise Structural + OmniQ perception lenses (Monade, Symbiote, Architect, Empath) ~$0.10-0.13

The cognitive cycle runs in the background. You can watch the progress bar in real time on the Intelligence page. A full cycle on a 1,000-document RAG takes approximately 2-5 minutes depending on the tier.

The cycle runs up to 5 iterations. Each iteration scans, analyzes, evaluates, generates perception rules, and optionally applies fixes. The system stops automatically when health metrics converge (stabilize).

4

Explore Your Results

After the cycle completes, you have several views to explore:

Home Dashboard

See your health score (0-100), document count, active observations, and a timeline graph tracking health across cycles.

Observations

Browse every finding from every lens. Observations are audited for accuracy -- suspect findings are filtered from metrics. Filter by type, view source documents, approve, reject, or apply fixes.

Documents

View and manage your editable documents. Open any document in the WYSIWYG editor to see cognitive annotations inline, edit the text, and export corrected versions.

Chat

Ask Seahorse questions in natural language. Use the source toggle (PDFs / RAG / All) to control where the search looks. Responses include clickable source citations with file names and page numbers.

For a detailed walkthrough of every dashboard page, see the Dashboard Guide.

Next Steps

  • Upload PDFs to start analyzing documents without a vector database.
  • Try the Document Editor for focused single-document analysis with inline annotations.
  • Configure write-back mode to let Seahorse apply fixes automatically. See Dashboard Guide.
  • Enable Smart Query for better search results across your documents.
  • Set up the API for programmatic access. See API Reference.
  • Upgrade your tier to unlock LLM-powered lenses and OmniQ perception. See Pricing.