User Guide

Dashboard Overview

The Sound Suite dashboard is accessible at http://localhost:3000 and provides a complete interface for managing cases, processing documents, searching, drafting, and analyzing legal documents.

Navigation Sidebar

The left sidebar provides access to all major sections:

Icon Section Description
Cases / Main dashboard — view cases, documents, processing status
Search /search AI-powered search across all case documents
Draft /draft Rich text editor for writing legal documents
Vectors /vectors Explore the vector database and chunk data
Workflows /workflow Document templates and structured writing workflows
Explorer /case-explorer PDF viewer with table of contents and document tree
MCP /mcp-explorer Manage and test the 14 MCP analysis tools
Management /case-management Create cases, manage directories and filings
Admin /admin System settings, health monitoring, and configuration

System Health Bar

The top of the dashboard shows real-time service status:

  • File Watcher — monitors directories for new PDFs
  • Job Queue — processes documents through the ingestion pipeline
  • MCP Server — serves analysis tools to AI clients
  • Database — SQLite connection status

Each service shows a green dot when healthy, yellow for degraded, and red for errors.

Processing Status

The main dashboard view shows documents organized by case with status indicators:

  • INDEXED (green) — Fully processed and searchable
  • PARTIAL (orange) — Some pages processed, others pending or failed
  • QUEUED (yellow) — Waiting in the processing queue
  • ERROR (red) — Processing failed, may need attention

Managing Cases

Creating a Case

  1. Navigate to Case Management from the sidebar
  2. Click the + button in the top-right corner
  3. Enter a case name and optional case number
  4. Select the directory path containing your case PDF files
  5. Click Create — Sound Suite begins monitoring the directory immediately

Case Structure

Each case can contain multiple filings organized in subdirectories:

/your-case-directory/
  ├── Briefs/
  │   ├── appellants-brief.pdf
  │   └── appellees-brief.pdf
  ├── Clerks-Record/
  │   └── clerks-record-vol1.pdf
  └── Reporters-Record/
      ├── rr-vol1.pdf
      ├── rr-vol2.pdf
      └── rr-vol3.pdf

Browsing Case Files

From the Case Explorer (sidebar), you can:

  • Browse the file tree on the left panel
  • Click any PDF to view it in the built-in PDF viewer
  • View the Table of Contents (TOC) panel on the right
  • Access Document Info including page count, exhibit count, and processing status
  • Navigate pages using the page selector or scroll
  • Zoom in/out and switch between page and width-fit modes

Document Processing

How Processing Works

When Sound Suite detects a new PDF in a watched directory:

  1. Hash Check — Computes SHA-256 to detect duplicates
  2. Record Creation — Creates a document record with QUEUED status
  3. Text Extraction — Uses pdfjs-dist to extract text from each page
  4. OCR Pass — Pages with fewer than 50 characters trigger tesseract.js OCR
  5. Exhibit Detection — Identifies and extracts embedded images
  6. Text Chunking — Splits text into overlapping chunks (default: 512 tokens, 50 token overlap)
  7. Embedding — Generates vector embeddings using the configured provider
  8. Indexing — Stores embeddings in LanceDB for semantic search

Processing Settings

Click Processing Settings on the dashboard to configure:

  • Concurrency — How many documents to process simultaneously
  • Retry count — Maximum retries for failed documents
  • OCR settings — Enable/disable OCR, adjust density threshold
  • Chunk size — Tokens per chunk (affects search granularity)

Monitoring Progress

The dashboard shows real-time processing progress:

  • Queued/Processing column shows documents currently being ingested
  • Stopped column shows paused documents
  • Errors column shows failed documents with error details
  • Total progress indicator: "X of Y documents indexed"

Re-processing Documents

To re-process a document (e.g., after changing embedding models):

  1. Navigate to the document in the dashboard
  2. Click the document to view its details
  3. Use the re-process option to reset it to QUEUED status

Search

Sound Suite provides three search modes accessible from the Search page:

AI Search (Deep Search)

The primary search mode uses AI to analyze your question and search across all case documents:

  • Select an AI provider (Anthropic, OpenAI, or Ollama)
  • Choose a model (e.g., Claude Opus 4.6, GPT-4o)
  • Select the case to search within
  • Toggle Deep Search for multi-query semantic analysis
  • Toggle Thinking mode for transparent AI reasoning
  • Adjust token limit for response length

Deep Search decomposes your question into multiple sub-queries, performs parallel vector searches, and synthesizes a comprehensive answer with citations.

Direct Search

Performs direct vector similarity search without AI processing:

  • Enter a search query
  • Results show the most semantically similar text chunks
  • Each result shows the source document, page number, and relevance score
  • Click any result to view the source document in context

Analysis Tools

Run specialized analysis tools directly from the search interface:

  • Select from the 14 available MCP tools
  • Configure tool parameters
  • View structured results
  • Tools include: Detect Contradictions, Analyze Citations, Extract Timeline, and more

Draft Editor

The draft editor is a full-featured rich text editor for writing legal documents, accessible at /draft.

Editor Features

  • Rich text editing with full formatting toolbar
  • Import .docx — Import existing Word documents
  • Export .docx — Export to Word format for filing
  • Export .pdf — Generate PDF output
  • Outline panel — Navigate document structure via headings
  • Version history — Track changes across drafts
  • AI Chat — Ask questions about your draft or request revisions
  • AI Suggest — Get AI-powered writing suggestions in context

Toolbar

The ribbon-style toolbar is organized into groups:

  • Font — Font family, size, color
  • Headings — H1 through H6
  • Text Format — Bold, italic, underline, strikethrough, highlight
  • Lists & Alignment — Ordered/unordered lists, text alignment
  • Insert — Tables, images, links, page breaks
  • View & Track — Show/hide formatting marks, track changes

AI Integration

The editor's right panel provides AI features:

  • Chat — Conversational interface for asking questions about your document
  • History — Session history of AI interactions
  • Workflows — Apply document workflow templates
  • Suggest — Context-aware writing suggestions
  • Version — Draft version management

Workflows

Workflows are structured document templates with predefined sections. Select a workflow from the sidebar to generate a document scaffold:

  • Appeal Brief — Statement of Case, Issues Presented, Statement of Facts, Arguments, Prayer
  • Motion — Introduction, Background, Legal Standard, Argument, Conclusion
  • Response — Preliminary Statement, Response to Arguments, Conclusion

Workflows also include analysis templates for running tools on your documents.

Vectors Explorer

The Vectors page at /vectors lets you inspect the vector database directly:

Table View

Browse all text chunks stored in LanceDB:

  • Statistics bar — Total chunks, documents, cases, text chunks, exhibit chunks, avg chunks per document
  • Filters — Filter by case, filing, document, type, text content, page range
  • Chunk table — Shows text preview, page number, chunk index, type (Text/Exhibit), source document, indexed date
  • Pagination — Navigate through large datasets (50 chunks per page)

Breakdown View

Visual breakdown of how chunks are distributed across cases and documents.

Page Report

Detailed per-page analysis showing chunk coverage and potential gaps.

Provider Info

The top bar shows the current embedding configuration:

  • Provider name and status
  • Model identifier
  • Vector dimensions