Workspace Structure

Rowboat stores all data locally in the ~/.rowboat/ directory. Everything is plain text - Markdown, JSON, and git history. You can inspect, edit, back up, or delete any file at any time.

Directory Overview

~/.rowboat/
├── knowledge/              # Your knowledge graph (Obsidian-compatible)
│   ├── People/            # Notes about individuals
│   ├── Organizations/     # Notes about companies and teams
│   ├── Projects/          # Notes about initiatives and workstreams
│   ├── Topics/            # Notes about recurring themes
│   └── Welcome.md         # Introduction to your knowledge graph
├── agents/                # Background agent state and outputs
├── config/                # Configuration files
│   ├── models.json        # AI model configuration
│   ├── mcp.json           # MCP server configuration
│   ├── security.json      # Allowed shell commands
│   ├── note_creation.json # Knowledge graph strictness
│   ├── agent_schedule.json # Background agent schedules
│   ├── deepgram.json      # Voice transcription (optional)
│   ├── brave-search.json  # Web search (optional)
│   └── exa-search.json    # Research search (optional)
├── sources/               # Raw email and meeting data
│   ├── gmail/             # Gmail threads
│   ├── fireflies/         # Fireflies transcripts
│   └── granola/           # Granola meeting notes
├── .git/                  # Version control for knowledge graph
└── knowledge_graph_state.json  # Change tracking for graph builder

Knowledge Graph

The knowledge/ directory is your work memory - an Obsidian-compatible vault with Markdown notes and backlinks.

People/

Notes about individuals you work with. Example: knowledge/People/Sarah Chen.md

# Sarah Chen

## Role
VP of Product at [[Acme Corp]]

## Recent Interactions

### 2026-01-15: Product roadmap discussion
- Wants to prioritize mobile app for Q2
- Concerned about API latency for international users
- Committed to providing designs by end of January

### 2025-12-10: Kickoff meeting
- Introduced the new authentication requirements
- Asked about our deployment timeline

## Open Questions
- What's the timeline for the API performance improvements?
- Do we need additional resources for mobile?

## Commitments

**From Sarah:**
- Design mockups by Jan 31
- Finalize scope for Q2 sprint by Feb 5

**From us:**
- Provide API benchmark results
- Share deployment schedule for mobile backend

Backlinks like [[Acme Corp]] connect notes together. Clicking a backlink in Obsidian (or Rowboat) navigates to that note.

Organizations/

Notes about companies, teams, and groups. Example: knowledge/Organizations/Acme Corp.md

# Acme Corp

## Key Contacts
- [[Sarah Chen]] - VP of Product
- [[Michael Park]] - CTO
- [[Lisa Wang]] - Engineering Manager

## Active Projects
- [[Mobile App Launch]] (Q2 2026)
- [[API Performance Optimization]] (ongoing)

## Relationship History
- Customer since 2024-06
- Contract renewal coming up in 2026-07
- Currently on Enterprise plan

## Recent Decisions
- 2026-01-15: Agreed to prioritize mobile over desktop features
- 2025-12-01: Extended contract through 2027

Projects/

Notes about ongoing initiatives, workstreams, and goals. Example: knowledge/Projects/Mobile App Launch.md

# Mobile App Launch

## Overview
Launch native mobile apps (iOS and Android) for [[Acme Corp]] by Q2 2026.

## Team
- [[Sarah Chen]] (Product lead)
- [[Alex Kim]] (iOS dev)
- [[Jordan Lee]] (Android dev)

## Milestones
- [ ] Design finalized - Jan 31
- [ ] API endpoints ready - Feb 15
- [ ] Beta build - Mar 1
- [ ] TestFlight/Play Console - Mar 15
- [ ] Public launch - Apr 1

## Decisions
- 2026-01-15: Using React Native for both platforms
- 2026-01-10: Authentication will use OAuth 2.0

## Blockers
- API latency for international users
- Need design assets from [[Sarah Chen]]

## Next Steps
- Finalize authentication flow
- Set up CI/CD pipeline for mobile builds

Topics/

Notes about recurring themes, concepts, and subject areas. Example: knowledge/Topics/API Architecture.md

# API Architecture

## Overview
Our REST API design patterns and standards.

## Related Projects
- [[Mobile App Launch]]
- [[API Performance Optimization]]

## Key Principles
- RESTful endpoints with versioning (v1, v2)
- JSON for request/response bodies
- OAuth 2.0 for authentication
- Rate limiting: 1000 requests/hour

## Recent Discussions
- 2026-01-15: [[Sarah Chen]] raised concerns about latency
- 2025-12-20: Decided to add caching layer for common queries

## Resources
- [API Documentation](https://docs.example.com/api)
- OpenAPI spec: `/api/openapi.json`

Welcome.md

An introduction file created on first run:

# Welcome to Rowboat

This vault is your work memory.

Rowboat extracts context from your emails and meetings and turns it
into long-lived, editable Markdown notes...

Configuration

The config/ directory contains all settings.

models.json

AI model configuration

{
  "provider": {
    "flavor": "openai",
    "apiKey": "sk-proj-..."
  },
  "model": "gpt-5.2",
  "knowledgeGraphModel": "gpt-4.1"
}

Provider Flavors
Fields

openai - OpenAI GPT models
anthropic - Anthropic Claude models
google - Google Gemini models
ollama - Local models via Ollama
openrouter - OpenRouter (multiple models, one key)
aigateway - Vercel AI Gateway
openai-compatible - Any OpenAI-compatible API

flavor - Provider type (required)
apiKey - API key for cloud providers
baseURL - Custom endpoint (optional for cloud, required for local)
headers - Custom HTTP headers (optional)
model - Main model for assistant
knowledgeGraphModel - Separate model for graph operations (optional)

mcp.json

MCP server configuration

{
  "mcpServers": {
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {
        "BRAVE_API_KEY": "BSA..."
      }
    },
    "slack": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-...",
        "SLACK_TEAM_ID": "T..."
      }
    }
  }
}

Always use the Settings dialog or addMcpServer builtin tool to modify this file. Manual edits may break validation.

security.json

Allowed shell commands Commands on this list run immediately without user approval:

{
  "allowedCommands": [
    "ls",
    "cat",
    "git status",
    "git log",
    "npm install",
    "python --version"
  ]
}

Commands not on the list trigger a one-time approval prompt.

note_creation.json

Knowledge graph strictness

{
  "strictness": "medium",
  "configured": true
}

Strictness Levels
Configuration

high - Meetings create notes. Emails only enrich existing notes.
medium - Both create notes, but emails require personalized content.
low - Capture broadly. Create notes for any identifiable human sender.

strictness - The level (“high”, “medium”, or “low”)
configured - Whether auto-analysis has run (set to false to re-analyze)

agent_schedule.json

Background agent schedules

{
  "agents": [
    {
      "id": "daily-briefing",
      "name": "Daily Briefing",
      "enabled": true,
      "schedule": "0 8 * * *",
      "instructions": "Generate a voice note with today's agenda..."
    }
  ]
}

Each agent has:

id - Unique identifier
name - Human-readable name
enabled - Whether the agent is active
schedule - Cron expression for when to run
instructions - What the agent should do

Optional Service Configs

{
  "apiKey": "your-deepgram-api-key"
}

Sources

The sources/ directory contains raw emails and meeting transcripts synced from external services.

gmail/

Gmail threads saved as Markdown:

sources/gmail/
├── thread_abc123.md
├── thread_def456.md
└── ...

Each file contains:

Thread subject
Participants
Message timestamps
Full message content

fireflies/

Fireflies meeting transcripts:

sources/fireflies/
├── meeting_2026-01-15.md
├── meeting_2026-01-14.md
└── ...

granola/

Granola meeting notes:

sources/granola/
├── standup_2026-01-15.md
├── planning_2026-01-14.md
└── ...

Source files are not meant for direct editing. They’re processed by the knowledge graph builder to create and update notes in knowledge/.

Agents

The agents/ directory contains state and outputs from background agents.

agents/
├── daily-briefing/
│   ├── state.json
│   └── outputs/
│       ├── 2026-01-15.md
│       └── 2026-01-14.md
└── email-drafter/
    ├── state.json
    └── drafts/
        ├── reply_to_sarah.md
        └── ...

Each agent gets its own subdirectory for:

state.json - Last run time, success/failure, etc.
outputs/ - Generated content (notes, drafts, reports)

Version Control

Rowboat automatically tracks changes to your knowledge graph using git. Location: ~/.rowboat/.git/

What’s Tracked

All notes in knowledge/
Automatic commits when notes are created or updated
Full git history for diffs and rollbacks

Using Git

You can use standard git commands:

cd ~/.rowboat

# View history
git log --oneline

# See what changed
git diff HEAD~1 knowledge/People/Sarah\ Chen.md

# Undo changes
git checkout HEAD~1 -- knowledge/People/Sarah\ Chen.md

# View all changes in the last day
git log --since="1 day ago" --stat

Rowboat’s version history is automatic - you don’t need to commit manually. But you can use git for advanced operations if needed.

State Files

knowledge_graph_state.json

Change tracking for graph builder

{
  "processedFiles": {
    "/full/path/to/sources/gmail/thread_abc.md": {
      "mtime": "2026-01-15T10:30:00.000Z",
      "hash": "a3f5e9d2c8b1...",
      "lastProcessed": "2026-01-15T10:35:00.000Z"
    }
  },
  "lastBuildTime": "2026-01-15T10:35:00.000Z"
}

This file tracks which source files have been processed to build the knowledge graph. Fields:

mtime - File modification time
hash - SHA-256 hash of file contents
lastProcessed - When the file was last processed
lastBuildTime - When the graph builder last ran

Delete this file to force reprocessing of all sources (useful for testing new strictness levels).

Backup and Portability

Since everything is plain text, you can:

Back Up

# Full backup
tar -czf rowboat-backup.tar.gz ~/.rowboat/

# Knowledge graph only
tar -czf knowledge-backup.tar.gz ~/.rowboat/knowledge/

# Use Time Machine, Backblaze, etc.
# (they'll automatically include ~/.rowboat/)

Restore

tar -xzf rowboat-backup.tar.gz -C ~/

Sync Across Machines

# Machine 1
cd ~/.rowboat/knowledge
git remote add origin [email protected]:you/rowboat-knowledge.git
git push -u origin main

# Machine 2
cd ~/.rowboat
rm -rf knowledge
git clone [email protected]:you/rowboat-knowledge.git knowledge

Be careful with API keys! If syncing ~/.rowboat/ to cloud storage, exclude config/*.json files containing sensitive credentials.

Open in Obsidian

Install Obsidian

Download from obsidian.md

Open as vault

File → Open vault → Open folder → Select ~/.rowboat/knowledge/

Browse and edit

All your notes appear with working backlinks. Edits sync back to Rowboat automatically.

Disk Usage

Typical disk usage:

knowledge/ - 5-50 MB (depends on how many notes)
sources/ - 50-500 MB (raw emails and transcripts)
.git/ - 10-100 MB (version history)
config/ - Less than 1 MB
agents/ - 1-10 MB

Total: Usually under 500 MB

If disk usage grows too large, you can archive old sources or prune git history.

Getting Started

Core Concepts

Desktop App

Integrations

Use Cases

Configuration

​Directory Overview

​Knowledge Graph

​People/

​Organizations/

​Projects/

​Topics/

​Welcome.md

​Configuration

​models.json

​mcp.json

​security.json

​note_creation.json

​agent_schedule.json

​Optional Service Configs

​Sources

​gmail/

​fireflies/

​granola/

​Agents

​Version Control

​What’s Tracked

​Using Git

​State Files

​knowledge_graph_state.json

​Backup and Portability

​Back Up

​Restore

​Sync Across Machines

​Open in Obsidian

​Disk Usage

​Next Steps

Configure AI Models

Explore Features

Build docs developers (and LLMs) love

Directory Overview

Knowledge Graph

People/

Organizations/

Projects/

Topics/

Welcome.md

Configuration

models.json

mcp.json

security.json

note_creation.json

agent_schedule.json

Optional Service Configs

Sources

gmail/

fireflies/

granola/

Agents

Version Control

What’s Tracked

Using Git

State Files

knowledge_graph_state.json

Backup and Portability

Back Up

Restore

Sync Across Machines

Open in Obsidian

Disk Usage

Next Steps