Skip to main content

​
What Gets Created

When Wolffish launches for the first time, it creates your workspace at ~/.wolffish/workspace/ and copies the default configuration files from the bundled defaults. Here’s what gets created:

​
The Workspace

~/.wolffish/workspace/
└── brain/
    ├── identity/
    │   ├── soul.md          # Agent personality
    │   └── user.md          # Your profile
    ├── prefrontal/
    │   └── agents.md        # Agent procedures and tool rules
    ├── cerebellum/          # Capabilities (shell, filesystem, git, etc.)
    ├── hippocampus/         # Memory (episodes, consolidated, knowledge)
    ├── motor/tasks/         # Task execution logs
    ├── basalganglia/        # Feedback logs
    ├── brainstem/           # Background process config
    ├── corpus/              # Event logs
    ├── conversations/       # Chat history
    └── config.json          # App configuration

​
Onboarding Flow

  1. Ollama Detection — Wolffish checks if Ollama is running at localhost:11434. If not found, it offers to guide you through installing it — or you can skip this step entirely and use cloud providers instead.
  2. Model Selection — If Ollama is available, pick a local model. Wolffish can pull models directly if they’re not already installed.
  3. Provider Configuration — Configure at least one provider: DeepSeek (recommended), Anthropic, or OpenAI API keys for cloud models, or Ollama for local inference. You need at least one to start chatting.

​
Customizing Your Agent

After the initial setup, you’ll want to edit two files:

user.md

Tell Wolffish about yourself — your name, preferences, projects, tech stack. The more context you provide, the more relevant the agent’s responses.

soul.md

Define the agent’s personality, tone, and behavioral guidelines. This is the character your agent takes on.

​
What Happens on Every Launch

After the first launch, each subsequent startup:
  1. Reads config.json for model and provider settings
  2. Initializes the 15 brain modules via the event bus (corpus)
  3. Indexes workspace markdown files into cortex.db (SQLite FTS5)
  4. Starts background processes (file watcher, cron scheduler)
  5. Checks provider health (Ollama, cloud APIs if configured)