Dawson-Core (v0.1.0)
Pre-releaseInstallation
DAWSON currently supports:
- macOS (full and tested)
- Linux (partial, untested)
After downloading DAWSON Core, install it by running:
python3 setup_dawson.py install(If you'd like to just test it temporarily, feel free to simply run the executable file in your terminal.)
The installer will:
- Configure DAWSON to automatically start when you log in.
- Create any required configuration files.
- Generate a TLS certificate for encrypted communication.
- Generate a unique authentication token.
- Start the DAWSON server.
Once installation is complete, DAWSON will continue running automatically in the background.
Connecting Beakshield
After DAWSON is running:
-
Open Beakshield.
-
Open System → DAWSON Server → Manage Connection.
-
Enter:
- IP Address: The host computer running DAWSON
- Port:
8443 - Authentication Token
- Certificate Fingerprint
The server currently always listens on port 8443.
The Authentication Token and Certificate Fingerprint are automatically generated the first time DAWSON starts. They are stored in:
DAWSON/
└── databank/
└── security/
├── auth-token.txt
└── auth-fingerprint.txt
Open these two files and copy their contents into the corresponding fields in Beakshield when adding your DAWSON server.
The authentication token acts as the password for your DAWSON server. Only clients that know this token are allowed to connect.
The certificate fingerprint allows Beakshield to verify that it is communicating with the correct DAWSON server rather than another machine impersonating it.
Once all four fields have been entered, save the connection. Beakshield will establish an encrypted and authenticated connection to your DAWSON server.
Configuring AI Providers
Before creating your first chat, you'll need to configure at least one AI provider.
In Beakshield, open:
System → Providers
Select the provider you want to configure and enter its API key (if required). Once saved, Beakshield will automatically retrieve the list of available models from that provider.
OpenAI
To use OpenAI models (such as GPT models), you'll first need an OpenAI account and API key.
-
Create an OpenAI account if you don't already have one.
-
Generate an API key from:
-
Paste the API key into the OpenAI provider configuration in Beakshield.
-
Click Save.
Anthropic
To use Claude models, you'll need an Anthropic account and API key.
-
Create an Anthropic account.
-
Generate an API key from:
-
Paste the API key into the Anthropic provider configuration in Beakshield.
-
Click Save.
Ollama
Ollama allows DAWSON to use locally hosted AI models running on your own computer.
Unlike cloud providers, Ollama does not require an API key.
Simply install Ollama from:
Download any models you wish to use and ensure the Ollama server is running on the same IP address as DAWSON.
After selecting the Ollama provider in Beakshield, simply click Save. DAWSON will automatically discover the locally installed models.
Note: At this time, Ollama support is limited to local models running on the same machine as DAWSON.
Your First Chat
After connecting:
- Press Start New Chat (on the home screen), or open the Chats page.
- Choose the model you want to use.
- Select the desired mode.
- Start chatting.
If you want DAWSON to work with files, attach one or more directories to the chat. Those directories become that chat's workspace and determine what the Squirebot is allowed to access depending on the selected mode.
Agent Modes
Each chat has a mode that determines what the Squirebot agent is allowed to do.
🥚 Egg
Egg mode is a standard AI chat.
The agent cannot:
- Read files
- Write files
- Execute commands
- Access your computer
Use this mode for normal conversations, brainstorming, writing, and general questions.
🐣 Fledgling
Fledgling introduces file access while keeping you in control.
The agent can:
- Read files inside the attached workspace
- Request approval before reading files outside the workspace
- Request approval before writing files inside the workspace
The agent cannot:
- Write outside the workspace
- Execute commands
- Use elevated privileges
This is the recommended mode for most development work.
⚔️ Warrior
Warrior mode allows the agent to work more independently within the attached workspace.
The agent can:
- Read files inside the workspace
- Write files inside the workspace
If it needs to read files outside the workspace, it will request your approval.
The agent cannot:
- Execute commands
- Use elevated privileges
- Perform unrestricted system access
Warrior mode is ideal for larger coding tasks where the AI needs to edit many files without constantly asking for permission.
👑 Ultimate
Ultimate removes the normal safety restrictions.
The agent can:
- Read any file
- Write any file
- Execute commands
- Perform unrestricted system operations
Ultimate is still in development due to its broad system access and is currently not available for use.
Removing a Provider
If you no longer want a provider's models to appear in Beakshield:
- Open System → Providers.
- Select the provider.
- Clear the API key.
- Click Save.
- Restart Beakshield.
Once Beakshield reconnects to DAWSON, the provider's models will no longer be available for new chats.
Note: Since Ollama does not use an API key, its models cannot be disabled this way. To remove Ollama models from Beakshield, stop the local Ollama server or uninstall Ollama.
Uninstalling
To remove DAWSON from your system:
python3 setup_dawson.py uninstallThis removes the background service and prevents DAWSON from automatically starting.
Your DAWSON directory, security credentials, and stored data are left intact unless you remove them manually.
Need Help?
If Beakshield cannot connect:
- Verify DAWSON is running.
- Confirm you entered the correct IP address.
- Verify port 8443 is being used.
- Double-check that both the Authentication Token and Certificate Fingerprint exactly match the values stored in
DAWSON/databank/security. - Ensure your firewall allows incoming connections on port 8443.