Skip to content

CONFIG.md

Latest commit

 

History

History
234 lines (173 loc) · 6.98 KB

CONFIG.md

File metadata and controls

234 lines (173 loc) · 6.98 KB

Configuration Guide

Terminal Intelligence supports configuration through a JSON file located at ~/.ti/config.json (or %USERPROFILE%\.ti\config.json on Windows).

Configuration File

On first run, if no config file exists, the application will automatically create a default config.json with example values at ~/.ti/config.json.

You can edit this file to customize your settings. The application will load it automatically on subsequent runs.

Configuration Examples

Ollama Configuration (Default)

{
  "agent": "ollama",
  "model": "qwen2.5-coder:3b",
  "ollama_url": "http://localhost:11434",
  "gemini_api": "",
  "gemini_model": "",
  "bedrock_api": "",
  "bedrock_model": "",
  "bedrock_region": "",
  "workspace": "/home/user/ti-workspace"
}

Gemini Configuration

{
  "agent": "gemini",
  "model": "",
  "ollama_url": "",
  "gemini_api": "your-gemini-api-key-here",
  "gemini_model": "gemini-3.1-flash-lite",
  "bedrock_api": "",
  "bedrock_model": "",
  "bedrock_region": "",
  "workspace": "/home/user/project-workspace"
}

AWS Bedrock Configuration

{
  "agent": "bedrock",
  "model": "",
  "ollama_url": "",
  "gemini_api": "",
  "gemini_model": "",
  "bedrock_api": "ACCESS_KEY_ID:SECRET_ACCESS_KEY",
  "bedrock_model": "anthropic.claude-sonnet-4-6",
  "bedrock_region": "us-east-1",
  "workspace": "/home/user/project-workspace"
}

Configuration Fields

Core Settings

  • agent (string, required): AI provider to use

    • Valid values: "ollama", "gemini", "bedrock"
    • Determines which AI service will handle requests

  • workspace (string, required): Workspace directory path

    • Absolute path to your workspace folder
    • Example: /home/user/ti-workspace or C:\Users\user\ti-workspace

Ollama Settings

  • ollama_url (string): Ollama server URL

    • Default: http://localhost:11434
    • Only used when agent is set to "ollama"
    • Change if running Ollama on a different host/port

  • model (string): Ollama model name

    • Examples: "qwen2.5-coder:3b", "deepseek-coder-v2:16b", "llama2"
    • Only used when agent is set to "ollama"

Gemini Settings

  • gemini_api (string): Google Gemini API key

    • Required when agent is set to "gemini"
    • Get your API key from Google AI Studio
    • Format: "AIza..."

  • gemini_model (string): Gemini model name

    • Examples: "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3-flash-preview"
    • Only used when agent is set to "gemini"

AWS Bedrock Settings

  • bedrock_api (string): AWS credentials

    • Required when agent is set to "bedrock"
    • Format: "ACCESS_KEY_ID:SECRET_ACCESS_KEY"
    • Example: "AKIAIOSFODNN7EXAMPLE:wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
    • Get credentials from AWS IAM Console

  • bedrock_model (string): Bedrock model identifier

    • Examples: "anthropic.claude-sonnet-4-6", "anthropic.claude-haiku-4-6", "anthropic.claude-opus-4-6"
    • Only used when agent is set to "bedrock"
    • Models automatically use inference profiles for optimal performance

  • bedrock_region (string): AWS region

    • Default: "us-east-1"
    • Examples: "us-east-1", "us-west-2", "eu-west-1"
    • Only used when agent is set to "bedrock"
    • Choose a region where Bedrock is available

Tested LLMs

Terminal Intelligence has been tested and verified to work with the following AI models:

Provider Model Notes
Ollama qwen2.5-coder:3b Recommended for coding tasks
qwen2.5-coder:1.5b Lightweight coding model
deepseek-coder-v2:16b Advanced coding capabilities
Gemini gemini-3-flash-preview Fast responses
gemini-3.1-pro-preview Advanced reasoning
gemini-3.1-flash-lite Lightweight and fast
AWS Bedrock anthropic.claude-sonnet-4-6 Best coding performance
anthropic.claude-haiku-4-6 Fast and cost-effective
anthropic.claude-opus-4-6 Highest intelligence

Note: AWS Bedrock models automatically use inference profiles for optimal performance and cross-region failover.

Command-Line Overrides

Configuration file values can be overridden using command-line flags:

# Override agent and model
ti --agent gemini --model gemini-3.1-flash-lite

# Override workspace
ti --workspace /path/to/workspace

# Override Ollama URL
ti --ollama-url http://remote-host:11434

# Override Gemini API key
ti --gemini-api YOUR_API_KEY

# Override Bedrock settings
ti --bedrock-api ACCESS_KEY:SECRET_KEY --bedrock-region us-west-2

Note: Command-line flags take precedence over config file values.

Configuration Best Practices

Security

  1. Protect API Keys: Never commit config.json with real API keys to version control
  2. Use Environment Variables: Consider using environment variables for sensitive credentials
  3. File Permissions: Ensure ~/.ti/config.json has restricted permissions (chmod 600 on Unix)

Performance

  1. Choose Appropriate Models:
    • Use smaller models (1.5b-3b) for faster responses
    • Use larger models (16b+) for complex coding tasks
  2. Local vs Cloud:
    • Ollama: No latency, runs locally, requires local resources
    • Gemini/Bedrock: Cloud-based, requires internet, no local resources needed

Workspace Organization

  1. Dedicated Workspace: Use a dedicated directory for TI projects
  2. Absolute Paths: Always use absolute paths for the workspace setting
  3. Backup: Regularly backup your workspace directory

Troubleshooting

Ollama Connection Issues

Error: Failed to connect to Ollama

Solutions:

  • Verify Ollama is running: ollama list
  • Check the ollama_url in config matches your Ollama server
  • Ensure firewall allows connections to Ollama port

Gemini API Issues

Error: Invalid Gemini API key

Solutions:

  • Verify your API key is correct in gemini_api field
  • Check API key has not expired
  • Ensure you have API quota remaining

Bedrock Authentication Issues

Error: Bedrock Generate (status 403): Invalid credentials

Solutions:

  • Verify AWS credentials format: ACCESS_KEY_ID:SECRET_ACCESS_KEY
  • Check IAM permissions include bedrock:InvokeModel
  • Ensure the model is available in your selected region
  • Verify your AWS account has Bedrock access enabled

Bedrock Model Not Found

Error: Bedrock Generate (status 404): Model may be invalid

Solutions:

  • Check model ID is correct (e.g., anthropic.claude-sonnet-4-6)
  • Verify model is available in your region
  • Request model access in AWS Bedrock console if needed

Interactive Configuration Editor

You can also edit configuration interactively from within TI:

  1. Press Ctrl+C to open the chat panel
  2. Type /config and press Enter
  3. Use the visual editor to modify settings
  4. Changes are saved automatically to ~/.ti/config.json

This provides a user-friendly way to update configuration without manually editing JSON files.