AI Profile Management

AI profiles allow you to configure and switch between different AI providers, models, and settings. Create profiles for different use cases, models, or accounts.

What is an AI Profile?

An AI profile stores:

• Provider - Which AI service (OpenAI, Anthropic, Gemini, etc.)
• Model - Which model to use (gpt-4o, claude-3.5-sonnet, etc.)
• Base URL - API endpoint (optional, for custom endpoints)
• Headers - Custom HTTP headers (optional)
• Settings - Provider-specific options (reasoning effort, etc.)
• API Key - Stored separately per space in .glyph/app/ai_secrets.json

Default Profiles

On first launch, Glyph creates 7 default profiles:

1. OpenAI - Empty model, requires API key
2. OpenAI-compatible - Points to http://localhost:11434/v1
3. OpenRouter - Empty model, requires API key
4. Anthropic - Empty model, requires API key
5. Gemini - Empty model, requires API key
6. Ollama - Empty model, allows private hosts
7. Codex (ChatGPT OAuth) - Model codex, OAuth authentication

Creating Profiles

Profile Storage Location

~/.config/glyph/ai.json  (Linux)
~/Library/Application Support/glyph/ai.json  (macOS)
%APPDATA%/glyph/ai.json  (Windows)

Manual Profile Creation

{
  "profiles": [
    {
      "id": "unique-uuid-here",
      "name": "My Custom Profile",
      "provider": "openai",
      "model": "gpt-4o",
      "base_url": null,
      "headers": [],
      "allow_private_hosts": false,
      "reasoning_effort": null
    }
  ],
  "active_profile_id": "unique-uuid-here"
}

Profile Structure

Required Fields

interface AiProfile {
  id: string;              // Unique UUID
  name: string;            // Display name
  provider: AiProviderKind; // See providers below
  model: string;           // Model ID (e.g., "gpt-4o")
}

Optional Fields

interface AiProfile {
  base_url?: string | null;        // Custom endpoint
  headers?: AiHeader[];            // Custom HTTP headers
  allow_private_hosts?: boolean;   // Allow http:// and private IPs
  reasoning_effort?: string | null; // For Codex reasoning models
}

Provider Values

type AiProviderKind = 
  | "openai"
  | "openai_compat"
  | "openrouter"
  | "anthropic"
  | "gemini"
  | "ollama"
  | "codex_chatgpt";

Custom Headers

interface AiHeader {
  key: string;    // Header name
  value: string;  // Header value
}

Example:

{
  "headers": [
    { "key": "X-Custom-Header", "value": "my-value" },
    { "key": "Authorization", "value": "Bearer token" }
  ]
}

Switching Profiles

Via Settings UI

Active Profile Storage

The active profile ID is stored in ai.json:

{
  "profiles": [...],
  "active_profile_id": "uuid-of-active-profile"
}

Changing profiles in the UI updates this field.

Use Cases

Multiple Accounts

Scenario: You have personal and work OpenAI accounts.

Solution: Create two OpenAI profiles with different names and API keys.

[
  {
    "id": "uuid-1",
    "name": "OpenAI (Personal)",
    "provider": "openai",
    "model": "gpt-4o"
  },
  {
    "id": "uuid-2",
    "name": "OpenAI (Work)",
    "provider": "openai",
    "model": "gpt-4o-mini"
  }
]

Set different API keys for each in .glyph/app/ai_secrets.json:

{
  "uuid-1": "sk-personal-key",
  "uuid-2": "sk-work-key"
}

Model Variants

Scenario: You want quick access to different models.

Solution: Create profiles for each model.

[
  {
    "id": "uuid-1",
    "name": "GPT-4o (Fast)",
    "provider": "openai",
    "model": "gpt-4o-mini"
  },
  {
    "id": "uuid-2",
    "name": "GPT-4 (Quality)",
    "provider": "openai",
    "model": "gpt-4"
  }
]

Switch profiles to use different models without reconfiguring.

Local and Cloud

Scenario: Use Ollama for private notes, OpenAI for general tasks.

Solution: Keep both profiles configured and switch as needed.

Provider Comparison

Scenario: Compare responses from different providers.

Solution: Create profiles for OpenAI, Anthropic, and Gemini with similar models. Switch profiles and ask the same question to compare.

API Key Management

Per-Profile API Keys

API keys are stored per profile ID in .glyph/app/ai_secrets.json:

{
  "profile-uuid-1": "sk-openai-key",
  "profile-uuid-2": "sk-ant-anthropic-key",
  "profile-uuid-3": "google-api-key"
}

Setting API Keys

In Settings → AI:

1. Select a profile
2. Click Set API Key
3. Paste your API key
4. Click Save

The key is stored under the profile’s UUID.

Clearing API Keys

Click Clear API Key to remove the key for the active profile.

Security

• Keys stored per space, not globally
• Each space has independent ai_secrets.json
• Add .glyph/ to .gitignore to avoid committing keys
• Keys never logged or sent to Glyph servers

Profile Examples

Azure OpenAI Profile

{
  "id": "azure-uuid",
  "name": "Azure OpenAI",
  "provider": "openai_compat",
  "model": "gpt-4",
  "base_url": "https://your-resource.openai.azure.com/openai/deployments/gpt-4",
  "headers": [
    { "key": "api-key", "value": "your-azure-api-key" },
    { "key": "api-version", "value": "2024-02-15-preview" }
  ],
  "allow_private_hosts": false
}

Local LLaMA Profile

{
  "id": "llama-uuid",
  "name": "LLaMA 3.1 (Local)",
  "provider": "ollama",
  "model": "llama3.1:70b",
  "base_url": "http://localhost:11434/v1",
  "headers": [],
  "allow_private_hosts": true
}

OpenRouter with Custom Headers

{
  "id": "or-uuid",
  "name": "OpenRouter",
  "provider": "openrouter",
  "model": "anthropic/claude-3.5-sonnet",
  "base_url": null,
  "headers": [
    { "key": "HTTP-Referer", "value": "https://glyph.app" },
    { "key": "X-Title", "value": "Glyph" }
  ],
  "allow_private_hosts": false
}

Troubleshooting

Profile doesn’t appear in UI

Cause: Invalid JSON in ai.json.

Solution: Validate JSON syntax at jsonlint.com .

”Unknown profile” error

Cause: Profile ID mismatch or profile deleted.

Solution: Check that active_profile_id matches a profile ID in the profiles array.

API key not working after switching profiles

Cause: Each profile has its own API key.

Solution: Set the API key for each profile separately.

Profile settings not saving

Cause: File permissions or Glyph config directory not writable.

Solution: Check file permissions on ai.json.

Duplicate profile names

Cause: Multiple profiles with the same name.

Solution: Profile IDs must be unique, but names can be the same. Rename for clarity.

Best Practices

Naming Conventions

Use descriptive names:

• ✅ OpenAI (Personal)
• ✅ GPT-4o-mini (Fast)
• ✅ Claude 3.5 Sonnet (Research)
• ❌ Profile 1
• ❌ Test

Profile Organization

Group profiles by:

• Provider: OpenAI, Anthropic, Gemini
• Use case: Research, coding, writing
• Speed: Fast (mini models), Quality (full models)
• Cost: Free (Ollama), Paid (OpenAI)

Backup Profiles

Backup ai.json before making changes:

cp ~/.config/glyph/ai.json ~/.config/glyph/ai.json.bak

Restore if needed:

cp ~/.config/glyph/ai.json.bak ~/.config/glyph/ai.json

Future Enhancements

Potential profile features (not yet implemented):

• Profile creation via UI
• Profile import/export
• Profile templates
• Per-space profile overrides
• Profile-specific system prompts

Next Steps