Register Your Model¶
Configure LLM models to use throughout the platform - from chat conversations to agent workflows.
Quick Start¶
Navigate to Model Manager from the workspace sidebar to register a new model.
Required Configuration¶
| Field | Description | Example |
|---|---|---|
| Model Name | Display name for this model | claude-sonnet-4-5 |
| Provider | Model provider (currently OpenAI-compatible) | openai |
| Model Type | Type of model | llm (language model) |
| Base URL | API endpoint | https://api.anthropic.com/v1/ |
| API Key | Your authentication key | sk-ant-api... |
| Model Identifier | Model ID from provider | claude-sonnet-4-5-20250929 |
Advanced Configuration¶
Optional parameters for fine-tuning model behavior:
Generation Control¶
| Parameter | Range | Description |
|---|---|---|
| Temperature | 0.0 - 2.0 | Randomness of outputs (higher = more creative) |
| Max Tokens | ≥ 1 | Maximum response length |
| Max Completion Tokens | ≥ 1 | Maximum tokens in completion |
| Top P | 0.0 - 1.0 | Nucleus sampling threshold |
Response Tuning¶
| Parameter | Range | Description |
|---|---|---|
| Frequency Penalty | -2.0 - 2.0 | Reduce repetition (positive = less repetitive) |
| Presence Penalty | -2.0 - 2.0 | Encourage topic diversity (positive = more diverse) |
| Stop Sequence | String or list | Tokens where model stops generating |
| Response Count (n) | 1 - 10 | Number of alternative responses to generate |
Debugging¶
| Parameter | Range | Description |
|---|---|---|
| Seed | Integer | Deterministic generation with same seed |
| Logprobs | Boolean | Return log probabilities of tokens |
| Top Logprobs | 0 - 20 | Number of most likely token alternatives |
Connection¶
| Parameter | Description |
|---|---|
| Timeout | Request timeout in seconds |
| Extra Body | Custom JSON parameters for specific providers |
Common Setups¶
Claude Sonnet 4.5¶
GPT-5¶
Gemini 2.5 Flash¶
Provider: openai
Base URL: https://generativelanguage.googleapis.com/v1beta/
Model: gemini-2.5-flash-latest
Managing Models¶
Edit Model¶
Click Edit on any model card to update configuration. Leave API Key blank to keep the existing key.
Delete Model¶
Click Delete to remove a model. You cannot delete models currently used by agents or workflows.
Title Generation Model¶
Set a default model for automatic conversation title generation:
- Click Title Model Settings in the header
- Select your preferred model
- Save
This model will be used when generating descriptive titles for new conversations.
Best Practices¶
API Keys: Store keys securely. The platform never displays full API keys after creation.
Naming: Use clear, descriptive names like gpt-4-turbo-preview instead of model-1.
Testing: Create test conversations to verify model configuration before using in workflows.
Default Parameters: Start without advanced parameters, then tune based on your use case.