AI Config
Configure the AI service (API key, model, etc.) used by Pluto AI Agent. Once configured, you can drive strategy, backtest, bot, and market workflows by chat—every action runs within system risk and permission boundaries.
AI Config configures the AI services (API Key, model, etc.) that power AI features across the system. The same configuration backs:
- AI Agent: conversational entry—use natural language for strategies, backtests, quotes, bot management, and more
- AI-driven trading: signals from AI analysis are executed under platform risk and compliance controls
- News & event-driven trading: parse news, sentiment, and macro events; generate and execute signals by event type
After configuration, the features above can be used. For what AI Agent does and how to use it, see AI Agent; this page focuses on configuration steps and field reference.
Quick start
If you already have an API Key from OpenAI, Claude, DeepSeek, or another supported provider, open the AI config page, pick the provider, paste the key, choose a model, save, then run Test connection to confirm—then you can start using AI Agent, AI-driven trading, and related features.
Configuration steps
Open the config page
In the system, open “AI config”, “Basic settings”, or “System settings” (wording varies); some products use a modal.
Choose provider and fill credentials
Pick a supported provider (e.g. OpenAI, Claude, DeepSeek), enter API Key and model name; for self-hosted or proxy endpoints, set Base URL.
Test connection
After saving, use Test connection to verify the API and model are reachable; on failure, check Key, Base URL, and network.
Enable and verify
Turn the AI config on, send a message from AI Agent, and confirm you get replies.
Multiple models supported
The system can connect to multiple providers and models. In AI config, select a provider and enter its API Key and model name; primary/backup is supported and the system can fail over when the primary is unavailable.
| Provider | Highlights | Typical use |
|---|---|---|
| OpenAI | GPT family, strong reasoning | General workloads, best quality |
| Claude | Anthropic, long text and logic | Complex analysis, long threads |
| DeepSeek | Cost-friendly, strong Chinese | Cost-sensitive, high call volume |
| Groq | Very fast inference, low latency | Latency-sensitive workloads |
| Gemini | Google family | Multimodal, stable availability |
| Grok | xAI family | Real-time and all-round |
After you pick a provider on the page, fill in that provider’s model name; choose based on cost, speed, quality, and network.
Choosing a model
Beginners can start with DeepSeek or Claude for value; for low latency (e.g. fast group replies) consider Groq; for top quality use GPT-4, Claude, or similar.
Configuration overview
| Item | Description |
|---|---|
| Where to configure | Usually under system settings, basic settings, or a dedicated “AI config” page; some UIs open a modal from another entry. |
| Required / common | Provider type, API Key, model name; Base URL when using self-hosted or proxy endpoints. |
| Primary / HA | When supported, configure a backup provider/model to auto-switch if the primary is down or times out. |
Configuration fields
| Field | Description |
|---|---|
| Enable / toggle | When off, AI Agent chat is disabled; when on and credentials are valid, AI features work. |
| Provider & model | Select provider type and enter that provider’s model name; trade off cost, speed, and quality. |
| API Key / Base URL | Paste the API Key from the provider console; for custom endpoints set Base URL, otherwise leave default. |
| Primary / backup | With both set, the system switches to backup on primary failure or timeout. |
After configuration
- Test connection: Always run Test connection after saving to verify the API and model; if it fails, check Key, Base URL, network, and firewall rules.
- Use with notifications: To use AI Agent in DingTalk, Telegram, etc., finish channel setup in Notification Config and enable smart reply; AI config and notification config are independent—both must be correct for group chat.
- Permissions & security: AI Agent actions still follow role permissions; if something is blocked, check the account. Do not leak API Keys—store them only inside the system for calling AI services.
Common issues
Troubleshooting order
- Confirm AI config is enabled; 2) verify API Key is valid and not expired; 3) run Test connection and read the error; 4) for proxy/self-host, verify Base URL and network reachability.
| Issue | What to try |
|---|---|
| Cannot chat after saving | Check enabled state, Key, and model; use Test connection; verify Base URL/network for proxy/self-host. |
| Very slow or timeout | Try a faster provider/model (e.g. Groq); configure backup if supported. |
| AI says no permission | The action is role-gated—grant strategy, backtest, trading-bot, or related permissions in user/role admin. |
| Which models are supported | Use the provider/model list on the AI config page and paste the matching Key. |
| How to set primary/backup | Primary = default model; backup = cheaper or faster failover when primary errors. |
Next steps
- AI Agent — Feature overview and usage
- Exchange Config, Notification Config — Other basics
- 5-minute quick start — Onboarding and navigation