LobeHub

Open-source AI chat client integration guide with custom providers and local knowledge base

​

1. Overview

LobeHub is an open-source, cross-platform AI chat client (formerly Lobe Chat) supporting multiple models, multimodal interactions, a plugin system, and local knowledge bases. With GravitexAI, you can access 100+ mainstream AI models using a single API key in LobeHub, without needing to register and pay for each provider separately.

​

2. Quick start

​

Step 1: Get a GravitexAI API Key

1. Visit the GravitexAI Console
2. Click Create API Key, name it, and save
3. Copy the key starting with sk- for later

​

Step 2: Install LobeHub

LobeHub offers three ways to use it:

​

Step 3: Open LobeHub settings

1. Click Personal Avatar → App Settings in the top-left to enter the settings page
2. Select Agent → AI Provider from the left navigation
3. Click the + button in the top-right to add a custom provider

​

Step 4: Create a custom AI provider

Click the Create button in the bottom-right corner to save.

​

Step 5. Verify connectivity

After successful creation, you’ll be redirected to the provider details page. You can:

1. Fill in API Key and Proxy URL: Confirm or modify the settings again
2. Enable advanced options (optional):
   • Use Responses API spec: Enable to use the new OpenAI request format (only supported by OpenAI models)
   • Use client request mode: Browser sends requests directly, which can improve response speed
3. Connectivity check: Select a model from the dropdown and click the Check button to test connectivity
4. Get model list: Click the Get Model List button to automatically fetch all models provided by GravitexAI

Once retrieved successfully, the Model List will display all models currently available from GravitexAI, grouped by type: chat, image, video, embeddings, TTS.

​

Step 6. Select a model and start chatting

From the model dropdown in the chat interface, select the model you want to use (e.g., claude-opus-5, gpt-5.6-luna, deepseek-v4-pro) and start chatting.

​

3. One-stop access to 100+ models

LobeHub has built-in official channels for multiple providers, requiring individual API keys. By connecting to GravitexAI, you only need to create one custom provider to cover all models:

​

4. LobeHub features

​

Plugin system

LobeHub provides a rich plugin marketplace (Settings → Plugins). Recommended plugins to enable:

• Web Search: Let the model access the internet in real time
• Code Interpreter: Run Python code and create data visualizations in conversations
• Chart Generation: Automatically generate flowcharts and mind maps
• Image Generation: Generate images through models like gpt-image-2, gemini-3-pro-image

​

Local knowledge base (RAG)

LobeHub supports uploading files to build a local vector database:

1. Go to the Knowledge Base page and create a new knowledge base
2. Upload PDF, Markdown, Word, Excel, TXT, and other files
3. The system automatically slices and vectorizes the content
4. When chatting, enable the knowledge base and the model will answer based on your private materials

​

Multimodal conversation

Support uploading images for visual understanding, image recognition, and OCR:

• Recommended vision models: gemini-3.6-flash, claude-opus-5, gpt-5.6-terra
• Drag images directly into the chat box and the model will automatically recognize them

​

Agent Market

LobeHub includes rich preset agents (translation, writing, coding, interviews, etc.). You can also:

• Enable community-shared agents with one click in the Agent Market
• Create custom agents: define personas, prompts, opening messages, knowledge bases, and plugins

​

Chat management and export

• Multi-branch conversations, message editing, and regeneration
• Chat export: Markdown, PNG, JSON
• Global search of chat history

​

5. Advanced configuration

​

Network proxy and self-hosting

If you need to use a proxy or self-host LobeHub:

1. Desktop: Settings → Network → Configure HTTP/HTTPS proxy
2. Self-hosted (Docker): Inject GravitexAI configuration via environment variables:
   OPENAI_API_KEY=sk-your-gravitex-key
   OPENAI_PROXY_URL=https://api.gravitex.ai/v1
   CUSTOM_MODELS=gpt-5.6-terra,claude-opus-5,deepseek-v4-pro
   
3. Client request mode: When enabled, the browser connects directly to the API (ensure the browser can access GravitexAI)

​

Keyboard shortcuts

​

6. Mobile

LobeHub desktop supports Windows / macOS / Linux. For mobile, access the lobechat.com web version via browser (responsive design) — no separate app installation required.

​

7. Troubleshooting

​

8. Best practices

​

Model selection strategy

Use different models for different tasks, select by complexity tier:

• Daily chat / simple Q&A: gemini-3.1-flash-lite, claude-haiku-4-5, deepseek-v4-flash
• Complex reasoning / long document analysis: claude-opus-5, gpt-5.6-sol, deepseek-v4-pro
• Coding: claude-opus-5, gpt-5.6-sol, claude-sonnet-5
• Image understanding: gemini-3.6-flash, claude-opus-5, gpt-5.6-terra
• Image generation: gpt-image-2, gemini-3-pro-image

For the complete model list and performance comparison, please refer to the Model Recommendations page.

​

Context management

• Regularly clean up unused old conversations
• Break complex tasks into multiple short conversations
• Use the Fork Conversation feature to explore different answers

​

Security and privacy

• Do not share API Keys in public
• Rotate keys regularly
• For sensitive conversations, consider self-hosting LobeHub with fully local data storage

​

Self-hosting recommendations

• Use Docker + PostgreSQL in production — do not use the built-in database
• Consider Cloudflare as a reverse proxy
• Manage keys centrally through environment variables for multi-instance deployment

​

9. Comparison of three approaches