Customization
Audience: Developers
Customization
There are two ways to customize GL Open DeepResearch capabilities:
1. Create New Research Profiles
What it is: Create custom profiles with different configurations (models, timeouts, depth, tools) using existing research engines.
When to use: When you want to customize behavior without changing the underlying research engine. For example:
Use a different LLM model (e.g., Claude instead of GPT-4)
Adjust research depth or timeout settings
Configure different tool sets
Create profiles optimized for specific use cases
How: Use the Profiles API to create, update, and manage profiles. Profiles are stored in the database and can be referenced by name in research requests.
See: Research profiles for detailed guide and examples.
2. Integrate New Open Source Deep Research Providers
What it is: Add support for new open-source deep research engines by implementing the adapter protocol.
When to use: When you want to use a different research engine that isn't currently supported (e.g., a new open-source deep research library).
Current providers: GL Open DeepResearch currently supports two open-source deep research providers:
Tongyi Deep Research — Multi-turn ReAct agent
GPT-Researcher — Automated research workflow
How: Implement the OrchestratorAdapter protocol and register the adapter in the Orchestrator Registry. See Core components and Core design principles for details.
See: Available Open Source Deep Research for provider details and Tongyi / GPT-Researcher for implementation examples.
Quick Comparison
Complexity
Low — API-based configuration
High — Requires code changes
Use case
Customize existing engines
Add new research engines
Storage
Database (via API)
Code (adapter implementation)
Examples
Different models, timeouts, tools
Tongyi, GPT-Researcher
For most use cases, creating new profiles is the recommended approach. Only integrate new providers when you need a research engine that isn't currently supported.
Last updated