added
Agents - Now in BETA
2 days ago
We've just delivered our new Agents Beta API to build single and multi-agent systems, making it easier than ever to build and run powerful agents.
Key Features
Built on Modern Standards
Using A2A Protocol for reliable multi-agent coordination
Intelligent Memory & Context
- Memory Stores: Persistent context across conversations with entity-based isolation.
- Knowledge Base: Ground responses in your data with dynamic discovery and querying
Multi-Modal Processing
- File Processing - Handle images and PDFs with support varying per given model
- Search & Web Scraping - Bring in real-time insights with built-in search and scraping tools
Flexible Tool Ecosystem
- HTTP Tools - Seamless external API integrations with parameter templating and authentication
- Function Tools - Custom business logic and access to internal systems with OpenAI function calling schema validation
- Code Tools - Custom Python logic and transformations in secure sandbox environments
Architecture
- Multi-Agent Systems - Build sophisticated hierarchies with specialized sub-agents and intelligent delegation.
- Synchronous and streaming endpoints for different use cases.
Orq handles memory, orchestration, and execution so you can focus on your agent’s role and behavior.
Example
Initiate an Agent call with a single call to the /v2/agents/run
API as follow:
curl --location 'https://my.orq.ai/v2/agents/stream-run' \
--header 'Content-Type: application/json' \
--header 'Authorization: $ORQ_API_KEY' \
--data '{
"key": "simple-agent-1",
"role": "Assistant",
"description": "A helpful assistant for general tasks",
"instructions": "Be helpful and concise",
"settings": {
"max_iterations": 3,
"max_execution_time": 300,
"tools": [
{
"type": "current_date"
}
]
},
"message": {
"role": "user",
"parts": [
{
"kind": "text",
"text": "What'\''s today'\''s date and day of the week?"
}
]
},
"model": "openai/gpt-4o",
"path": "Default/agents",
"memory_stores": [],
"team_of_agents": []
}'
To learn more about agents, see Agents and browse through the advanced usages of our new API.