Create an Azure AI Service agent which calls an external http endpoint without a client intervention

---

---

# Request for Professional Developer Assistance ## Project: Azure AI Agent with Direct HTTP Call Capability ### Executive Summary I need a professional developer to create a Weather Agent that can make direct HTTP calls to external weather APIs from within the Azure AI Agent Service itself, without requiring client-side execution. Despite multiple attempts using different approaches, the agent consistently fails to make external HTTP calls due to apparent platform limitations. ### Current Situation We have attempted to create a Weather Agent using Azure AI Agent Service (formerly Azure OpenAI Assistants API) that can autonomously fetch weather data from external APIs. However, all attempts have failed with the same issue: the agent cannot make outbound HTTP requests. ### Technical Context **Environment:** - Platform: Azure AI Agent Service (Azure AI Foundry) - Model: GPT-3.5-turbo - SDK: azure-ai-projects (Python) - External API: Weather service endpoint or OpenWeatherMap API **What We've Tried:** 1. **OpenAPI Tool Specification (Server-side execution)** - Created OpenAPI 3.0 specification - Configured with anonymous authentication - Agent attempts execution but no HTTP calls are made - No logs appear in target service 2. **Function Tools (Client-side execution)** - Works, but requires client to execute HTTP calls - Not the desired solution 3. **Direct HTTP Instructions** - Explicitly instructed agent to make HTTP calls autonomously - Agent still returns `requires_action` status - Confirms it cannot make direct calls ### Specific Requirements 1. **Agent Must:** - Make HTTP GET/POST requests directly to external weather APIs - Process JSON responses autonomously - Return formatted weather information to users - NOT require any client-side code execution 2. **Example Flow:** ``` User: "What's the weather in Seattle?" Agent: [Makes HTTP call to weather API directly] Agent: [Processes response] Agent: "The weather in Seattle is 15°C with cloudy skies." ``` 3. **No Client Involvement:** - The agent should handle the entire flow internally - No `requires_action` status should be returned - No client-side function execution needed ### Current Code Structure ```python # Agent Configuration agent_instructions = """You are a weather agent that makes direct HTTP calls...""" tools = [ dict( type="openapi", openapi=dict( name="weather_api", spec=dict( openapi="3.0.0", servers=[dict(url="https://api.weather.com")], paths=dict( weather=dict( get=dict( operationId="getWeather", parameters=[...], responses=dict(...) ) ) ) ), auth=dict(type="anonymous") ) ) ] # Agent Creation agent = client.agents.create_agent( model="gpt-35-turbo", name="WeatherAgent", instructions=agent_instructions, tools=tools ) # Execution run = client.agents.runs.create_and_process( thread_id=thread.id, agent_id=agent.id ) # Result: Agent cannot make HTTP calls, returns errors ``` ### Deliverables Needed 1. **Working Weather Agent** that can: - Make direct HTTP calls from within Azure AI Agent Service - Handle authentication if needed (API keys, etc.) - Process responses without client intervention 2. **Deployment Script** that: - Deploys the agent to Azure AI Agent Service - Configures all necessary permissions/settings - Handles any special Azure configurations needed 3. **Test Script** that: - Demonstrates the agent making direct HTTP calls - Shows no client-side execution is required - Validates the end-to-end flow 4. **Documentation** explaining: - How the solution works - Any Azure configurations required - Any limitations or special considerations ### Constraints & Considerations 1. Must work within Azure AI Agent Service (not custom hosted solution) 2. Must use official Azure SDKs where possible 3. Should handle standard weather API endpoints (OpenWeatherMap, Weather.gov, etc.) 4. Must not require proxy servers or intermediate services ### Questions for the Developer 1. Is it technically possible for Azure AI Agents to make direct external HTTP calls? 2. If not possible with current platform, what are the recommended alternatives? 3. Are there specific Azure configurations or permissions that enable external HTTP calls? 4. Is there a different Azure service that would better suit this requirement? ### Access & Resources - Azure subscription with AI Agent Service enabled - Access to Azure AI Foundry workspace - Test weather API endpoints available - Existing codebase and failed attempts for reference ### Success Criteria The solution is successful when: 1. User asks agent about weather 2. Agent makes HTTP call directly (no client code runs) 3. Agent returns weather information 4. No `requires_action` status or client-side execution occurs ### Budget & Timeline Please provide: - Feasibility assessment - Estimated hours/cost - Proposed approach - Timeline for delivery ### Contact Information [Your contact details here] --- **Note:** If this is not technically possible due to Azure AI Agent Service limitations, please provide alternative architectures that could achieve similar functionality while minimizing client-side code requirements.

Keyword: Software Development

Price: $200.0

AI Agent Development Python