Merge pull request #2226 from oracle-devrel/lsa_crewai01

added crewai integration, rel 1

Author: anshoracle

ai/gen-ai-agents/crewai-oci-integration/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Luigi Saetta
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

ai/gen-ai-agents/crewai-oci-integration/README.md

@@ -0,0 +1,107 @@
+# CrewAI ↔ OCI Generative AI Integration
+
+This repository provides examples and configuration guidelines for integrating **[CrewAI](https://github.com/joaomdmoura/crewAI)** with **Oracle Cloud Infrastructure (OCI) Generative AI** services.  
+The goal is to demonstrate how CrewAI agents can seamlessly leverage OCI-hosted models through the **LiteLLM gateway**.
+
+Reviewed: 31.10.2025
+
+---
+
+## 🔐 Security Configuration
+
+Before running the demos, you must configure access credentials for OCI.
+
+In these examples, we use a **locally stored key pair** for authentication.  
+Ensure your local OCI configuration (`~/.oci/config` and private key) is correctly set up and accessible to the Python SDK.
+
+To correctly start the **LiteLLM gateway** you need to create and configure correctly a **config.yml** file. To create this file use the [template](./config_template.yml).
+
+In addition, you should be **enabled** to use OCI Generative AI Service in your tenant. If you haven't yet used OCI GenAI ask to your tenant's admin to setup the **needed policies**.
+
+---
+
+## 🧩 Demos Included
+
+- [Simple CrewAI Agent](./simple_test_crewai_agent.py) — basic CrewAI agent interacting with an LLM through OCI
+- [OCi Consumption Report](./crew_agent_mcp02.py)
+- *(More demos to be added soon)*
+
+---
+
+## 📦 Dependencies
+
+The project relies on the following main packages:
+
+| Dependency | Purpose |
+|-------------|----------|
+| **CrewAI** | Framework for creating multi-agent workflows |
+| **OCI Python SDK** | Access OCI services programmatically |
+| **LiteLLM (Gateway)** | OpenAI-compatible proxy for accessing OCI Generative AI models |
+
+To connect CrewAI to OCI models, we use a **LiteLLM gateway**, which exposes OCI GenAI via an **OpenAI-compatible** REST API.
+
+---
+
+## ⚙️ Environment Setup
+
+1. **Create a Conda environment**
+```bash
+conda create -n crewai python=3.11
+```
+
+2. **Activate** the environment
+```
+conda activate crewai
+```
+
+3. **Install** the required **packages**
+```
+pip install -U oci litellm "litellm[proxy]" crewai
+```
+
+4. Run the LiteLLM Gateway
+
+Start the LiteLLM gateway using your configuration file (config.yml):
+```
+./start_gateway.sh
+```
+
+Make sure the gateway starts successfully and is listening on the configured port (e.g., http://localhost:4000/v1).
+
+🧠 Test the Integration
+
+Run the sample CrewAI agent to verify that CrewAI can connect to OCI through LiteLLM:
+
+```
+python simple_test_crewai_agent.py
+```
+
+If the setup is correct, you should see the agent’s output using an OCI model.
+
+## Integrate Agents with **MCP** servers.
+Install this additional package:
+
+```
+pip install 'crewai-tools[mcp]'
+```
+
+You can test the integration with **MCP** using [OCI Consumption report](./crew_agent_mcp02.py) that generates a report
+of the consumption in your tenant (top 5 compartments, for 4 weeks).
+
+To have this demo up&running:
+* download the code for the MCP server from [here](https://github.com/oracle-devrel/technology-engineering/blob/main/ai/gen-ai-agents/mcp-oci-integration/mcp_consumption.py)
+* start the MCP server, on a free port (for example 9500)
+* register the URL, in [source](./crew_agent_mcp02.py), in the section:
+```
+server_params = {
+    "url": "http://localhost:9500/mcp",
+    "transport": "streamable-http"
+}
+```
+
+If you don't want to secure (with JWT) the communication with the MCP server, put 
+```
+ENABLE_JWT_TOKEN = False
+```
+in the config.py file.
+

ai/gen-ai-agents/crewai-oci-integration/config_template.yml

@@ -0,0 +1,36 @@
+# config.yaml for litellm with OCI Grok models
+litellm_settings:
+  drop_params: true            
+  # drop unsupported params instead of 500 errors
+  additional_drop_params: ["max_retries"]
+
+# Common OCI connection parameters
+common_oci: &common_oci
+  provider: oci
+  oci_region: us-chicago-1
+  oci_serving_mode: ON_DEMAND
+  supports_tool_calls: true
+  oci_user: your-oci-user-ocid
+  oci_fingerprint: your-oci-api-key-fingerprint
+  oci_tenancy: your-oci-tenancy-ocid
+  oci_compartment_id:   your-oci-compartment-ocid
+  oci_key_file:   /path/to/your/oci_api_key.pem
+  api_key: key4321
+  
+
+# List of models
+model_list:
+  - model_name: grok4-oci
+    litellm_params:
+      <<: *common_oci           # merge common OCI params
+      model: oci/xai.grok-4
+
+  - model_name: grok4-fast-oci
+    litellm_params:
+      <<: *common_oci
+      model: oci/xai.grok-4-fast-reasoning
+
+general_settings:
+  telemetry: false
+  proxy_logging: false
+  allow_model_alias: true

ai/gen-ai-agents/crewai-oci-integration/crew_agent_mcp01.py

@@ -0,0 +1,58 @@
+"""
+CrewAI agent with MCP
+
+This one is doing Deep research using internet search tools via MCP server.
+
+see: 
+    https://docs.crewai.com/en/mcp/overview
+    https://docs.crewai.com/en/mcp/multiple-servers
+"""
+import os
+from crewai import Agent, Task, Crew, LLM
+from crewai_tools import MCPServerAdapter
+
+# Disable telemetry, tracing, and logging
+os.environ["CREWAI_LOGGING_ENABLED"] = "false"
+os.environ["CREWAI_TELEMETRY_ENABLED"] = "false"
+os.environ["CREWAI_TRACING_ENABLED"] = "false"
+
+llm = LLM(
+    model="grok4-fast-oci",
+    # LiteLLM proxy endpoint
+    base_url="http://localhost:4000/v1",
+    api_key="sk-local-any",
+    temperature=0.2,
+    max_tokens=4000,
+)
+
+server_params = {
+    "url": "http://localhost:8500/mcp",
+    "transport": "streamable-http"
+}
+
+# Create agent with MCP tools
+with MCPServerAdapter(server_params, connect_timeout=60) as mcp_tools:
+    print(f"Available tools: {[tool.name for tool in mcp_tools]}")
+
+    research_agent = Agent(
+        role="Research Analyst",
+        goal="Find and analyze information using advanced search tools",
+        backstory="Expert researcher with access to multiple data sources",
+        llm=llm,
+        tools=mcp_tools,
+        verbose=True
+    )
+
+    # Create task
+    research_task = Task(
+        description="Research the latest developments in AI agent frameworks",
+        expected_output="Comprehensive research report with citations",
+        agent=research_agent
+    )
+
+    # Create and run crew
+    crew = Crew(agents=[research_agent], tasks=[research_task])
+    
+    result = crew.kickoff()
+
+    print(result)

ai/gen-ai-agents/crewai-oci-integration/crew_agent_mcp02.py

@@ -0,0 +1,79 @@
+"""
+CrewAI agent with MCP
+
+This one is analyzing tenant consumption via MCP server.
+
+see: 
+    https://docs.crewai.com/en/mcp/overview
+    https://docs.crewai.com/en/mcp/multiple-servers
+"""
+import os
+from datetime import datetime
+from crewai import Agent, Task, Crew, LLM
+from crewai_tools import MCPServerAdapter
+
+# Disable telemetry, tracing, and logging
+os.environ["CREWAI_LOGGING_ENABLED"] = "false"
+os.environ["CREWAI_TELEMETRY_ENABLED"] = "false"
+os.environ["CREWAI_TRACING_ENABLED"] = "false"
+
+llm = LLM(
+    model="grok4-oci",
+    # LiteLLM proxy endpoint
+    base_url="http://localhost:4000/v1",
+    api_key="sk-local-any",
+    temperature=0.,
+    max_tokens=4000,
+)
+
+# OCI consumption
+server_params = {
+    "url": "http://localhost:9500/mcp",
+    "transport": "streamable-http"
+}
+
+# Create agent with MCP tools
+with MCPServerAdapter(server_params, connect_timeout=60) as mcp_tools:
+    print(f"Available tools: {[tool.name for tool in mcp_tools]}")
+
+    research_agent = Agent(
+        role="OCI Consumption Analyst",
+        goal="Find and analyze information about OCI tenant consumption.",
+        backstory="Expert analyst with access to multiple data sources",
+        llm=llm,
+        tools=mcp_tools,
+        max_iter=30,
+        max_retry_limit=5,
+        verbose=True
+    )
+
+    # Create task
+    research_task = Task(
+        description="Identify the top 5 compartments by consumption (amount) for the OCI tenant "
+        "in the  weeks of the month of september 2025, analyze the trends and provide insights on usage patterns."
+        "Analyze fully the top 5 compartments. Use only the amount, not the quantity.",
+        expected_output="Comprehensive report with data-backed insights.",
+        agent=research_agent
+    )
+
+    # Create and run crew
+    crew = Crew(agents=[research_agent], tasks=[research_task])
+    
+    result = crew.kickoff()
+
+    print(result)
+
+    # --- Save the result to a Markdown file ---
+    # Create an output directory if it doesn’t exist
+    output_dir = "reports"
+    os.makedirs(output_dir, exist_ok=True)
+
+    # Use timestamped filename for clarity
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    output_path = os.path.join(output_dir, f"oci_consumption_report_{timestamp}.md")
+
+    # Write the result
+    with open(output_path, "w", encoding="utf-8") as f:
+        f.write(str(result))
+
+    print(f"\n✅ Report saved successfully to: {output_path}")