making the rocket text more like a button, making it more obvious for users to hover over agents for more info, adding to troubleshooting doc pointers to lambda logs, cloudformation output is a link to the secret, bug fix in external MCP agents with custom names.

kaleko · kaleko · commit 537e71322a4e · 2025-08-29T20:04:50.000Z
diff --git a/docs/custom-MCP-agent.md b/docs/custom-MCP-agent.md
@@ -147,9 +147,9 @@ In the AWS account where the IDP solution is deployed, create a secret with your
 
 1. **Navigate to AWS Secrets Manager** in the AWS Console
 2. **Find the External MCP Agents Secret**:
-   - Look for a secret named `{StackName}/external-mcp-agents/credentials` (where StackName is your IDP stack name).
+   - Use the `ExternalMCPAgentsSecretConsoleURL` link from your CloudFormation stack outputs to go directly to the secret
+   - Alternatively, look for a secret named `{StackName}/external-mcp-agents/credentials` (where StackName is your IDP stack name)
    - This secret is automatically created by the CloudFormation template with an empty array `[]`
-   - You can find the exact secret name in your CloudFormation stack outputs
 
 3. **Update Secret with JSON Array Structure**:
    ```json
diff --git a/docs/troubleshooting.md b/docs/troubleshooting.md
@@ -42,6 +42,14 @@ This guide provides solutions for common issues and optimization techniques for
 | **DynamoDB capacity exceeded** | Check CloudWatch metrics for throttling. Consider increasing provisioned capacity or switching to on-demand capacity. |
 | **S3 permission errors** | Verify bucket policies and IAM role permissions. Check for cross-account access issues. |
 
+### Agent Processing Issues
+
+| Issue | Resolution |
+|-------|------------|
+| **Agent query shows "processing failed"** | Check CloudWatch logs for the Agent Processing Lambda function (`{StackName}-AgentProcessorFunction-*`). Look for specific error messages, timeout issues, or permission errors. |
+| **External MCP agent not appearing** | Verify the External MCP Agents secret is properly configured with valid JSON array format. Check CloudWatch logs for agent registration errors. |
+| **Agent responses are incomplete** | Check CloudWatch logs for token limits, model throttling, or timeout issues in the Agent Processing function. |
+
 ## Performance Considerations
 
 ### Resource Sizing
diff --git a/lib/idp_common_pkg/idp_common/agents/factory/registry.py b/lib/idp_common_pkg/idp_common/agents/factory/registry.py
@@ -36,7 +36,7 @@
     agent_name="Analytics Agent",
     agent_description="""
     Converts natural language questions into SQL queries and generates visualizations from document data.
-    This agent has access to all databases within the IDP system, including metering databases which track
+    This agent has access to many databases within the IDP system, including metering databases which track
     the document processing volume and statistics, document-specific tables for different classes of documents,
     entity-specific information like names of people, numbers, and other entities extracted from documents,
     as well as evaluation tables which include confidence scores for extracted entities as well as
@@ -154,9 +154,8 @@ def wrapper(config=None, session=None, model_id=None, **kwargs):
         # Determine agent name and ID
         if "agent_name" in mcp_config and mcp_config["agent_name"]:
             agent_name = mcp_config["agent_name"]
-            # Create ID from name with UUID for uniqueness
-            name_no_spaces = agent_name.replace(" ", "_").lower()
-            agent_id = f"{name_no_spaces}_{str(uuid.uuid4())[:8]}"
+            # Create ID from name with no spaces - assume users provide unique names
+            agent_id = agent_name.replace(" ", "_").lower()
         else:
             agent_name = f"External MCP Agent {i}"
             agent_id = f"external-mcp-agent-{i}"
diff --git a/src/ui/src/components/document-agents-layout/AgentQueryInput.jsx b/src/ui/src/components/document-agents-layout/AgentQueryInput.jsx
@@ -414,21 +414,29 @@ const AgentQueryInput = ({ onSubmit, isSubmitting, selectedResult }) => {
       <form onSubmit={handleSubmit}>
         <SpaceBetween size="l">
           <Box>
-            <SpaceBetween direction="horizontal" size="s" alignItems="center">
-              <Box fontSize="body-m" fontWeight="bold">
+            <div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center' }}>
+              <Box fontSize="heading-xs" fontWeight="bold">
                 Select from available agents
               </Box>
-              <Button variant="link" onClick={() => setShowMcpInfoModal(true)} fontSize="body-s">
+              <Button
+                variant="normal"
+                onClick={() => setShowMcpInfoModal(true)}
+                fontSize="body-s"
+                style={{
+                  borderColor: '#e9ecef',
+                }}
+              >
                 🚀 NEW: Integrate your own systems with MCP!
               </Button>
-            </SpaceBetween>
+            </div>
             <div
               style={{
                 maxHeight: '200px',
                 overflowY: 'auto',
                 border: '1px solid #d5dbdb',
                 padding: '8px',
                 position: 'relative',
+                marginTop: '16px',
               }}
             >
               {isLoadingAgents && (
@@ -446,10 +454,21 @@ const AgentQueryInput = ({ onSubmit, isSubmitting, selectedResult }) => {
                   {availableAgents.map((agent) => (
                     <div
                       key={agent.agent_id}
-                      onMouseEnter={(e) => handleMouseEnter(agent.agent_id, e)}
+                      onMouseEnter={(e) => {
+                        handleMouseEnter(agent.agent_id, e);
+                        e.currentTarget.style.backgroundColor = '#f8f9fa';
+                      }}
                       onMouseMove={handleMouseMove}
-                      onMouseLeave={handleMouseLeave}
-                      style={{ width: '100%' }}
+                      onMouseLeave={(e) => {
+                        handleMouseLeave();
+                        e.currentTarget.style.backgroundColor = 'transparent';
+                      }}
+                      style={{
+                        width: '100%',
+                        padding: '4px',
+                        borderRadius: '4px',
+                        transition: 'background-color 0.2s ease',
+                      }}
                     >
                       <Checkbox
                         checked={selectedAgents.includes(agent.agent_id)}
diff --git a/src/ui/src/components/document-agents-layout/tools-panel.jsx b/src/ui/src/components/document-agents-layout/tools-panel.jsx
@@ -24,8 +24,8 @@ const ToolsPanel = () => {
     >
       <div>
         <p>
-          Use Agent Analysis to interact with AI agents using natural language and receive interactive visualizations and
-          insights.
+          Use Agent Analysis to interact with AI agents using natural language and receive interactive visualizations
+          and insights.
         </p>
         <h3>How to use</h3>
         <ol>
diff --git a/template.yaml b/template.yaml
@@ -6414,6 +6414,6 @@ Outputs:
       REACT_APP_APPSYNC_GRAPHQL_URL=${GraphQLApi.GraphQLUrl}
       REACT_APP_AWS_REGION=${AWS::Region}
       REACT_APP_SETTINGS_PARAMETER=${SettingsParameter}
-  ExternalMCPAgentsSecretName:
-    Description: Name of the Secrets Manager secret for External MCP Agent credentials (JSON array format)
-    Value: !Sub "${AWS::StackName}/external-mcp-agents/credentials"
+  ExternalMCPAgentsSecretConsoleURL:
+    Description: External MCP Agents secret console URL - configure MCP server credentials here (JSON array format)
+    Value: !Sub "https://${AWS::Region}.console.aws.amazon.com/secretsmanager/secret?name=${AWS::StackName}/external-mcp-agents/credentials&region=${AWS::Region}"
