You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
"instructions": "You are a privacy agent that REMOVES direct PII. Focus on clearly personal identifiers:\n\nEmail addresses\nPhone numbers\nSocial Security Numbers\nFull names (but not generic titles)\nStreet addresses\nCredit card numbers\nDriver's license numbers\n\nResponse Format:\n\ndetected: true if any PII was found, false otherwise\ntypes: array of PII types found (e.g., ['email', 'name', 'phone'])\nredacted: the input text with PII replaced by [REDACTED], keeping the text readable and natural\n\nExamples:\n\nInput: \"I work at Acme Corp in Berlin as a manager\"\n\nOutput: detected=false, types=[], redacted='I work at Acme Corp in Berlin as a manager'\n\n\nInput: \"Contact John Smith at john@email.com or 555-1234\"\n\nOutput: detected=true, types=['name', 'email', 'phone'], redacted='Contact [REDACTED] at [REDACTED] or [REDACTED]'\n\n\nInput: \"The CEO from Microsoft contacted me\"\n\nOutput: detected=false, types=[], redacted='The CEO from Microsoft contacted me'"
19
+
},
20
+
"supervisor-agent": {
21
+
"enabled": true,
22
+
"model": {
23
+
"name": "claude-3-7-sonnet-latest",
24
+
"parameters": {}
25
+
},
26
+
"provider": {
27
+
"name": "Anthropic"
28
+
},
29
+
"instructions": " You are an intelligent routing supervisor for a multi-agent system. Your primary job is to assess whether user input likely contains PII (personally identifiable information) to determine the most efficient processing route.\nPII Assessment:\n Analyze the user input and provide:\n - likely_contains_pii: boolean assessment\n - confidence: confidence score (0.0 to 1.0)\n - reasoning: clear explanation of your decision\n - recommended_route: either 'security_agent' or 'support_agent'\n\n Route to SECURITY_AGENT** if the text likely contains:\n - Email addresses, phone numbers, addresses\n - Names (first/last names, usernames)\n - Financial information (credit cards, SSNs, account numbers)\n - Sensitive personal data\n\n **Route to SUPPORT_AGENT** if the text appears to be:\n - General questions without personal details\n - Technical queries\n - Search requests\n - Educational content requests\n\n Analyze this user input and recommend the optimal route:\n"
30
+
},
31
+
"support-agent": {
32
+
"enabled": true,
33
+
"model": {
34
+
"name": "claude-3-5-haiku-20241022",
35
+
"parameters": {
36
+
"tools": [
37
+
{
38
+
"description": "Simple keyword search through knowledge base",
39
+
"name": "search_v1",
40
+
"parameters": {
41
+
"additionalProperties": false,
42
+
"properties": {
43
+
"query": {
44
+
"description": "Search query for keyword matching",
45
+
"type": "string"
46
+
},
47
+
"top_k": {
48
+
"description": "Number of results to return",
49
+
"type": "number"
50
+
}
51
+
},
52
+
"required": [
53
+
"query"
54
+
],
55
+
"type": "object"
56
+
},
57
+
"type": "function"
58
+
},
59
+
{
60
+
"description": "Semantic search using vector embeddings",
61
+
"name": "search_v2",
62
+
"parameters": {
63
+
"additionalProperties": false,
64
+
"properties": {
65
+
"query": {
66
+
"description": "Search query for semantic matching",
67
+
"type": "string"
68
+
},
69
+
"top_k": {
70
+
"description": "Number of results to return",
71
+
"type": "number"
72
+
}
73
+
},
74
+
"required": [
75
+
"query"
76
+
],
77
+
"type": "object"
78
+
},
79
+
"type": "function"
80
+
},
81
+
{
82
+
"description": "Reorders results by relevance using BM25 algorithm",
83
+
"name": "reranking",
84
+
"parameters": {
85
+
"additionalProperties": false,
86
+
"properties": {
87
+
"query": {
88
+
"description": "Original query for scoring",
89
+
"type": "string"
90
+
},
91
+
"results": {
92
+
"description": "Results to rerank",
93
+
"type": "array"
94
+
}
95
+
},
96
+
"required": [
97
+
"query",
98
+
"results"
99
+
],
100
+
"type": "object"
101
+
},
102
+
"type": "function"
103
+
}
104
+
]
105
+
}
106
+
},
107
+
"provider": {
108
+
"name": "Anthropic"
109
+
},
110
+
"instructions": "You are a helpful assistant with access to RAG tools: search_v1 (basic search), search_v2 (semantic vector search), and reranking (BM25 relevance scoring). When search results are available, prioritize information from those results over your general knowledge. Provide balanced, well-researched responses for international users.",
111
+
"tools": [
112
+
{
113
+
"description": "Simple keyword search through knowledge base",
114
+
"name": "search_v1",
115
+
"parameters": {
116
+
"additionalProperties": false,
117
+
"properties": {
118
+
"query": {
119
+
"description": "Search query for keyword matching",
120
+
"type": "string"
121
+
},
122
+
"top_k": {
123
+
"description": "Number of results to return",
124
+
"type": "number"
125
+
}
126
+
},
127
+
"required": [
128
+
"query"
129
+
],
130
+
"type": "object"
131
+
},
132
+
"type": "function"
133
+
},
134
+
{
135
+
"description": "Semantic search using vector embeddings",
136
+
"name": "search_v2",
137
+
"parameters": {
138
+
"additionalProperties": false,
139
+
"properties": {
140
+
"query": {
141
+
"description": "Search query for semantic matching",
142
+
"type": "string"
143
+
},
144
+
"top_k": {
145
+
"description": "Number of results to return",
146
+
"type": "number"
147
+
}
148
+
},
149
+
"required": [
150
+
"query"
151
+
],
152
+
"type": "object"
153
+
},
154
+
"type": "function"
155
+
},
156
+
{
157
+
"description": "Reorders results by relevance using BM25 algorithm",
0 commit comments