Hugging Face's logo

narcolepticchicken
/
agent-cost-optimizer

Safetensors
Model card Files
xet
 Community
agent-cost-optimizer
File size: 4,033 Bytes
9a3f54b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
# Agent Cost Optimizer Configuration

project_name: "agent-cost-optimizer"
trace_storage_path: "./traces"

# Model tiers: 1=tiny, 2=cheap, 3=medium, 4=frontier, 5=specialist, 6=verifier
models:
  tiny_local:
    model_id: "tiny_local"
    provider: "local"
    cost_per_1k_input: 0.0001
    cost_per_1k_output: 0.0002
    cost_per_1k_reasoning: 0.0
    latency_ms_estimate: 200
    strength_tier: 1
    max_context: 32768
    supports_tools: true
    supports_reasoning: false
    cache_discount_rate: 0.5
  
  cheap_cloud:
    model_id: "cheap_cloud"
    provider: "cloud"
    cost_per_1k_input: 0.0005
    cost_per_1k_output: 0.001
    cost_per_1k_reasoning: 0.0
    latency_ms_estimate: 500
    strength_tier: 2
    max_context: 128000
    supports_tools: true
    supports_reasoning: false
    cache_discount_rate: 0.5
  
  medium:
    model_id: "medium"
    provider: "cloud"
    cost_per_1k_input: 0.003
    cost_per_1k_output: 0.006
    cost_per_1k_reasoning: 0.0
    latency_ms_estimate: 800
    strength_tier: 3
    max_context: 128000
    supports_tools: true
    supports_reasoning: false
    cache_discount_rate: 0.5
  
  frontier:
    model_id: "frontier"
    provider: "cloud"
    cost_per_1k_input: 0.01
    cost_per_1k_output: 0.03
    cost_per_1k_reasoning: 0.01
    latency_ms_estimate: 1500
    strength_tier: 4
    max_context: 200000
    supports_tools: true
    supports_reasoning: true
    cache_discount_rate: 0.5
  
  specialist:
    model_id: "specialist"
    provider: "cloud"
    cost_per_1k_input: 0.015
    cost_per_1k_output: 0.045
    cost_per_1k_reasoning: 0.015
    latency_ms_estimate: 2000
    strength_tier: 5
    max_context: 128000
    supports_tools: true
    supports_reasoning: true
    cache_discount_rate: 0.5

tools:
  search:
    tool_name: "search"
    cost_per_call: 0.002
    latency_ms_estimate: 500
    cacheable: false
    requires_verification: false
    max_retries: 3
  
  retrieve:
    tool_name: "retrieve"
    cost_per_call: 0.001
    latency_ms_estimate: 300
    cacheable: true
    requires_verification: false
    max_retries: 3
  
  code_execution:
    tool_name: "code_execution"
    cost_per_call: 0.005
    latency_ms_estimate: 1000
    cacheable: false
    requires_verification: true
    max_retries: 2
  
  linter:
    tool_name: "linter"
    cost_per_call: 0.001
    latency_ms_estimate: 200
    cacheable: false
    requires_verification: false
    max_retries: 1
  
  file_read:
    tool_name: "file_read"
    cost_per_call: 0.0005
    latency_ms_estimate: 100
    cacheable: true
    requires_verification: false
    max_retries: 3
  
  compliance_check:
    tool_name: "compliance_check"
    cost_per_call: 0.01
    latency_ms_estimate: 1500
    cacheable: false
    requires_verification: true
    max_retries: 2

verifiers:
  verifier_medium:
    verifier_model_id: "verifier_medium"
    cost_per_call: 0.005
    latency_ms_estimate: 800
    confidence_threshold: 0.8

# Routing policy
routing_policy:
  name: "default"
  type: "cascade"
  threshold_confidence: 0.7
  max_cascade_depth: 3
  enable_verifier_fallback: true
  enable_escalation: true

# Cost weights
model_cost_weight: 1.0
tool_cost_weight: 1.0
verifier_cost_weight: 1.0
latency_weight: 0.1
retry_penalty_weight: 2.0
false_done_penalty: 10.0
unsafe_cheap_model_penalty: 20.0
missed_escalation_penalty: 15.0

# Module toggles
enable_telemetry: true
enable_classifier: true
enable_router: true
enable_context_budgeter: true
enable_cache_layout: true
enable_tool_gate: true
enable_verifier_budgeter: true
enable_retry_optimizer: true
enable_meta_tool_miner: true
enable_early_termination: true

# Cache-aware layout
cache_prefix_stable:
  - system_rules
  - tool_descriptions
  - user_preferences
cache_suffix_dynamic:
  - user_message
  - retrieved_docs
  - recent_trace
  - artifacts

# Early termination
doom_max_cost_ratio: 3.0
doom_max_retries: 3
doom_no_progress_steps: 5
doom_verifier_disagreement_threshold: 2

# Meta-tool mining
meta_tool_min_frequency: 5
meta_tool_min_success_rate: 0.8