File size: 11,170 Bytes
83aabd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
{
  "openapi": "3.0.0",
  "info": {
    "description": "The API recommends sentences to be added or removed from a given input prompt to make it more responsible.",
    "version": "1.0.0",
    "title": "Responsible Prompt Recommendation API",
    "license": {
      "name": "Apache 2.0",
      "url": "https://www.apache.org/licenses/LICENSE-2.0.html"
    }
  },
  "paths": {
    "/recommend": {
      "get": {
        "tags": [
          "recommend"
        ],
        "summary": "Obtain responsible prompt recommendations",
        "description": "Given an input prompt with one or more sentences, returns up to five recommendations of sentences and its respective positive value representations to be added at the end of the input prompt, and up to five recommendations for removal of potentially harmful sentences in the given input prompt.",
        "parameters": [
          {
            "name": "prompt",
            "in": "query",
            "description": "Provide an input prompt that would be sent to an LLM. Note that this recommendation is performed at prompting time and this API is not generating content. It uses a curated dataset of values and prompts to compute the similarity to the input sentences provided and the dataset sentences. Please note that similarity informed is cosine similarity. Please note that the index informed in the removal recommendations is 0-indexed.",
            "required": true,
            "schema": {
              "type": "string"
            }
          },
          {
            "name": "add_lower_threshold",
            "in": "header",
            "description": "The lower threshold is used to select centroids of sentences based on the similarity of the provided input and the centroid of existing clusters of sentences based on values. The default value for the lower threshold is 0.3",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.3"
          },
          {
            "name": "add_upper_threshold",
            "in": "header",
            "description": "The upper sentence threshold is being used as a ceiling value to avoid redundance, meaning that we will not recommend adding something that is already there. The default value for the upper threshold is 0.5",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.5"
          },
          {
            "name": "remove_lower_threshold",
            "in": "header",
            "description": "The lower threshold is used to select centroids of sentences based on the similarity of the provided input and the centroid of existing clusters of sentences based on values.. The default value for the lower removal threshold is 0.3",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.1"
          },
          {
            "name": "remove_upper_threshold",
            "in": "header",
            "description": "An upper removal threshold limit is used to prevent false positives and remove a sentence in the prompt similar to one of our adversarial prompts. The default value for the upper removal threshold is 0.5",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.5"
          },
          {
            "name": "model_id",
            "in": "header",
            "description": "The model id to be used. The default choice is: sentence-transformers/all-minilm-l6-v2",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "sentence-transformers/all-minilm-l6-v2"
          }
        ],
        "responses": {
          "200": {
            "description": "successful recommendation"
          },
          "404": {
            "description": "invalid prompt"
          }
        }
      }
    },
    "/get_thresholds": {
      "get": {
        "tags": [
          "thresholds"
        ],
        "summary": "Function that recommends thresholds given an array of prompts",
        "description": "Given an input array of prompts, returns thresholds for a selected model.",
        "parameters": [
          {
            "name": "prompts",
            "in": "query",
            "description": "Prompt array.",
            "required": true,
            "schema": {
              "type": "string",
              "maxLength": 1024
            }
          },
          {
            "name": "model_id",
            "in": "header",
            "description": "The model id to be used. The default choice is: sentence-transformers/all-minilm-l6-v2",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "sentence-transformers/all-minilm-l6-v2"
          }
        ],
        "responses": {
          "200": {
            "description": "successful threshold display"
          },
          "404": {
            "description": "no threshold display"
          }
        }
      }
    },
    "/recommend_local": {
      "get": {
        "tags": [
          "recommend_local"
        ],
        "summary": "Obtain responsible prompt recommendations from a model running locally",
        "description": "Given an input prompt with one or more sentences, returns up to five recommendations of sentences and its respective positive value representations to be added at the end of the input prompt, and up to five recommendations for removal of potentially harmful sentences in the given input prompt.",
        "parameters": [
          {
            "name": "prompt",
            "in": "query",
            "description": "Provide an input prompt that will be sent to a local LLM. Note that this recommendation is performed at prompting time and this model is not generating content. It uses a curated dataset of values and prompts to compute the similarity to the input sentences provided and the dataset sentences. Please note that similarity informed is cosine similarity. Please note that the index informed in the removal recommendations is 0-indexed.",
            "required": true,
            "schema": {
              "type": "string"
            }
          },
          {
            "name": "add_lower_threshold",
            "in": "header",
            "description": "The lower threshold is used to select centroids of sentences based on the similarity of the provided input and the centroid of existing clusters of sentences based on values. The default value for the lower threshold is 0.3",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.3"
          },
          {
            "name": "add_upper_threshold",
            "in": "header",
            "description": "The upper sentence threshold is being used as a ceiling value to avoid redundance, meaning that we will not recommend adding something that is already there. The default value for the upper threshold is 0.5",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.5"
          },
          {
            "name": "remove_lower_threshold",
            "in": "header",
            "description": "The lower threshold is used to select centroids of sentences based on the similarity of the provided input and the centroid of existing clusters of sentences based on values.. The default value for the lower removal threshold is 0.3",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.1"
          },
          {
            "name": "remove_upper_threshold",
            "in": "header",
            "description": "An upper removal threshold limit is used to prevent false positives and remove a sentence in the prompt similar to one of our adversarial prompts. The default value for the upper removal threshold is 0.5",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.5"
          },
          {
            "name": "model_id",
            "in": "header",
            "description": "The local model id to be used. The default choice is: sentence-transformers/all-minilm-l6-v2",
            "required": true,
            "schema": {
              "type": "string"
            },
            "example": "sentence-transformers/all-minilm-l6-v2"
          },
          {
            "name": "model_path",
            "in": "header",
            "description": "The local model files path. The default choice is: ./models/all-MiniLM-L6-v2/",
            "required": true,
            "schema": {
              "type": "string"
            },
            "example": "./models/all-MiniLM-L6-v2/"
          }
        ],
        "responses": {
          "200": {
            "description": "successful recommendation"
          },
          "404": {
            "description": "invalid prompt"
          }
        }
      }
    },
    "/demo_inference": {
      "get": {
        "tags": [
          "demo_inference"
        ],
        "summary": "Obtain a response to the user query from LLMs hosted on HuggingFace.",
        "description": "Given an input prompt, returns the response from text completion using LLMs hosted on HuggingFace.",
        "parameters": [
          {
            "name": "prompt",
            "in": "query",
            "description": "Text input used to generate a response.",
            "required": false,
            "schema": {
              "type": "string"
            }
          },
          {
            "name": "model_id",
            "in": "query",
            "description": "The id from HuggingFace of the LLM to be accessed. The default choice is: meta-llama/Llama-4-Scout-17B-16E-Instruct",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "meta-llama/Llama-4-Scout-17B-16E-Instruct"
          },
          {
            "name": "max_new_tokens",
            "in": "query",
            "description": "An upper bound for the number of tokens that can be generated for a response",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "1000"
          },
          {
            "name": "temperature",
            "in": "query",
            "description": "What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.",
            "required": false,
            "schema": {
              "type": "string"
            },
            "example": "0.5"
          }
        ],
        "responses": {
          "200": {
            "description": "successful text completion"
          },
          "404": {
            "description": "invalid input parameters or model unavailable"
          }
        }    
      }
    }
  }
}