Wiki source code of Intelligence artificielle

Last modified by Aurelie Bertrand on 2025/11/14 11:59
Show last authors
1 {{ddtoc/}}
2
3 ----
4
5 This section allows you to configure the options for features using artificial intelligence: [[data enrichment with open data>>doc:Digdash.user_guide.studio.open_data.WebHome]], generation of a transformation function, [[display of highlights>>doc:Digdash.user_guide.dashboard.Display_highlights.WebHome]] and [[DigDash Agent>>doc:Digdash.Agent.WebHome]].
6
7 [[image:AI_en.png||alt="Artificial intelligence"]]
8
9 = Open data =
10
11 Here you can activate the [[open data functionality>>doc:Digdash.user_guide.studio.open_data.WebHome]]:
12
13 * Check the **Enable the open data functionality** box.
14 ➡ The** Enrich with open data** command is then available in the data model context menu.
15 [[image:Model_enrich_open_data_EN.png||alt="Enrich model" height="504" width="707"]]
16
17 It is also possible to automatically detect whether open data compatible with your data is available, when your data model is created. To do this:
18
19 * Check the **Enable open data functionality when creating a data model** box **.**
20 ➡ If open data compatible with your data is available, it will be offered to you when you create your data model (after clicking on the **Finish** button when configuring the data model in the Studio).
21
22 The field **The identifier of the role containing the open data models **contains the identifier of the role dedicated to Digdash Open Data models.
23
24 = LLM (Large Language Model) =
25
26 Here you can activate and specify the LLM (large language model) used by the AI wizard for [[the generation of transformation functions>>doc:Digdash.user_guide.studio.Create_datamodel.transform_data.transform_data_tab.WebHome||anchor="IA"]].
27
28 1. Tick the **Activate LLM** box.
29 1. Select the **LLM Provider **from the drop-down list.
30
31 (% class="box infomessage" %)
32 (((
33 The **Ollama** LLM provider requires local installation. You can consult the Ollama documentation [[https:~~/~~/github.com/ollama/ollama/>>https://github.com/ollama/ollama/]] for more details.
34 )))
35
36 (% start="3" %)
37 1. Fill in the following information:(((
38 |=(% style="width: 181px;" %) |=(% style="width: 286px;" %)OpenAI|=(% style="width: 363px;" %)Google Gemini|=(% style="width: 482px;" %)Ollama
39 |=(% style="width: 181px;" %)**Server URL**|(% style="width:286px" %)https:~/~/api.openai.com|(% style="width:363px" %)(((
40 https:~/~/generativelanguage.googleapis.com/v1beta/models/nom_du_modèle
41
42 The URL contains the model used by Gemini. Thus,// model_name //should be replaced by the name of the chosen model. For example:
43 https:~/~/generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-latest:generateContent
44
45 Click on the following link to consult the list of available models:
46 [[https:~~/~~/ai.google.dev/gemini-api/docs/models/gemini?hl=fr>>https://ai.google.dev/gemini-api/docs/models/gemini?hl=fr]]
47 )))|(% style="width:482px" %)(((
48 Enter the server URL in the following format:
49 //http:~/~/[server]:[port]
50 For example : //[[http:~~/~~/lab1234.lan.digdash.com:11434>>http://lab3090.lan.digdash.com:11434]]
51 )))
52 |=(% style="width: 181px;" %)**API key**|(% style="width:286px" %)(((
53 Enter your API key.
54
55 See the paragraph [[Configuring an OpenAI API key>>doc:.LLM_API_key.WebHome||anchor="OpenAI"]] if necessary.
56 )))|(% style="width:363px" %)(((
57 Enter your API key.
58
59 See the paragraph [[Configuring a Gemini API key>>doc:.LLM_API_key.WebHome||anchor="Gemini"]] if necessary.
60 )))|(% style="width:482px" %)Ollama does not require an API key.
61 |=(% style="width: 181px;" %)**Model**|(% style="width:286px" %)(((
62 Enter the name of the chosen model.
63 For example, gpt-3.5-turbo.
64
65 Click on the following link to consult the list of available models:
66 [[https:~~/~~/platform.openai.com/docs/models>>https://platform.openai.com/docs/models]]
67 )))|(% style="width:363px" %)(((
68 The model is not entered here but directly in the server URL.
69 The field must remain empty.
70 )))|(% style="width:482px" %)(((
71 Enter the model identifier.
72
73 We recommend the following models:
74
75 * **Codestral 22B**: LLM specialising in code generation (small, high-performance model).
76 With the level of [[quantization level>>https://huggingface.co/docs/optimum/concept_guides/quantization]] Q4_K_M, the identifier is //codestral:22b-v0.1-q4_K_M//
77 * **Llama 3.3 70B **: General-purpose LLM for code-generating tasks.
78 With the [[quantization level>>https://huggingface.co/docs/optimum/concept_guides/quantization]] Q4_K_M, the identifier is //llama3.3:70b-instruct-q4_K_M//
79
80 Click on the following link to consult the list of available models:
81 [[https:~~/~~/ollama.com/search>>https://ollama.com/search]]
82 )))
83 )))
84 1. Click on **Save**.
85
86 To return to the default values, click **Reset**.
87
88 == Setting prompts ==
89
90 The prompts (or instructions) used by default for generating data transformations are stored in the ///home/digdash/webapps/ddenterpriseapi/WEB-INF/classes/resources/llm// directory //.// There is a prompt for each provider. ❗These prompts MUST NOT be changed.
91
92 You can define a custom prompt on the same model, keeping the last lines :
93
94 {{{request: #/*REQUEST_CLIENT*/#}}}
95
96 In order to be taken into account, it must be named //custom.prompt// and placed in the ///home/digdash/appdata/default/Enterprise Server/ddenterpriseapi/config// directory //.//
97
98 = Highlights{{id name="highlight"/}} =
99
100 (% class="box" %)
101 (((
102 💡 Consult the page [[View highlights>>doc:Digdash.user_guide.dashboard.Display_highlights.WebHome]] page for more details on the use of highlights.
103 )))
104
105 Some default values for [[highlights>>doc:Digdash.user_guide.dashboard.Display_highlights.WebHome]] parameters ars eset by default. You can change these values if you wish to influence the way in which highlights are identified.
106
107 |=(% style="background-color: grey; width: 438px;" %)(% style="color:#ffffff" %)Parameter|=(% style="background-color: grey; width: 1053px;" %)(% style="color:#ffffff" %)Description
108 |(% style="width:438px" %)**Minimum number of members for Mega**|(% style="width:1053px" %)(((
109 Minimum number of members the dimension must contain to be able to identify a mega contributor.
110 By default, the dimension must contain at least 5 members.
111 )))
112 |(% style="width:438px" %)**Minimum percentage for Mega**|(% style="width:1053px" %)Minimum percentage of the total sum that the member must represent to be a mega contributor.
113 By default, a member must contribute at least 40% of a given measure to be able to be a mega contributor.
114 |(% style="width:438px" %)**Minimum number K for TopK**|(% style="width:1053px" %)Minimum number of members contributing at least the//"Minimum percentage for TopK"// to a given measure (total sum).
115 By default, at least 2 members must contribute at least 40% of a given measure to be top contributors.
116 |(% style="width:438px" %)**Maximum number K for TopK**|(% style="width:1053px" %)Maximum number of members contributing at least the//"Minimum percentage for TopK"// to a given measure (total sum).
117 By default, a maximum of 5 members must contribute at least 40% of a given measure to be top contributors.
118 |(% style="width:438px" %)**Percentage K for TopK**|(% style="width:1053px" %)Percentage of the number of members used to determine K.
119 By default, K is equal to 33%. A minimum of 6 members is required to obtain a Top2.
120 |(% style="width:438px" %)**Minimum percentage for TopK**|(% style="width:1053px" %)Minimum percentage of a given measure to which the K best members must contribute to be Top contributors.
121 By default, the K best members must contribute at least 40% to a given measure (total sum) to be Top contributors.
122 |(% style="width:438px" %)**Lowest lower bound for correlation**|(% style="width:1053px" %)Minimum correlation coefficient for a dimension with 10 members for correlation to be taken into account.
123 By default, the minimum correlation coefficient is 0.7.
124 |(% style="width:438px" %)**Highest lower bound for correlation**|(% style="width:1053px" %)Minimum correlation coefficient for a dimension with 50 or more members for correlation to be taken into account.
125 By default, the minimum correlation coefficient is 0.3.
126 |(% style="width:438px" %)**Limit for correlation**|(% style="width:1053px" %)Limit of the correlation coefficient beyond which the relationship is no longer considered a correlation.
127
128 (% class="box infomessage" %)
129 (((
130 ℹ The correlation search is only performed if the selected measure has a Mega Contributor, Top K Contributors or Pareto.
131 )))
132
133 = Agent =
134
135 This section allows you to configure the server parameters for the use of DigDash Agent. See paragraph [[Configuring DigDash server settings>>doc:Digdash.Agent.install_guide_Agent.WebHome||anchor="Paramètres_serveur"]] for a detailed description.