Update documentations

ditrit · Dec 3, 2024 · f626940 · f626940
1 parent d88103a
commit f626940
Show file tree

Hide file tree

Showing 4 changed files with 116 additions and 43 deletions.
diff --git a/.env.example b/.env.example
@@ -0,0 +1 @@
+DECRYPTION_KEY=123456789
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -91,7 +91,7 @@ In order to handle a new AI, you need to add it to:
 - Create a new handler in `handlers` folder and in the new handler class should inherit from `BaseHandler`.
   - Currently you need to implement the `generate` method.
 - Add the new handler in the factory `handlers/Factory.py`.
-- Add a section in `configuration/configurationManager.py` with the name of the new AI and its settings.
+- Add a file called `configuration_description.json` that will describe the new AI settings and how to configure it.
 
 ### Example of Handler
 
@@ -129,46 +129,97 @@ class Factory:
 
 ### Example of configuration
 
-Finally, you need to add the new AI in the `configuration.json` file.
+Finally, you will have a `configuration` similar to:
 
 ```json
 {
     "pluginPreferences":{
         "default": "ollama"
     },
-    "ai-models":{
-        "ollama":
-        {
-            "base_url": "http://localhost:11434/api",
-            "models": ["mistral"],
-            "defaultModel": "mistral",
-            "modelFiles": {
-                "generate": {
-                    "default": "default-mistral-modelfile_generate",
-                    "@ditrit/kubernator-plugin": "default-kubernetes-mistral-modelfile_generate",
-                    "@ditrit/githubator-plugin": "default-githubactions-mistral-modelfile_generate"
-                },
-                "message": {
-                    "default": "default-mistral-modelfile_message",
-                    "@ditrit/kubernator-plugin": "default-kubernetes-mistral-modelfile_message",
-                    "@ditrit/githubator-plugin": "default-githubactions-mistral-modelfile_message"
-                }
+    "ollama":
+    {
+        "base_url": "http://localhost:11434/api",
+        "models": ["mistral"],
+        "defaultModel": "mistral",
+        "modelFiles": {
+            "generate": {
+                "default": "default-mistral-modelfile_generate",
+                "@ditrit/kubernator-plugin": "default-kubernetes-mistral-modelfile_generate",
+                "@ditrit/githubator-plugin": "default-githubactions-mistral-modelfile_generate"
+            },
+            "message": {
+                "default": "default-mistral-modelfile_message",
+                "@ditrit/kubernator-plugin": "default-kubernetes-mistral-modelfile_message",
+                "@ditrit/githubator-plugin": "default-githubactions-mistral-modelfile_message"
             }
-        },
-        "MyAI": {
-            "base_url": "http://localhost:8080/api",
-            "otherSetting": "value"
-        }    
+        }
+    },
+    "MyAI": {
+        "base_url": "http://localhost:8080/api",
+        "otherSetting": "value"
     }
 }
 ```
 
+A configuration_description.json file is essential for integrating any AI system with Leto-Modelizer-Admin, as it defines the fields administrators must fill out to configure the AI properly.
+
+Here are the description of the configuration fields:
+
+- **`handler`**
+  - **Purpose**: Identifies the AI system or plugin this configuration belongs to (e.g., "ollama").
+  - **Use**: Associates the field with a specific handler or module.
+
+- **`key`**
+  - **Purpose**: A unique identifier for the field used internally by the system.
+  - **Use**: Represents the configuration parameter (e.g., `base_url`, `defaultModel`).
+
+- **`type`**
+  - **Purpose**: Specifies the type of input expected in the UI.
+  - **Values**:
+    - `"text"`: Single-line text input.
+    - `"select"`: To create a dropdown menu.
+    - `"textarea"`: Multi-line input for longer text.
+
+- **`values`**
+  - **Purpose**: Defines predefined selectable options (e.g., dropdown menu).
+  - **Use**: Empty if free-form input is allowed; can be populated for fixed choices.
+
+- **`defaultValue`**
+  - **Purpose**: The pre-filled value shown to the user when the field is displayed.
+  - **Use**: Ensures default configurations or suggested values are provided (e.g., `http://localhost:11434/api`).
+
+- **`label`**
+  - **Purpose**: The user-facing name of the field displayed in the UI.
+  - **Use**: Briefly explains what the field represents (e.g., "Ollama server URL").
+
+- **`title`**
+  - **Purpose**: A short explanatory tooltip shown when hovering over the field in the UI.
+  - **Use**: Provides additional context (e.g., "Define the URL of the Ollama server").
+
+- **`description`**
+  - **Purpose**: A detailed explanation of the field's purpose or expected input.
+  - **Use**: Guides users on how to fill the field, especially for complex settings.
+
+- **`pluginDependent`**
+  - **Purpose**: Indicates whether the field is specific to a plugin.
+  - **Values**:
+    - `true`: Applies only to a specific plugin (e.g., plugin-specific instructions).
+    - `false`: Applies globally.
+
+- **`required`**
+  - **Purpose**: Specifies if the field must be completed for the configuration to be valid.
+  - **Values**:
+    - `true`: Field is mandatory.
+    - `false`: Field is optional.
+
+
+See the `configuration_description.json` of Ollama for an example.
+
 ### Adding a new Ollama model file
 
-First add your modelfile in the `src/handlers/Ollama/ModelFiles` folder.
 You can find how to create your model file [here](https://github.com/ollama/ollama/blob/main/docs/modelfile.md)
 
-Then add your model file in the `configuration.json` file, in the `ai-models.ollama.modelFiles` section.
+Then add your model file in the `description_configuration.json` file, then through Leto-Modelizer-Admin you can configure it.
 
 ## How to launch e2e tests with Behave
 

diff --git a/README.md b/README.md
@@ -13,10 +13,22 @@ Built on top of Python's robust web framework, the Leto-Modelizer Proxy API offe
 
 Currently this project use Ollama or Gemini in order to generate code.
 
+## .env
+
+In order to configure the API, you need to create a `.env` file in the root of the project.
+You can find an example of this file [here](.env.example).
+
+The `.env` file should contain the following variables:
+
+| Variable       | Description                          |
+| -------------- | ------------------------------------ |
+| DECRYPTION_KEY | Key for decrypting the configuration |
+
+
 ## Configuration
 
-To configure the API, you need to edit the `configuration/configuration.json` file (default location).
-Or you can set the environment variables `API_CONFIGURATION` and put the `configuration.json` file in that path`.
+To configure the Proxy API, you need to have the **Leto-Modelizer-Admin running**.
+Then, you have to go in the 'AI settings' section to configure the Proxy API.
 
 Currently the configuration has the following settings:
 
@@ -26,7 +38,7 @@ Currently the configuration has the following settings:
 | ollama             | A dictionary containing the ollama configuration (cf: next section).                            |
 | Gemini             | A dictionary containing the Gemini configuration (cf: next section).                            |
 
-Here is an example of the `configuration.json` file:
+Here is an example of the `configuration`:
 
 ```json
 {
@@ -71,7 +83,8 @@ Here is an example of the `configuration.json` file:
 ```
 
 The mandatory `default` key is used to specify which AI model to use by default for every plugin.
-Moreover, the AI precised for each plugin in the `pluginPreferences` key must exist in the `configuration.json`.
+Moreover, the AI precised for each plugin in the `pluginPreferences` key must exist in the configuration.
+
 
 ### Ollama
 
@@ -90,7 +103,7 @@ Currently only the default model is used. But later, we will be able to handle m
 ### Gemini
 
 Gemini can be found here: https://github.com/google-gemini/
-**Don't forget to set the `key` in the `configuration.json` by creating an account and getting an API key.**
+**Don't forget to set the `key` in the configuration by creating an account and getting an API key.**
 Gemini has the following settings:
 
 | Setting            | Description                                                                                              |
@@ -137,7 +150,7 @@ cf: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/ins
 ### Gemini API
 
 For Gemini, you need to create an account and get an API key, on this website: https://ai.google.dev/
-Then you just need to add it in the `configuration.json` file, in the `ai-models` section.
+Then you just need to add it in the `configuration`, in the `gemini` section of the Leto-Modelizer-Admin.
 
 
 ## How to launch the API using hypercorn
@@ -206,8 +219,9 @@ docker build -t leto-modelizer-ai-proxy .
 Then run the image:
 
 ```sh
-docker compose up -f docker-compose-nvidia.yaml
+docker compose --env-file .env -f docker-compose-nvidia.yaml up
 ```
+**NOTE:** Even using Ollama with docker, you have to pull the model files first (connect to the container then) using the command ```ollama pull mistral```.
 
 ### Without NVIDIA GPU
 
@@ -220,7 +234,7 @@ docker build -t leto-modelizer-ai-proxy .
 Then run the image:
 
 ```sh
-docker compose up -f docker-compose.yaml
+docker compose --env-file .env -f docker-compose.yaml up
 ```
 
 The initialize script initializes everything the handlers needs in order to work (for instance the Ollama handler needs to be initialized with the model files).
@@ -231,6 +245,7 @@ python3 initialize.py
 Once your docker is running and initialized, you can request it on this url: ```http://localhost:8585/```
 
 **NOTE:** You do not need to launch the initialize script everytime you launch the API with docker, once is enough since it target the AI (i.e model files for Ollama).
+**NOTE:** Even using Ollama with docker, you have to pull the model files first (connect to the container then) using the command ```ollama pull mistral```.
 
 ## Endpoint
 

diff --git a/changelog.md b/changelog.md
@@ -5,21 +5,27 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html)
 
+## [Unreleased]
+
+### Added
+
+ - Configuration is now handle through the Leto-Modelizer-Admin, the user give the configuration for the AIs and not through the file anymore.
+
 ## [1.0.0] - 2024/10/15
 
 ### Added
 
- - Setup the project
- - Create Ollama handler in order to generate code through the /api/diagram route
- - Handle Modelfiles for Ollama (terraform, kubernetes and github actions)
- - Migrate to python 3.12
- - Migrate from uvicorn to hypercorn
- - Add new Ollama model files (for generate and message)
- - Handle new /api/message endpoint for Ollama, that send messages to the Ollama AI and get a response with the associated context
+ - Setup the project.
+ - Create Ollama handler in order to generate code through the /api/diagram route.
+ - Handle Modelfiles for Ollama (terraform, kubernetes and github actions).
+ - Migrate to python 3.12.
+ - Migrate from uvicorn to hypercorn.
+ - Add new Ollama model files (for generate and message).
+ - Handle new /api/message endpoint for Ollama, that send messages to the Ollama AI and get a response with the associated context.
  - Added new Gemini model files (for generate and message). Conversation with a context is not supported.
- - Added Docker compose (works only with nvidia gpu)
+ - Added Docker compose (works only with nvidia gpu).
  - The initialize script is now separated from the main script. It can be launched from the root folder, anytime.
- - Add health endpoint
+ - Add health endpoint.
 
 [1.0.0]: https://github.com/ditrit/leto-modelizer-ai-proxy/blob/main/changelog.md#1.0.0