feat(config, engine): add support for Mistral AI provider and engine (#436)

* docs(CONTRIBUTING.md): update `TODO.md` reference (#435)

Signed-off-by: Emmanuel Ferdman <[email protected]>

* feat(config, engine): add support for Mistral AI provider and engine

* ```
feat(package): add mistralai and zod dependencies
```

* fix: recreate package-lock.json with node20

* fix: recreate package-lock.json with node v20.18.1 based on branch dev

---------

Signed-off-by: Emmanuel Ferdman <[email protected]>
Co-authored-by: Emmanuel Ferdman <[email protected]>
Co-authored-by: pedro-valentim <>

Author: pedro-valentim

package-lock.json

@@ -88,6 +88,7 @@
     "@clack/prompts": "^0.6.1",
     "@dqbd/tiktoken": "^1.0.2",
     "@google/generative-ai": "^0.11.4",
+    "@mistralai/mistralai": "^1.3.5",
     "@octokit/webhooks-schemas": "^6.11.0",
     "@octokit/webhooks-types": "^6.11.0",
     "axios": "^1.3.4",
@@ -99,6 +100,7 @@
     "ini": "^3.0.1",
     "inquirer": "^9.1.4",
     "openai": "^4.57.0",
-    "punycode": "^2.3.1"
+    "punycode": "^2.3.1",
+    "zod": "^3.23.8"
   }
 }

package.json

@@ -86,6 +86,48 @@ export const MODEL_LIST = {
     'llama-3.1-70b-versatile', // Llama 3.1 70B (Preview)
     'gemma-7b-it', // Gemma 7B
     'gemma2-9b-it' // Gemma 2 9B
+  ],
+
+  mistral: [
+    'ministral-3b-2410',
+    'ministral-3b-latest',
+    'ministral-8b-2410',
+    'ministral-8b-latest',
+    'open-mistral-7b',
+    'mistral-tiny',
+    'mistral-tiny-2312',
+    'open-mistral-nemo',
+    'open-mistral-nemo-2407',
+    'mistral-tiny-2407',
+    'mistral-tiny-latest',
+    'open-mixtral-8x7b',
+    'mistral-small',
+    'mistral-small-2312',
+    'open-mixtral-8x22b',
+    'open-mixtral-8x22b-2404',
+    'mistral-small-2402',
+    'mistral-small-2409',
+    'mistral-small-latest',
+    'mistral-medium-2312',
+    'mistral-medium',
+    'mistral-medium-latest',
+    'mistral-large-2402',
+    'mistral-large-2407',
+    'mistral-large-2411',
+    'mistral-large-latest',
+    'pixtral-large-2411',
+    'pixtral-large-latest',
+    'codestral-2405',
+    'codestral-latest',
+    'codestral-mamba-2407',
+    'open-codestral-mamba',
+    'codestral-mamba-latest',
+    'pixtral-12b-2409',
+    'pixtral-12b',
+    'pixtral-12b-latest',
+    'mistral-embed',
+    'mistral-moderation-2411',
+    'mistral-moderation-latest',
   ]
 };
 
@@ -101,6 +143,8 @@ const getDefaultModel = (provider: string | undefined): string => {
       return MODEL_LIST.gemini[0];
     case 'groq':
       return MODEL_LIST.groq[0];
+    case 'mistral':
+      return MODEL_LIST.mistral[0];
     default:
       return MODEL_LIST.openai[0];
   }
@@ -257,14 +301,15 @@ export const configValidators = {
       CONFIG_KEYS.OCO_AI_PROVIDER,
       [
         'openai',
+        'mistral',
         'anthropic',
         'gemini',
         'azure',
         'test',
         'flowise',
         'groq'
-      ].includes(value) || value.startsWith('ollama') || value.startsWith('mlx'),
-      `${value} is not supported yet, use 'ollama', 'mlx', anthropic', 'azure', 'gemini', 'flowise' or 'openai' (default)`
+      ].includes(value) || value.startsWith('ollama'),
+      `${value} is not supported yet, use 'ollama', 'mlx', 'anthropic', 'azure', 'gemini', 'flowise', 'mistral' or 'openai' (default)`
     );
 
     return value;
@@ -310,6 +355,7 @@ export enum OCO_AI_PROVIDER_ENUM {
   TEST = 'test',
   FLOWISE = 'flowise',
   GROQ = 'groq',
+  MISTRAL = 'mistral',
   MLX = 'mlx'
 }
 

src/commands/config.ts

@@ -3,6 +3,7 @@ import { OpenAIClient as AzureOpenAIClient } from '@azure/openai';
 import { GoogleGenerativeAI as GeminiClient } from '@google/generative-ai';
 import { AxiosInstance as RawAxiosClient } from 'axios';
 import { OpenAI as OpenAIClient } from 'openai';
+import { Mistral as MistralClient } from '@mistralai/mistralai';
 
 export interface AiEngineConfig {
   apiKey: string;
@@ -17,7 +18,8 @@ type Client =
   | AzureOpenAIClient
   | AnthropicClient
   | RawAxiosClient
-  | GeminiClient;
+  | GeminiClient
+  | MistralClient;
 
 export interface AiEngine {
   config: AiEngineConfig;

src/engine/Engine.ts

@@ -0,0 +1,82 @@
+import axios from 'axios';
+import { Mistral } from '@mistralai/mistralai';
+import { OpenAI } from 'openai';
+import { GenerateCommitMessageErrorEnum } from '../generateCommitMessageFromGitDiff';
+import { tokenCount } from '../utils/tokenCount';
+import { AiEngine, AiEngineConfig } from './Engine';
+import {
+  AssistantMessage as MistralAssistantMessage,
+  SystemMessage as MistralSystemMessage,
+  ToolMessage as MistralToolMessage,
+  UserMessage as MistralUserMessage
+} from '@mistralai/mistralai/models/components';
+
+export interface MistralAiConfig extends AiEngineConfig {}
+export type MistralCompletionMessageParam = Array<
+| (MistralSystemMessage & { role: "system" })
+| (MistralUserMessage & { role: "user" })
+| (MistralAssistantMessage & { role: "assistant" })
+| (MistralToolMessage & { role: "tool" })
+>
+
+export class MistralAiEngine implements AiEngine {
+  config: MistralAiConfig;
+  client: Mistral;
+
+  constructor(config: MistralAiConfig) {
+    this.config = config;
+
+    if (!config.baseURL) {
+      this.client = new Mistral({ apiKey: config.apiKey });
+    } else {
+      this.client = new Mistral({ apiKey: config.apiKey, serverURL: config.baseURL });
+    }
+  }
+
+  public generateCommitMessage = async (
+    messages: Array<OpenAI.Chat.Completions.ChatCompletionMessageParam>
+  ): Promise<string | null> => {
+    const params = {
+      model: this.config.model,
+      messages: messages as MistralCompletionMessageParam,
+      topP: 0.1,
+      maxTokens: this.config.maxTokensOutput
+    };
+
+    try {
+      const REQUEST_TOKENS = messages
+        .map((msg) => tokenCount(msg.content as string) + 4)
+        .reduce((a, b) => a + b, 0);
+
+      if (
+        REQUEST_TOKENS >
+        this.config.maxTokensInput - this.config.maxTokensOutput
+      )
+        throw new Error(GenerateCommitMessageErrorEnum.tooMuchTokens);
+
+      const completion = await this.client.chat.complete(params);
+
+      if (!completion.choices)
+        throw Error('No completion choice available.')
+      
+      const message = completion.choices[0].message;
+
+      if (!message || !message.content)
+        throw Error('No completion choice available.')
+
+      return message.content as string;
+    } catch (error) {
+      const err = error as Error;
+      if (
+        axios.isAxiosError<{ error?: { message: string } }>(error) &&
+        error.response?.status === 401
+      ) {
+        const mistralError = error.response.data.error;
+
+        if (mistralError) throw new Error(mistralError.message);
+      }
+
+      throw err;
+    }
+  };
+}

src/engine/mistral.ts

@@ -6,6 +6,7 @@ import { FlowiseEngine } from '../engine/flowise';
 import { GeminiEngine } from '../engine/gemini';
 import { OllamaEngine } from '../engine/ollama';
 import { OpenAiEngine } from '../engine/openAi';
+import { MistralAiEngine } from '../engine/mistral';
 import { TestAi, TestMockType } from '../engine/testAi';
 import { GroqEngine } from '../engine/groq';
 import { MLXEngine } from '../engine/mlx';
@@ -44,6 +45,9 @@ export function getEngine(): AiEngine {
     case OCO_AI_PROVIDER_ENUM.GROQ:
       return new GroqEngine(DEFAULT_CONFIG);
 
+    case OCO_AI_PROVIDER_ENUM.MISTRAL:
+      return new MistralAiEngine(DEFAULT_CONFIG);
+
     case OCO_AI_PROVIDER_ENUM.MLX:
       return new MLXEngine(DEFAULT_CONFIG);