feat: Add adaptors for ollama and openai

package.json

@@ -89,6 +89,8 @@
               "codellama/CodeLlama-13b-hf",
               "Phind/Phind-CodeLlama-34B-v2",
               "WizardLM/WizardCoder-Python-34B-V1.0",
+              "ollama/CodeLlama-7b-code",
+              "openai/chat-gpt-3.5-turbo",
               "Custom"
             ],
             "default": "bigcode/starcoder",
@@ -180,6 +182,22 @@
             "default": null,
             "description": "Tokenizer configuration for the model, check out the documentation for more details"
           },
+          "llm.adaptor": {
+            "type": "string",
+            "enum": [
+              "huggingface",
+              "ollama",
+              "openai",
+              "tgi"
+            ],
+            "default": "huggingface",
+            "desciption": "Completion provider to adapt to"
+          },
+            "llm.requestBody": {
+            "type": ["object", "null"],
+            "default": null,
+            "description": "Request body to be sent to the adaptor"
+          },
           "llm.documentFilter": {
             "type": [
               "object",

src/configTemplates.ts

@@ -1,4 +1,4 @@
-const templateKeys = ["bigcode/starcoder", "codellama/CodeLlama-13b-hf", "Phind/Phind-CodeLlama-34B-v2", "WizardLM/WizardCoder-Python-34B-V1.0", "Custom"] as const;
+const templateKeys = ["bigcode/starcoder", "codellama/CodeLlama-13b-hf", "Phind/Phind-CodeLlama-34B-v2", "WizardLM/WizardCoder-Python-34B-V1.0", "ollama/CodeLlama-7b-code", "openai/chat-gpt-3.5-turbo", "Custom"] as const;
 export type TemplateKey = typeof templateKeys[number];
 
 export interface TokenizerPathConfig {
@@ -24,6 +24,8 @@ export interface Config {
     contextWindow: number;
     tokensToClear: string[];
     tokenizer: TokenizerPathConfig | TokenizerRepoConfig | TokenizerUrlConfig | null;
+    adaptor?: string;
+    requestBody?: object & { model?: string };
 }
 
 const StarCoderConfig: Config = {
@@ -67,9 +69,41 @@ const WizardCoderPython34Bv1Config: Config = {
     }
 }
 
+const OllamaAdaptorDefaultConfig: Config = {
+    modelIdOrEndpoint: "http://localhost:11435/api/generate",
+    "fillInTheMiddle.enabled": true,
+    "fillInTheMiddle.prefix": "<PRE>",
+    "fillInTheMiddle.middle": "<MID>",
+    "fillInTheMiddle.suffix": "<SUF>",
+    temperature: 0.2,
+    contextWindow: 8192,
+    tokensToClear: ["<EOT>"],
+    tokenizer: {
+        repository: "codellama/CodeLlama-7b-hf",
+    },
+    adaptor: "ollama",
+    requestBody: { model: "codellama:7b-code" }
+}
+
+const OpenAIAdaptorDefaultConfig: Config = {
+    modelIdOrEndpoint: "https://api.openai.com/v1/chat/completions",
+    "fillInTheMiddle.enabled": true,
+    "fillInTheMiddle.prefix": "<im_start>",
+    "fillInTheMiddle.middle": "<im_end>",
+    "fillInTheMiddle.suffix": "<im_end>",
+    temperature: 0.2,
+    contextWindow: 8192,
+    tokensToClear: ["<endoftext>"],
+    tokenizer: null,
+    adaptor: "openai",
+    requestBody: { model: "gpt-3.5-turbo" }
+}
+
 export const templates: Partial<Record<TemplateKey, Config>> = {
     "bigcode/starcoder": StarCoderConfig,
     "codellama/CodeLlama-13b-hf": CodeLlama13BConfig,
     "Phind/Phind-CodeLlama-34B-v2": PhindCodeLlama34Bv2Config,
     "WizardLM/WizardCoder-Python-34B-V1.0": WizardCoderPython34Bv1Config,
-}
+    "ollama/CodeLlama-7b-code": OllamaAdaptorDefaultConfig,
+    "openai/chat-gpt-3.5-turbo": OpenAIAdaptorDefaultConfig,
+}

src/extension.ts

@@ -157,6 +157,8 @@ export function activate(context: vscode.ExtensionContext) {
 				fim: config.get("fillInTheMiddle") as number,
 				context_window: config.get("contextWindow") as number,
 				tls_skip_verify_insecure: config.get("tlsSkipVerifyInsecure") as boolean,
+				adaptor: config.get("adaptor") as string,
+				request_body: config.get("requestBody") as object | null,
 				ide: "vscode",
 				tokenizer_config: config.get("tokenizer") as object | null,
 			};
@@ -329,4 +331,4 @@ async function delay(milliseconds: number, token: vscode.CancellationToken): Pro
             resolve(token.isCancellationRequested)
         }, milliseconds);
     });
-}
+}