refactor: update GITHUB_MODELS_BASE URL and enhance vector index logging

Author: pelikhan

packages/core/src/constants.ts

@@ -221,7 +221,7 @@ export const OPENROUTER_API_CHAT_URL = "https://openrouter.ai/api/v1/chat/comple
 export const OPENROUTER_SITE_URL_HEADER = "HTTP-Referer";
 export const OPENROUTER_SITE_NAME_HEADER = "X-Title";
 
-export const GITHUB_MODELS_BASE = "https://models.github.ai/inference/chat/completions";
+export const GITHUB_MODELS_BASE = "https://models.github.ai/inference";
 
 export const DOCS_CONFIGURATION_URL =
   "https://microsoft.github.io/genaiscript/getting-started/configuration/";

packages/core/src/llmsdata.ts

@@ -292,6 +292,7 @@ export default {
         vision: "openai/gpt-4.1",
         reasoning: "openai/o3",
         reasoning_small: "openai/o3-mini",
+        embeddings: "openai/text-embedding-3-small",
       },
       models: {
         "o1-preview": {

packages/core/src/openai.ts

@@ -6,9 +6,14 @@ import { host } from "./host.js";
 import {
   AZURE_AI_INFERENCE_VERSION,
   AZURE_OPENAI_API_VERSION,
+  MODEL_PROVIDER_ALIBABA,
+  MODEL_PROVIDER_AZURE_AI_INFERENCE,
   MODEL_PROVIDER_AZURE_OPENAI,
   MODEL_PROVIDER_AZURE_SERVERLESS_MODELS,
   MODEL_PROVIDER_AZURE_SERVERLESS_OPENAI,
+  MODEL_PROVIDER_GITHUB,
+  MODEL_PROVIDER_HUGGINGFACE,
+  MODEL_PROVIDER_OPENAI,
   MODEL_PROVIDER_OPENAI_HOSTS,
   OPENROUTER_API_CHAT_URL,
   OPENROUTER_SITE_NAME_HEADER,
@@ -195,24 +200,28 @@ export const OpenAIChatCompletion: ChatCompletionHandler = async (req, cfg, opti
   let url = "";
   const toolCalls: ChatCompletionToolCall[] = [];
 
-  if (cfg.type === "openai" || cfg.type === "localai" || cfg.type === "alibaba") {
+  if (
+    cfg.type === MODEL_PROVIDER_OPENAI ||
+    cfg.type === "localai" ||
+    cfg.type === MODEL_PROVIDER_ALIBABA
+  ) {
     url = trimTrailingSlash(cfg.base) + "/chat/completions";
     if (url === OPENROUTER_API_CHAT_URL) {
       (headers as any)[OPENROUTER_SITE_URL_HEADER] = process.env.OPENROUTER_SITE_URL || TOOL_URL;
       (headers as any)[OPENROUTER_SITE_NAME_HEADER] = process.env.OPENROUTER_SITE_NAME || TOOL_NAME;
     }
-  } else if (cfg.type === "azure") {
+  } else if (cfg.type === MODEL_PROVIDER_AZURE_OPENAI) {
     delete postReq.model;
     const version = cfg.version || AZURE_OPENAI_API_VERSION;
     trace?.itemValue(`version`, version);
     url = trimTrailingSlash(cfg.base) + "/" + family + `/chat/completions?api-version=${version}`;
-  } else if (cfg.type === "azure_ai_inference") {
+  } else if (cfg.type === MODEL_PROVIDER_AZURE_AI_INFERENCE) {
     const version = cfg.version;
     trace?.itemValue(`version`, version);
     url = trimTrailingSlash(cfg.base) + `/chat/completions`;
     if (version) url += `?api-version=${version}`;
     (headers as any)["extra-parameters"] = "pass-through";
-  } else if (cfg.type === "azure_serverless_models") {
+  } else if (cfg.type === MODEL_PROVIDER_AZURE_SERVERLESS_MODELS) {
     const version = cfg.version || AZURE_AI_INFERENCE_VERSION;
     trace?.itemValue(`version`, version);
     url =
@@ -223,15 +232,15 @@ export const OpenAIChatCompletion: ChatCompletionHandler = async (req, cfg, opti
     (headers as any)["extra-parameters"] = "pass-through";
     delete postReq.model;
     delete postReq.stream_options;
-  } else if (cfg.type === "azure_serverless") {
+  } else if (cfg.type === MODEL_PROVIDER_AZURE_SERVERLESS_OPENAI) {
     const version = cfg.version || AZURE_AI_INFERENCE_VERSION;
     trace?.itemValue(`version`, version);
     url = trimTrailingSlash(cfg.base) + "/" + family + `/chat/completions?api-version=${version}`;
     // https://learn.microsoft.com/en-us/azure/machine-learning/reference-model-inference-api?view=azureml-api-2&tabs=javascript#extensibility
     (headers as any)["extra-parameters"] = "pass-through";
     delete postReq.model;
-  } else if (cfg.type === "github") {
-    url = cfg.base;
+  } else if (cfg.type === MODEL_PROVIDER_GITHUB) {
+    url = trimTrailingSlash(cfg.base) + "/chat/completions";
     const { prefix } = /^(?<prefix>[^-]+)-([^\/]+)$/.exec(postReq.model)?.groups || {};
     const patch = {
       gpt: "openai",
@@ -247,7 +256,7 @@ export const OpenAIChatCompletion: ChatCompletionHandler = async (req, cfg, opti
       postReq.model = `${patch}/${postReq.model}`;
       dbg(`updated model to ${postReq.model}`);
     }
-  } else if (cfg.type === "huggingface") {
+  } else if (cfg.type === MODEL_PROVIDER_HUGGINGFACE) {
     // https://github.com/huggingface/text-generation-inference/issues/2946
     delete postReq.model;
     url =

packages/core/src/vectorsearch.ts

@@ -26,6 +26,8 @@ import type {
   WorkspaceFileIndex,
   WorkspaceFileWithScore,
 } from "./types.js";
+import { genaiscriptDebug } from "./debug.js";
+const dbg = genaiscriptDebug("vector");
 
 interface EmbeddingsResponse {
   /**
@@ -126,6 +128,7 @@ export async function vectorCreateIndex(
   assert(!!indexName);
   options = options || {};
   const { type = "local", embeddingsModel, cancellationToken, trace } = options || {};
+  dbg(`create index %s %s %s`, indexName, type, embeddingsModel);
 
   let factory: WorkspaceFileIndexCreator;
   if (type === "azure_ai_search") factory = azureAISearchIndex;
@@ -154,13 +157,18 @@ export async function vectorCreateIndex(
   checkCancelled(cancellationToken);
 
   if (!options.vectorSize) {
+    dbg(`sniffing vector size for %s`, indexName);
     const sniff = await cachedEmbedder(
+      
       `Lorem ipsum dolor sit amet, consectetur adipiscing elit
 sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.`,
       configuration,
       options,
     );
-    options.vectorSize = sniff.data[0].length;
+    const vectorSize = sniff.data?.[0]?.length;
+    dbg(`sniffed vector size: %o`, sniff.data);
+    if (isNaN(vectorSize)) throw new Error("embeddings: unable to determine vector size");
+      options.vectorSize = vectorSize;
   }
 
   return await factory(indexName, configuration, cachedEmbedder, options);