@retrivora-ai/rag-engine 2.5.5 → 2.5.6

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/server.mjs CHANGED
@@ -326,7 +326,7 @@ var init_package = __esm({
326
326
  "package.json"() {
327
327
  package_default = {
328
328
  name: "@retrivora-ai/rag-engine",
329
- version: "2.5.5",
329
+ version: "2.5.6",
330
330
  description: "Retrivora AI is a plug-and-play AI engine for RAG chat experiences \u2014 generic vector DB + LLM provider, embeddable or standalone.",
331
331
  author: "Abhinav Alkuchi",
332
332
  license: "UNLICENSED",
@@ -1770,7 +1770,7 @@ var MilvusProvider_exports = {};
1770
1770
  __export(MilvusProvider_exports, {
1771
1771
  MilvusProvider: () => MilvusProvider
1772
1772
  });
1773
- import axios3 from "axios";
1773
+ import axios4 from "axios";
1774
1774
  var MilvusProvider;
1775
1775
  var init_MilvusProvider = __esm({
1776
1776
  "src/providers/vectordb/MilvusProvider.ts"() {
@@ -1782,7 +1782,7 @@ var init_MilvusProvider = __esm({
1782
1782
  const opts = config.options;
1783
1783
  const baseUrl = opts.baseUrl || opts.uri || (opts.host ? `http://${opts.host}:${opts.port || 19530}` : void 0);
1784
1784
  if (!baseUrl) throw new Error("[MilvusProvider] options.baseUrl (or uri / host+port) is required");
1785
- this.http = axios3.create({
1785
+ this.http = axios4.create({
1786
1786
  baseURL: baseUrl,
1787
1787
  headers: __spreadValues(__spreadValues({
1788
1788
  "Content-Type": "application/json"
@@ -1883,7 +1883,7 @@ var QdrantProvider_exports = {};
1883
1883
  __export(QdrantProvider_exports, {
1884
1884
  QdrantProvider: () => QdrantProvider
1885
1885
  });
1886
- import axios4 from "axios";
1886
+ import axios5 from "axios";
1887
1887
  import crypto from "crypto";
1888
1888
  var QdrantProvider;
1889
1889
  var init_QdrantProvider = __esm({
@@ -1900,7 +1900,7 @@ var init_QdrantProvider = __esm({
1900
1900
  this.contentField = opts.contentField || "content";
1901
1901
  this.metadataField = opts.metadataField || "metadata";
1902
1902
  this.isFlatPayload = !!opts.flatPayload;
1903
- this.http = axios4.create({
1903
+ this.http = axios5.create({
1904
1904
  baseURL: baseUrl,
1905
1905
  timeout: 15e3,
1906
1906
  // 15s timeout for vector DB operations
@@ -1968,7 +1968,7 @@ var init_QdrantProvider = __esm({
1968
1968
  try {
1969
1969
  await this.http.get(`/collections/${this.indexName}`);
1970
1970
  } catch (err) {
1971
- if (axios4.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
1971
+ if (axios5.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
1972
1972
  const opts = this.config.options;
1973
1973
  const dimensionsForCreate = opts.dimensions || 1536;
1974
1974
  await this.http.put(`/collections/${this.indexName}`, {
@@ -1998,7 +1998,7 @@ var init_QdrantProvider = __esm({
1998
1998
  });
1999
1999
  } catch (err) {
2000
2000
  let status;
2001
- if (axios4.isAxiosError(err)) status = (_a2 = err.response) == null ? void 0 : _a2.status;
2001
+ if (axios5.isAxiosError(err)) status = (_a2 = err.response) == null ? void 0 : _a2.status;
2002
2002
  if (status === 409) return;
2003
2003
  console.warn(`[QdrantProvider] \u26A0\uFE0F Could not ensure index for "${fullPath}":`, err instanceof Error ? err.message : String(err));
2004
2004
  }
@@ -2122,7 +2122,7 @@ var ChromaDBProvider_exports = {};
2122
2122
  __export(ChromaDBProvider_exports, {
2123
2123
  ChromaDBProvider: () => ChromaDBProvider
2124
2124
  });
2125
- import axios5 from "axios";
2125
+ import axios6 from "axios";
2126
2126
  var ChromaDBProvider;
2127
2127
  var init_ChromaDBProvider = __esm({
2128
2128
  "src/providers/vectordb/ChromaDBProvider.ts"() {
@@ -2135,7 +2135,7 @@ var init_ChromaDBProvider = __esm({
2135
2135
  const opts = config.options;
2136
2136
  const baseUrl = opts.baseUrl || (opts.host ? `http://${opts.host}:${opts.port || 8e3}` : void 0);
2137
2137
  if (!baseUrl) throw new Error("[ChromaDBProvider] options.baseUrl is required");
2138
- this.http = axios5.create({
2138
+ this.http = axios6.create({
2139
2139
  baseURL: baseUrl,
2140
2140
  headers: { "Content-Type": "application/json" }
2141
2141
  });
@@ -2149,7 +2149,7 @@ var init_ChromaDBProvider = __esm({
2149
2149
  const { data } = await this.http.get(`/api/v1/collections/${this.indexName}`);
2150
2150
  this.collectionId = data.id;
2151
2151
  } catch (err) {
2152
- if (axios5.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
2152
+ if (axios6.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
2153
2153
  const { data } = await this.http.post("/api/v1/collections", {
2154
2154
  name: this.indexName
2155
2155
  });
@@ -2230,7 +2230,7 @@ var RedisProvider_exports = {};
2230
2230
  __export(RedisProvider_exports, {
2231
2231
  RedisProvider: () => RedisProvider
2232
2232
  });
2233
- import axios6 from "axios";
2233
+ import axios7 from "axios";
2234
2234
  var RedisProvider;
2235
2235
  var init_RedisProvider = __esm({
2236
2236
  "src/providers/vectordb/RedisProvider.ts"() {
@@ -2242,7 +2242,7 @@ var init_RedisProvider = __esm({
2242
2242
  const opts = config.options;
2243
2243
  const baseUrl = opts.baseUrl || opts.url || (opts.host ? `http://${opts.host}:${opts.port || 6379}` : void 0);
2244
2244
  if (!baseUrl) throw new Error("[RedisProvider] options.baseUrl is required");
2245
- this.http = axios6.create({
2245
+ this.http = axios7.create({
2246
2246
  baseURL: baseUrl,
2247
2247
  headers: __spreadValues({
2248
2248
  "Content-Type": "application/json"
@@ -2312,7 +2312,7 @@ var WeaviateProvider_exports = {};
2312
2312
  __export(WeaviateProvider_exports, {
2313
2313
  WeaviateProvider: () => WeaviateProvider
2314
2314
  });
2315
- import axios7 from "axios";
2315
+ import axios8 from "axios";
2316
2316
  var WeaviateProvider;
2317
2317
  var init_WeaviateProvider = __esm({
2318
2318
  "src/providers/vectordb/WeaviateProvider.ts"() {
@@ -2324,7 +2324,7 @@ var init_WeaviateProvider = __esm({
2324
2324
  const opts = config.options;
2325
2325
  const baseUrl = opts.baseUrl || opts.url;
2326
2326
  if (!baseUrl) throw new Error("[WeaviateProvider] options.baseUrl is required");
2327
- this.http = axios7.create({
2327
+ this.http = axios8.create({
2328
2328
  baseURL: baseUrl,
2329
2329
  headers: __spreadValues(__spreadValues({
2330
2330
  "Content-Type": "application/json"
@@ -2464,7 +2464,7 @@ var UniversalVectorProvider_exports = {};
2464
2464
  __export(UniversalVectorProvider_exports, {
2465
2465
  UniversalVectorProvider: () => UniversalVectorProvider
2466
2466
  });
2467
- import axios8 from "axios";
2467
+ import axios9 from "axios";
2468
2468
  var UniversalVectorProvider;
2469
2469
  var init_UniversalVectorProvider = __esm({
2470
2470
  "src/providers/vectordb/UniversalVectorProvider.ts"() {
@@ -2481,7 +2481,7 @@ var init_UniversalVectorProvider = __esm({
2481
2481
  }
2482
2482
  async initialize() {
2483
2483
  var _a2;
2484
- this.http = axios8.create({
2484
+ this.http = axios9.create({
2485
2485
  baseURL: this.opts.baseUrl,
2486
2486
  headers: __spreadValues({
2487
2487
  "Content-Type": "application/json"
@@ -3061,7 +3061,9 @@ var LLM_PROVIDERS = [
3061
3061
  "qwen",
3062
3062
  "rest",
3063
3063
  "universal_rest",
3064
- "custom"
3064
+ "custom",
3065
+ "custom_embedding",
3066
+ "local_embedding"
3065
3067
  ];
3066
3068
  var EMBEDDING_PROVIDERS = [
3067
3069
  "openai",
@@ -3070,7 +3072,9 @@ var EMBEDDING_PROVIDERS = [
3070
3072
  "qwen",
3071
3073
  "rest",
3072
3074
  "universal_rest",
3073
- "custom"
3075
+ "custom",
3076
+ "custom_embedding",
3077
+ "local_embedding"
3074
3078
  ];
3075
3079
  var PROVIDERS_WITH_EMBEDDINGS = [
3076
3080
  "openai",
@@ -3078,7 +3082,10 @@ var PROVIDERS_WITH_EMBEDDINGS = [
3078
3082
  "gemini",
3079
3083
  "qwen",
3080
3084
  "rest",
3081
- "universal_rest"
3085
+ "universal_rest",
3086
+ "custom",
3087
+ "custom_embedding",
3088
+ "local_embedding"
3082
3089
  ];
3083
3090
 
3084
3091
  // src/config/serverConfig.ts
@@ -3824,8 +3831,8 @@ var OllamaProvider = class {
3824
3831
  const baseUrl = config.baseUrl || "http://localhost:11434";
3825
3832
  const modelName = config.model;
3826
3833
  try {
3827
- const axios10 = (await import("axios")).default;
3828
- const { data } = await axios10.get(`${baseUrl}/api/tags`);
3834
+ const axios11 = (await import("axios")).default;
3835
+ const { data } = await axios11.get(`${baseUrl}/api/tags`);
3829
3836
  const models = data.models || [];
3830
3837
  const hasModel = models.some((m) => m.name === modelName || m.name.startsWith(`${modelName}:`));
3831
3838
  return {
@@ -4942,6 +4949,158 @@ ${context != null ? context : "None"}` },
4942
4949
  }
4943
4950
  };
4944
4951
 
4952
+ // src/llm/providers/CustomEmbeddingProvider.ts
4953
+ init_templateUtils();
4954
+ import axios3 from "axios";
4955
+ var CustomEmbeddingProvider = class {
4956
+ constructor(config) {
4957
+ var _a2, _b;
4958
+ this.model = config.model || "custom-embedding-model";
4959
+ this.apiKey = config.apiKey;
4960
+ const options = (_a2 = config.options) != null ? _a2 : {};
4961
+ this.opts = {
4962
+ baseUrl: config.baseUrl || options.baseUrl,
4963
+ embedPath: options.embedPath || "/embeddings",
4964
+ pingPath: options.pingPath || "/health",
4965
+ queryPrefix: options.queryPrefix || "",
4966
+ documentPrefix: options.documentPrefix || "",
4967
+ embedPayloadTemplate: options.embedPayloadTemplate,
4968
+ embedExtractPath: options.embedExtractPath,
4969
+ headers: options.headers || {},
4970
+ timeout: options.timeout || 3e4
4971
+ };
4972
+ this.baseUrl = (_b = this.opts.baseUrl) != null ? _b : "";
4973
+ if (!this.baseUrl) {
4974
+ throw new Error("[CustomEmbeddingProvider] baseUrl is required in config or config.options");
4975
+ }
4976
+ const headers = __spreadValues(__spreadValues({
4977
+ "Content-Type": "application/json"
4978
+ }, this.apiKey ? { Authorization: `Bearer ${this.apiKey}` } : {}), this.opts.headers || {});
4979
+ this.http = axios3.create({
4980
+ baseURL: this.baseUrl.replace(/\/$/, ""),
4981
+ headers,
4982
+ timeout: this.opts.timeout
4983
+ });
4984
+ }
4985
+ /**
4986
+ * Dummy chat implementation. CustomEmbeddingProvider is dedicated to embedding generation.
4987
+ */
4988
+ async chat(_messages, _context, _options) {
4989
+ throw new Error("[CustomEmbeddingProvider] chat() is not supported on a dedicated embedding provider.");
4990
+ }
4991
+ /**
4992
+ * Generate vector embedding for a single string.
4993
+ */
4994
+ async embed(text, options) {
4995
+ var _a2;
4996
+ const taskType = (_a2 = options == null ? void 0 : options.taskType) != null ? _a2 : "document";
4997
+ const prefix = taskType === "query" ? this.opts.queryPrefix : this.opts.documentPrefix;
4998
+ const formattedText = prefix ? `${prefix}${text}` : text;
4999
+ const modelToUse = (options == null ? void 0 : options.model) || this.model;
5000
+ const path2 = this.opts.embedPath || "/embeddings";
5001
+ let payload;
5002
+ if (this.opts.embedPayloadTemplate) {
5003
+ payload = buildPayload(this.opts.embedPayloadTemplate, {
5004
+ input: formattedText,
5005
+ model: modelToUse
5006
+ });
5007
+ } else {
5008
+ payload = {
5009
+ model: modelToUse,
5010
+ input: formattedText
5011
+ };
5012
+ }
5013
+ try {
5014
+ const { data } = await this.http.post(path2, payload);
5015
+ const vector = this.extractVector(data);
5016
+ if (vector) return vector;
5017
+ throw new Error(`[CustomEmbeddingProvider] Failed to extract embedding vector from response at ${path2}`);
5018
+ } catch (err) {
5019
+ console.error("[CustomEmbeddingProvider] Embedding request failed:", {
5020
+ baseUrl: this.baseUrl,
5021
+ path: path2,
5022
+ model: modelToUse,
5023
+ error: (err == null ? void 0 : err.message) || err
5024
+ });
5025
+ throw err;
5026
+ }
5027
+ }
5028
+ /**
5029
+ * Batch embedding generation for multiple texts.
5030
+ */
5031
+ async batchEmbed(texts, options) {
5032
+ var _a2;
5033
+ if (!Array.isArray(texts) || texts.length === 0) return [];
5034
+ const taskType = (_a2 = options == null ? void 0 : options.taskType) != null ? _a2 : "document";
5035
+ const prefix = taskType === "query" ? this.opts.queryPrefix : this.opts.documentPrefix;
5036
+ const formattedTexts = prefix ? texts.map((t) => `${prefix}${t}`) : texts;
5037
+ const modelToUse = (options == null ? void 0 : options.model) || this.model;
5038
+ const path2 = this.opts.embedPath || "/embeddings";
5039
+ if (!this.opts.embedPayloadTemplate) {
5040
+ try {
5041
+ const payload = {
5042
+ model: modelToUse,
5043
+ input: formattedTexts
5044
+ };
5045
+ const { data } = await this.http.post(path2, payload);
5046
+ const batchVectors = this.extractBatchVectors(data);
5047
+ if (batchVectors && batchVectors.length === texts.length) {
5048
+ return batchVectors;
5049
+ }
5050
+ } catch (e) {
5051
+ }
5052
+ }
5053
+ return Promise.all(texts.map((text) => this.embed(text, options)));
5054
+ }
5055
+ /**
5056
+ * Ping provider endpoint for health check.
5057
+ */
5058
+ async ping() {
5059
+ try {
5060
+ const pingPath = this.opts.pingPath || "/health";
5061
+ const res = await this.http.get(pingPath);
5062
+ return res.status >= 200 && res.status < 300;
5063
+ } catch (e) {
5064
+ try {
5065
+ const embedPath = this.opts.embedPath || "/embeddings";
5066
+ await this.http.post(embedPath, { input: "ping", model: this.model });
5067
+ return true;
5068
+ } catch (e2) {
5069
+ return false;
5070
+ }
5071
+ }
5072
+ }
5073
+ extractVector(data) {
5074
+ if (!data) return void 0;
5075
+ if (this.opts.embedExtractPath) {
5076
+ const extracted = resolvePath(data, this.opts.embedExtractPath);
5077
+ if (Array.isArray(extracted) && typeof extracted[0] === "number") {
5078
+ return extracted;
5079
+ }
5080
+ }
5081
+ const directVector = resolvePath(data, "data[0].embedding") || resolvePath(data, "embedding") || resolvePath(data, "embeddings[0]") || resolvePath(data, "data[0].vector") || resolvePath(data, "vector");
5082
+ if (Array.isArray(directVector) && typeof directVector[0] === "number") {
5083
+ return directVector;
5084
+ }
5085
+ if (Array.isArray(data) && typeof data[0] === "number") {
5086
+ return data;
5087
+ }
5088
+ return void 0;
5089
+ }
5090
+ extractBatchVectors(data) {
5091
+ if (!data) return void 0;
5092
+ if (typeof data === "object" && data !== null && Array.isArray(data.data)) {
5093
+ const list = data.data;
5094
+ const vectors = list.map((item) => (item == null ? void 0 : item.embedding) || (item == null ? void 0 : item.vector)).filter((v) => Array.isArray(v));
5095
+ if (vectors.length > 0) return vectors;
5096
+ }
5097
+ if (Array.isArray(data) && Array.isArray(data[0]) && typeof data[0][0] === "number") {
5098
+ return data;
5099
+ }
5100
+ return void 0;
5101
+ }
5102
+ };
5103
+
4945
5104
  // src/llm/LLMFactory.ts
4946
5105
  var customProviders = /* @__PURE__ */ new Map();
4947
5106
  var LLMFactory = class _LLMFactory {
@@ -4984,6 +5143,8 @@ var LLMFactory = class _LLMFactory {
4984
5143
  "rest",
4985
5144
  "universal_rest",
4986
5145
  "custom",
5146
+ "custom_embedding",
5147
+ "local_embedding",
4987
5148
  ...Array.from(customProviders.keys())
4988
5149
  ];
4989
5150
  }
@@ -5002,6 +5163,9 @@ var LLMFactory = class _LLMFactory {
5002
5163
  return new GroqProvider(llmConfig, embeddingConfig);
5003
5164
  case "qwen":
5004
5165
  return new QwenProvider(llmConfig, embeddingConfig);
5166
+ case "custom_embedding":
5167
+ case "local_embedding":
5168
+ return new CustomEmbeddingProvider(llmConfig);
5005
5169
  case "rest":
5006
5170
  case "universal_rest":
5007
5171
  case "custom":
@@ -12478,7 +12642,7 @@ _LicenseValidator.SUCCESS_CACHE_TTL_MS = 5 * 60 * 1e3;
12478
12642
  var LicenseValidator = _LicenseValidator;
12479
12643
 
12480
12644
  // src/config/LicenseConfigResolver.ts
12481
- import axios9 from "axios";
12645
+ import axios10 from "axios";
12482
12646
  var LicenseConfigResolver = class {
12483
12647
  /**
12484
12648
  * Validate and resolve configuration from license key
@@ -12502,7 +12666,7 @@ var LicenseConfigResolver = class {
12502
12666
  */
12503
12667
  static async fetchTierConfig(licenseKey, projectId) {
12504
12668
  try {
12505
- const response = await axios9.post(
12669
+ const response = await axios10.post(
12506
12670
  this.RETRIVORA_CONFIG_URL,
12507
12671
  { licenseKey, projectId },
12508
12672
  {
@@ -12533,7 +12697,7 @@ var LicenseConfigResolver = class {
12533
12697
  */
12534
12698
  static async fetchProviderCredentials(licenseKey, projectId) {
12535
12699
  try {
12536
- const response = await axios9.post(
12700
+ const response = await axios10.post(
12537
12701
  this.RETRIVORA_CREDENTIALS_URL,
12538
12702
  { licenseKey, projectId },
12539
12703
  {
@@ -12731,12 +12895,12 @@ var _TelemetryService = class _TelemetryService {
12731
12895
  });
12732
12896
  }
12733
12897
  /**
12734
- * Record query execution event
12898
+ * Record query execution event (PII-safe)
12735
12899
  */
12736
12900
  recordQueryExecuted(data) {
12737
12901
  this.recordEvent({
12738
12902
  eventType: "query_executed" /* QUERY_EXECUTED */,
12739
- queryText: data.queryText,
12903
+ queryLength: data.queryLength,
12740
12904
  resultCount: data.resultCount,
12741
12905
  durationMs: data.durationMs,
12742
12906
  operationId: data.operationId
@@ -12982,7 +13146,7 @@ var LicensedRetrivora = class {
12982
13146
  const durationMs = Date.now() - startTime;
12983
13147
  if (telemetry) {
12984
13148
  telemetry.recordQueryExecuted({
12985
- queryText,
13149
+ queryLength: queryText.length,
12986
13150
  resultCount: ((_a2 = response.sources) == null ? void 0 : _a2.length) || 0,
12987
13151
  durationMs,
12988
13152
  operationId
@@ -13063,6 +13227,7 @@ export {
13063
13227
  ConfigResolver,
13064
13228
  ConfigValidator,
13065
13229
  ConfigurationException,
13230
+ CustomEmbeddingProvider,
13066
13231
  DocumentChunker,
13067
13232
  DocumentParser,
13068
13233
  EmbeddingFailedException,
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@retrivora-ai/rag-engine",
3
- "version": "2.5.5",
3
+ "version": "2.5.6",
4
4
  "description": "Retrivora AI is a plug-and-play AI engine for RAG chat experiences — generic vector DB + LLM provider, embeddable or standalone.",
5
5
  "author": "Abhinav Alkuchi",
6
6
  "license": "UNLICENSED",