@retrivora-ai/rag-engine 0.1.3 → 0.1.4

package/dist/chunk-GD3QJFNN.mjs

@@ -0,0 +1,424 @@
+import {
+  __spreadProps,
+  __spreadValues
+} from "./chunk-I4E63NIC.mjs";
+
+// src/llm/providers/OpenAIProvider.ts
+import OpenAI from "openai";
+var OpenAIProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error("[OpenAIProvider] llmConfig.apiKey is required");
+    this.client = new OpenAI({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h;
+    const systemContent = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+    const systemMessage = {
+      role: "system",
+      content: systemContent.includes("{{context}}") ? systemContent.replace("{{context}}", context) : `${systemContent}
+
+Context:
+${context}`
+    };
+    const formattedMessages = [
+      systemMessage,
+      ...messages.map((m) => ({
+        role: m.role,
+        content: m.content
+      }))
+    ];
+    const completion = await this.client.chat.completions.create({
+      model: this.llmConfig.model,
+      messages: formattedMessages,
+      max_tokens: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+      temperature: (_e = (_d = options == null ? void 0 : options.temperature) != null ? _d : this.llmConfig.temperature) != null ? _e : 0.7,
+      stop: options == null ? void 0 : options.stop
+    });
+    return (_h = (_g = (_f = completion.choices[0]) == null ? void 0 : _f.message) == null ? void 0 : _g.content) != null ? _h : "";
+  }
+  async embed(text, options) {
+    const results = await this.batchEmbed([text], options);
+    return results[0];
+  }
+  async batchEmbed(texts, options) {
+    var _a, _b, _c, _d, _e;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-3-small";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new OpenAI({ apiKey }) : this.client;
+    const response = await client.embeddings.create({ model, input: texts });
+    return response.data.map((d) => d.embedding);
+  }
+  async ping() {
+    try {
+      await this.client.models.list();
+      return true;
+    } catch (err) {
+      console.error("[OpenAIProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
+// src/llm/providers/AnthropicProvider.ts
+import Anthropic from "@anthropic-ai/sdk";
+var AnthropicProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error("[AnthropicProvider] llmConfig.apiKey is required");
+    this.client = new Anthropic({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c;
+    const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Use the context below to answer the user's question accurately.
+
+Context:
+${context}`;
+    const system = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+    const anthropicMessages = messages.map((m) => ({
+      role: m.role === "assistant" ? "assistant" : "user",
+      content: m.content
+    }));
+    const response = await this.client.messages.create({
+      model: this.llmConfig.model,
+      max_tokens: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+      system,
+      messages: anthropicMessages
+    });
+    const block = response.content[0];
+    return block.type === "text" ? block.text : "";
+  }
+  /**
+   * Anthropic does not offer an embedding API.
+   * This method throws with a clear error so developers know to configure
+   * a separate embedding provider (OpenAI or Ollama).
+   */
+  async embed(text, options) {
+    void text;
+    void options;
+    throw new Error(
+      '[AnthropicProvider] Anthropic does not provide an embedding API. Set embedding.provider to "openai" or "ollama" in your RagConfig.'
+    );
+  }
+  async batchEmbed(texts, options) {
+    void texts;
+    void options;
+    throw new Error(
+      "[AnthropicProvider] Anthropic does not provide an embedding API."
+    );
+  }
+  async ping() {
+    try {
+      await this.client.models.list();
+      return true;
+    } catch (err) {
+      console.error("[AnthropicProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
+// src/llm/providers/OllamaProvider.ts
+import axios from "axios";
+var OllamaProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    var _a;
+    const baseURL = (_a = llmConfig.baseUrl) != null ? _a : "http://localhost:11434";
+    this.http = axios.create({ baseURL, timeout: 12e4 });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c, _d, _e;
+    const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Use the provided context to answer the user's question.
+
+Context:
+${context}`;
+    const system = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+    const { data } = await this.http.post("/api/chat", {
+      model: this.llmConfig.model,
+      stream: false,
+      options: {
+        temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+        num_predict: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024
+      },
+      messages: [
+        { role: "system", content: system },
+        ...messages.map((m) => ({ role: m.role, content: m.content }))
+      ]
+    });
+    return data.message.content;
+  }
+  async embed(text, options) {
+    var _a, _b, _c, _d, _e, _f, _g;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "nomic-embed-text";
+    const baseURL = (_f = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.baseUrl) != null ? _e : this.llmConfig.baseUrl) != null ? _f : "http://localhost:11434";
+    const client = baseURL !== ((_g = this.llmConfig.baseUrl) != null ? _g : "http://localhost:11434") ? axios.create({ baseURL, timeout: 6e4 }) : this.http;
+    const { data } = await client.post("/api/embeddings", {
+      model,
+      prompt: text
+    });
+    return data.embedding;
+  }
+  async batchEmbed(texts, options) {
+    const vectors = [];
+    for (const text of texts) {
+      vectors.push(await this.embed(text, options));
+    }
+    return vectors;
+  }
+  async ping() {
+    try {
+      await this.http.get("/api/tags");
+      return true;
+    } catch (err) {
+      console.error("[OllamaProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
+// src/llm/providers/UniversalLLMAdapter.ts
+import axios2 from "axios";
+
+// src/utils/templateUtils.ts
+function isRecord(value) {
+  return typeof value === "object" && value !== null;
+}
+function resolvePath(obj, path) {
+  if (!path || !obj) return void 0;
+  return path.replace(/\[(\w+)\]/g, ".$1").replace(/^\./, "").split(".").reduce((current, segment) => {
+    if (!isRecord(current) && !Array.isArray(current)) {
+      return void 0;
+    }
+    return current[segment];
+  }, obj);
+}
+function buildPayload(template, vars) {
+  let raw = template;
+  for (const [key, val] of Object.entries(vars)) {
+    const stringifiedVal = val === void 0 ? "null" : JSON.stringify(val);
+    raw = raw.replace(new RegExp(`"\\{\\{${key}\\}\\}"`, "g"), stringifiedVal);
+    raw = raw.replace(new RegExp(`\\{\\{${key}\\}\\}`, "g"), stringifiedVal);
+  }
+  try {
+    return JSON.parse(raw);
+  } catch (err) {
+    throw new Error(`Failed to parse payload template: ${err.message}
+Raw payload: ${raw}`);
+  }
+}
+function mergeDefined(base, override) {
+  const merged = __spreadValues({}, base || {});
+  if (!override) {
+    return merged;
+  }
+  for (const [key, value] of Object.entries(override)) {
+    if (value !== void 0 && value !== null && value !== "") {
+      merged[key] = value;
+    }
+  }
+  return merged;
+}
+
+// src/config/UniversalProfiles.ts
+var OPENAI_BASE = {
+  chatPath: "/chat/completions",
+  embedPath: "/embeddings",
+  responseExtractPath: "choices[0].message.content",
+  embedExtractPath: "data[0].embedding",
+  chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"max_tokens":{{maxTokens}},"temperature":{{temperature}}}'
+};
+var LLM_PROFILES = {
+  "openai-compatible": OPENAI_BASE,
+  "litellm": __spreadProps(__spreadValues({}, OPENAI_BASE), {
+    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"max_tokens":{{maxTokens}},"temperature":{{temperature}},"stream":false}'
+  }),
+  "anthropic-claude": {
+    chatPath: "/v1/messages",
+    responseExtractPath: "content[0].text",
+    headers: { "anthropic-version": "2023-06-01" },
+    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"max_tokens":{{maxTokens}}}'
+  },
+  "google-gemini": {
+    chatPath: "/v1beta/models/{{model}}:generateContent",
+    responseExtractPath: "candidates[0].content.parts[0].text",
+    chatPayloadTemplate: '{"contents":[{"parts":[{"text":"{{messages}}"}]}]}'
+  },
+  "github-copilot": __spreadProps(__spreadValues({}, OPENAI_BASE), {
+    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"temperature":{{temperature}}}'
+  }),
+  "ollama-standard": {
+    chatPath: "/api/chat",
+    embedPath: "/api/embeddings",
+    responseExtractPath: "message.content",
+    embedExtractPath: "embedding",
+    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"stream":false}',
+    embedPayloadTemplate: '{"model":"{{model}}","prompt":"{{input}}"}'
+  }
+};
+
+// src/llm/providers/UniversalLLMAdapter.ts
+var UniversalLLMAdapter = class {
+  constructor(config) {
+    var _a, _b, _c, _d, _e, _f, _g;
+    this.model = config.model;
+    const llmConfig = config;
+    const options = (_a = llmConfig.options) != null ? _a : {};
+    let profile = {};
+    if (typeof options.profile === "string") {
+      profile = LLM_PROFILES[options.profile] || {};
+    } else if (typeof options.profile === "object" && options.profile !== null) {
+      profile = options.profile;
+    }
+    this.opts = __spreadValues(__spreadValues({}, profile), (_b = llmConfig.options) != null ? _b : {});
+    this.systemPrompt = (_c = llmConfig.systemPrompt) != null ? _c : "You are a helpful AI assistant. Use the provided context to answer the user.";
+    this.maxTokens = (_d = llmConfig.maxTokens) != null ? _d : 1024;
+    this.temperature = (_e = llmConfig.temperature) != null ? _e : 0;
+    const baseUrl = (_f = llmConfig.baseUrl) != null ? _f : this.opts.baseUrl;
+    if (!baseUrl) {
+      throw new Error("[UniversalLLMAdapter] baseUrl is required in config or config.options");
+    }
+    this.http = axios2.create({
+      baseURL: baseUrl,
+      headers: __spreadValues(__spreadValues({
+        "Content-Type": "application/json"
+      }, config.apiKey ? { Authorization: `Bearer ${config.apiKey}` } : {}), this.opts.headers || {}),
+      timeout: (_g = this.opts.timeout) != null ? _g : 6e4
+    });
+  }
+  async chat(messages, context) {
+    var _a, _b;
+    const path = (_a = this.opts.chatPath) != null ? _a : "/chat/completions";
+    const formattedMessages = [
+      { role: "system", content: `${this.systemPrompt}
+
+Context:
+${context != null ? context : "None"}` },
+      ...messages
+    ];
+    let payload;
+    if (this.opts.chatPayloadTemplate) {
+      payload = buildPayload(this.opts.chatPayloadTemplate, {
+        model: this.model,
+        messages: formattedMessages,
+        maxTokens: this.maxTokens,
+        temperature: this.temperature
+      });
+    } else {
+      payload = {
+        model: this.model,
+        messages: formattedMessages,
+        max_tokens: this.maxTokens,
+        temperature: this.temperature
+      };
+    }
+    const { data } = await this.http.post(path, payload);
+    const extractPath = (_b = this.opts.responseExtractPath) != null ? _b : "choices[0].message.content";
+    const result = resolvePath(data, extractPath);
+    if (result === void 0) {
+      throw new Error(`[UniversalLLMAdapter] Could not extract text from path '${extractPath}' in response.`);
+    }
+    return String(result);
+  }
+  async embed(text) {
+    var _a, _b;
+    const path = (_a = this.opts.embedPath) != null ? _a : "/embeddings";
+    let payload;
+    if (this.opts.embedPayloadTemplate) {
+      payload = buildPayload(this.opts.embedPayloadTemplate, {
+        model: this.model,
+        input: text
+      });
+    } else {
+      payload = {
+        model: this.model,
+        input: text
+      };
+    }
+    const { data } = await this.http.post(path, payload);
+    const extractPath = (_b = this.opts.embedExtractPath) != null ? _b : "data[0].embedding";
+    const vector = resolvePath(data, extractPath);
+    if (!Array.isArray(vector)) {
+      throw new Error(`[UniversalLLMAdapter] Expected a number array at '${extractPath}' for embeddings.`);
+    }
+    return vector;
+  }
+  async batchEmbed(texts) {
+    const vectors = [];
+    for (const text of texts) {
+      vectors.push(await this.embed(text));
+    }
+    return vectors;
+  }
+  async ping() {
+    try {
+      if (this.opts.pingPath) {
+        await this.http.get(this.opts.pingPath);
+      }
+      return true;
+    } catch (err) {
+      console.error("[UniversalLLMAdapter] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
+// src/llm/LLMFactory.ts
+var LLMFactory = class _LLMFactory {
+  static create(llmConfig, embeddingConfig) {
+    var _a;
+    switch (llmConfig.provider) {
+      case "openai":
+        return new OpenAIProvider(llmConfig, embeddingConfig);
+      case "anthropic":
+        return new AnthropicProvider(llmConfig, embeddingConfig);
+      case "ollama":
+        return new OllamaProvider(llmConfig, embeddingConfig);
+      case "rest":
+      case "universal_rest":
+      case "custom":
+        return new UniversalLLMAdapter(llmConfig);
+      default:
+        if (llmConfig.baseUrl || ((_a = llmConfig.options) == null ? void 0 : _a.baseUrl)) {
+          return new UniversalLLMAdapter(llmConfig);
+        }
+        throw new Error(
+          `[LLMFactory] Unknown provider "${llmConfig.provider}". Supported: openai | anthropic | ollama | rest | custom`
+        );
+    }
+  }
+  /**
+   * Creates a dedicated embedding-only provider.
+   * Useful when the LLM provider (e.g. Anthropic) doesn't support embeddings.
+   */
+  static createEmbeddingProvider(embeddingConfig) {
+    const fakeLLMConfig = {
+      provider: embeddingConfig.provider,
+      model: embeddingConfig.model,
+      apiKey: embeddingConfig.apiKey,
+      baseUrl: embeddingConfig.baseUrl,
+      options: embeddingConfig.options
+    };
+    return _LLMFactory.create(fakeLLMConfig, embeddingConfig);
+  }
+};
+
+export {
+  mergeDefined,
+  OpenAIProvider,
+  AnthropicProvider,
+  OllamaProvider,
+  LLMFactory
+};

package/dist/chunk-HUGLYKD6.mjs

@@ -0,0 +1,84 @@
+import {
+  BaseVectorProvider
+} from "./chunk-VOIWNO5O.mjs";
+import {
+  __spreadValues
+} from "./chunk-I4E63NIC.mjs";
+
+// src/providers/vectordb/ChromaDBProvider.ts
+import axios from "axios";
+var ChromaDBProvider = class extends BaseVectorProvider {
+  constructor(config) {
+    super(config);
+    this.collectionId = "";
+    const opts = config.options;
+    const baseUrl = opts.baseUrl;
+    if (!baseUrl) throw new Error("[ChromaDBProvider] baseUrl is required");
+    this.http = axios.create({
+      baseURL: baseUrl,
+      headers: { "Content-Type": "application/json" }
+    });
+  }
+  async initialize() {
+    const { data } = await this.http.get(`/api/v1/collections/${this.indexName}`);
+    this.collectionId = data.id;
+  }
+  async upsert(doc, namespace) {
+    await this.batchUpsert([doc], namespace);
+  }
+  async batchUpsert(docs, namespace) {
+    const payload = {
+      ids: docs.map((d) => d.id),
+      embeddings: docs.map((d) => d.vector),
+      documents: docs.map((d) => d.content),
+      metadatas: docs.map((d) => __spreadValues(__spreadValues({}, d.metadata || {}), namespace ? { namespace } : {}))
+    };
+    await this.http.post(`/api/v1/collections/${this.collectionId}/add`, payload);
+  }
+  // eslint-disable-next-line @typescript-eslint/no-unused-vars
+  async query(vector, topK, namespace, _filter) {
+    const payload = {
+      query_embeddings: [vector],
+      n_results: topK,
+      where: namespace ? { namespace: { $eq: namespace } } : void 0
+    };
+    const { data } = await this.http.post(`/api/v1/collections/${this.collectionId}/query`, payload);
+    const matches = [];
+    if (data.ids && data.ids[0]) {
+      for (let i = 0; i < data.ids[0].length; i++) {
+        matches.push({
+          id: data.ids[0][i],
+          score: data.distances ? 1 - data.distances[0][i] : 0,
+          content: data.documents ? data.documents[0][i] : "",
+          metadata: data.metadatas ? data.metadatas[0][i] : {}
+        });
+      }
+    }
+    return matches;
+  }
+  async delete(id, _namespace) {
+    await this.http.post(`/api/v1/collections/${this.collectionId}/delete`, {
+      ids: [id],
+      where: _namespace ? { namespace: { $eq: _namespace } } : void 0
+    });
+  }
+  async deleteNamespace(_namespace) {
+    await this.http.post(`/api/v1/collections/${this.collectionId}/delete`, {
+      where: { namespace: { $eq: _namespace } }
+    });
+  }
+  async ping() {
+    try {
+      await this.http.get("/api/v1/heartbeat");
+      return true;
+    } catch (e) {
+      return false;
+    }
+  }
+  async disconnect() {
+  }
+};
+
+export {
+  ChromaDBProvider
+};

package/dist/chunk-I4E63NIC.mjs

@@ -0,0 +1,24 @@
+var __defProp = Object.defineProperty;
+var __defProps = Object.defineProperties;
+var __getOwnPropDescs = Object.getOwnPropertyDescriptors;
+var __getOwnPropSymbols = Object.getOwnPropertySymbols;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __propIsEnum = Object.prototype.propertyIsEnumerable;
+var __defNormalProp = (obj, key, value) => key in obj ? __defProp(obj, key, { enumerable: true, configurable: true, writable: true, value }) : obj[key] = value;
+var __spreadValues = (a, b) => {
+  for (var prop in b || (b = {}))
+    if (__hasOwnProp.call(b, prop))
+      __defNormalProp(a, prop, b[prop]);
+  if (__getOwnPropSymbols)
+    for (var prop of __getOwnPropSymbols(b)) {
+      if (__propIsEnum.call(b, prop))
+        __defNormalProp(a, prop, b[prop]);
+    }
+  return a;
+};
+var __spreadProps = (a, b) => __defProps(a, __getOwnPropDescs(b));
+
+export {
+  __spreadValues,
+  __spreadProps
+};

package/dist/chunk-QEYVWVT5.mjs

@@ -0,0 +1,102 @@
+import {
+  BaseVectorProvider
+} from "./chunk-VOIWNO5O.mjs";
+import {
+  __spreadValues
+} from "./chunk-I4E63NIC.mjs";
+
+// src/providers/vectordb/MongoDBProvider.ts
+import { MongoClient } from "mongodb";
+var MongoDBProvider = class extends BaseVectorProvider {
+  constructor(config) {
+    super(config);
+    const opts = config.options;
+    if (!opts.uri || !opts.database || !opts.collection) {
+      throw new Error("[MongoDBProvider] options uri, database, and collection are required.");
+    }
+    this.client = new MongoClient(opts.uri);
+    this.dbName = opts.database;
+    this.collectionName = opts.collection;
+    this.embeddingKey = opts.embeddingKey || "embedding";
+    this.contentKey = opts.contentKey || "content";
+    this.metadataKey = opts.metadataKey || "metadata";
+  }
+  async initialize() {
+    await this.client.connect();
+    this.db = this.client.db(this.dbName);
+    this.collection = this.db.collection(this.collectionName);
+  }
+  async upsert(doc, namespace) {
+    const document = __spreadValues({
+      _id: doc.id,
+      [this.embeddingKey]: doc.vector,
+      [this.contentKey]: doc.content,
+      [this.metadataKey]: doc.metadata || {}
+    }, namespace ? { namespace } : {});
+    await this.collection.updateOne({ _id: doc.id }, { $set: document }, { upsert: true });
+  }
+  async batchUpsert(docs, namespace) {
+    const operations = docs.map((doc) => ({
+      updateOne: {
+        filter: { _id: doc.id },
+        update: {
+          $set: __spreadValues({
+            [this.embeddingKey]: doc.vector,
+            [this.contentKey]: doc.content,
+            [this.metadataKey]: doc.metadata || {}
+          }, namespace ? { namespace } : {})
+        },
+        upsert: true
+      }
+    }));
+    await this.collection.bulkWrite(operations);
+  }
+  async query(vector, topK, namespace, filter) {
+    const pipeline = [
+      {
+        $vectorSearch: __spreadValues({
+          index: this.config.options.indexName || "vector_index",
+          path: this.embeddingKey,
+          queryVector: vector,
+          numCandidates: Math.max(topK * 10, 100),
+          limit: topK
+        }, filter || namespace ? { filter: __spreadValues(__spreadValues({}, filter || {}), namespace ? { namespace } : {}) } : {})
+      },
+      {
+        $project: {
+          score: { $meta: "vectorSearchScore" },
+          content: `$${this.contentKey}`,
+          metadata: `$${this.metadataKey}`
+        }
+      }
+    ];
+    const results = await this.collection.aggregate(pipeline).toArray();
+    return results.map((res) => ({
+      id: res["_id"].toString(),
+      score: res["score"],
+      content: res["content"],
+      metadata: res["metadata"]
+    }));
+  }
+  async delete(id, namespace) {
+    await this.collection.deleteOne(__spreadValues({ _id: id }, namespace ? { namespace } : {}));
+  }
+  async deleteNamespace(namespace) {
+    await this.collection.deleteMany({ namespace });
+  }
+  async ping() {
+    try {
+      await this.db.command({ ping: 1 });
+      return true;
+    } catch (e) {
+      return false;
+    }
+  }
+  async disconnect() {
+    await this.client.close();
+  }
+};
+
+export {
+  MongoDBProvider
+};