@umituz/web-ai 1.0.1 → 1.0.2

package/dist/index.cjs

@@ -132,6 +132,14 @@ var AIError = class _AIError extends Error {
   }
 };
 
+// src/core/errors/groq-error.ts
+var GroqError = class extends AIError {
+  constructor(type, message, cause, metadata = {}) {
+    super(type, message, cause, metadata);
+    this.name = "GroqError";
+  }
+};
+
 // src/core/errors/error-classifier.ts
 var DEFAULT_RETRYABLE_TYPES = /* @__PURE__ */ new Set([
   "NETWORK_ERROR" /* NETWORK_ERROR */,
@@ -410,6 +418,8 @@ function calculateExponentialBackoff(params) {
 // src/core/constants/identifier.constants.ts
 var MESSAGE_ID_PREFIX = "msg-";
 var CHAT_MESSAGE_ID_SUFFIX_LENGTH = 9;
+var CACHE_KEY_TYPE_COMPLETION = "completion";
+var CACHE_KEY_TYPE_CHAT = "chat";
 var USER_ROLE = "user";
 var ASSISTANT_ROLE = "assistant";
 var QUEUE_PRIORITY_USER_REQUEST = 10;
@@ -1881,12 +1891,22 @@ var DEFAULT_GENERATION_CONFIG = {
   maxTokens: 1024,
   topP: 1
 };
+var DEFAULT_MODELS = {
+  TEXT: GROQ_MODELS.LLAMA_3_1_8B_INSTANT,
+  FAST: GROQ_MODELS.LLAMA_3_1_8B_INSTANT,
+  EXPERIMENTAL: GROQ_MODELS.GPT_OSS_20B
+};
 var GROQ_BASE_URL = "https://api.groq.com/openai/v1";
+var DEFAULT_BASE_URL = GROQ_BASE_URL;
 var CHAT_COMPLETIONS_PATH = "/chat/completions";
+var API_ENDPOINTS = {
+  CHAT: CHAT_COMPLETIONS_PATH
+};
 var GROQ_TIMEOUTS = {
   DEFAULT: 3e4,
   STREAMING: 6e4
 };
+var TIMEOUTS = GROQ_TIMEOUTS;
 var HTTP_HEADERS = {
   CONTENT_TYPE: "Content-Type",
   APPLICATION_JSON: "application/json",
@@ -1945,13 +1965,65 @@ var GroqChatRequestBuilder = class _GroqChatRequestBuilder {
   }
 };
 
+// src/providers/groq/groq-config-store.ts
+var GroqConfigStore = class {
+  config = null;
+  initialize(config) {
+    const validation = validateApiKey(config.apiKey, { minLength: 1 });
+    if (!validation.valid) {
+      throw new GroqError("INVALID_API_KEY" /* INVALID_API_KEY */, validation.reason ?? "Invalid API key");
+    }
+    const normalized = {
+      apiKey: config.apiKey.trim(),
+      baseUrl: config.baseUrl ?? DEFAULT_BASE_URL,
+      timeoutMs: config.timeoutMs ?? TIMEOUTS.DEFAULT,
+      ...config.models !== void 0 ? { models: config.models } : {},
+      ...config.textModel !== void 0 ? { textModel: config.textModel } : {}
+    };
+    this.config = normalized;
+    return normalized;
+  }
+  get() {
+    if (!this.config) {
+      throw new GroqError(
+        "MISSING_CONFIG" /* MISSING_CONFIG */,
+        "Client not initialized. Call configureProvider first."
+      );
+    }
+    return this.config;
+  }
+  getOrNull() {
+    return this.config;
+  }
+  isInitialized() {
+    return this.config !== null;
+  }
+  setApiKey(apiKey) {
+    const current = this.get();
+    const validation = validateApiKey(apiKey, { minLength: 1 });
+    if (!validation.valid) {
+      throw new GroqError("INVALID_API_KEY" /* INVALID_API_KEY */, validation.reason ?? "Invalid API key");
+    }
+    this.config = { ...current, apiKey: apiKey.trim() };
+  }
+  reset() {
+    this.config = null;
+  }
+};
+
 // src/providers/groq/groq-http-transport.ts
 var GroqHttpTransport = class {
-  constructor(config) {
-    this.config = config;
+  constructor(configSource) {
+    this.configSource = configSource;
   }
-  config;
+  configSource;
   pendingAbortControllers = /* @__PURE__ */ new Set();
+  getConfig() {
+    if ("get" in this.configSource && typeof this.configSource.get === "function") {
+      return this.configSource.get();
+    }
+    return this.configSource;
+  }
   cancelPendingRequests() {
     for (const controller of this.pendingAbortControllers) {
       controller.abort();
@@ -1966,7 +2038,7 @@ var GroqHttpTransport = class {
   }
   async streamChatCompletion(request, options = {}) {
     const url = `${this.baseUrl()}${CHAT_COMPLETIONS_PATH}`;
-    const timeout = this.config.timeoutMs ?? GROQ_TIMEOUTS.STREAMING;
+    const timeout = this.getConfig().timeoutMs ?? GROQ_TIMEOUTS.STREAMING;
     const controller = new AbortController();
     const timeoutId = setTimeout(() => controller.abort(), timeout);
     const onExternalAbort = () => controller.abort();
@@ -1993,11 +2065,11 @@ var GroqHttpTransport = class {
     }
   }
   baseUrl() {
-    return this.config.baseUrl ?? GROQ_BASE_URL;
+    return this.getConfig().baseUrl ?? GROQ_BASE_URL;
   }
   async request(endpoint, body) {
     const url = `${this.baseUrl()}${endpoint}`;
-    const timeout = this.config.timeoutMs ?? GROQ_TIMEOUTS.DEFAULT;
+    const timeout = this.getConfig().timeoutMs ?? GROQ_TIMEOUTS.DEFAULT;
     const controller = new AbortController();
     const timeoutId = setTimeout(() => controller.abort(), timeout);
     this.pendingAbortControllers.add(controller);
@@ -2020,7 +2092,7 @@ var GroqHttpTransport = class {
   buildHeaders() {
     return {
       [HTTP_HEADERS.CONTENT_TYPE]: HTTP_HEADERS.APPLICATION_JSON,
-      [HTTP_HEADERS.AUTHORIZATION]: `${HTTP_HEADERS.BEARER_PREFIX}${this.config.apiKey}`
+      [HTTP_HEADERS.AUTHORIZATION]: `${HTTP_HEADERS.BEARER_PREFIX}${this.getConfig().apiKey}`
     };
   }
   async throwFromErrorResponse(response) {
@@ -2064,21 +2136,40 @@ var GroqHttpTransport = class {
 
 // src/providers/groq/groq-http-client.ts
 var GroqHttpClient = class {
+  configStore;
   transport;
   executor;
-  constructor(config, deps = {}) {
-    this.transport = deps.transport ?? new GroqHttpTransport(config);
+  constructor(configOrDeps, maybeDeps = {}) {
+    const isConfig = typeof configOrDeps === "object" && configOrDeps !== null && "apiKey" in configOrDeps && typeof configOrDeps.apiKey === "string";
+    const deps = isConfig ? maybeDeps : configOrDeps ?? {};
+    this.configStore = deps.configStore ?? new GroqConfigStore();
+    if (isConfig) {
+      this.configStore.initialize(configOrDeps);
+    }
+    this.transport = deps.transport ?? new GroqHttpTransport(this.configStore);
     const retryManager2 = deps.retryManager ?? new RetryManager();
     const queue = deps.queue ?? new RequestQueue();
     const deduplicator = deps.deduplicator ?? new RequestDeduplicator();
     this.executor = new CompositeExecutor([
       new DedupeExecutor(deduplicator, {
-        generateKey: (request) => deduplicatorKey(request, config.apiKey)
+        generateKey: (request) => deduplicatorKey(request, this.configStore.isInitialized() ? this.configStore.get().apiKey : "")
       }),
       new RetryExecutor(retryManager2),
       new QueueExecutor(queue, QUEUE_PRIORITY_USER_REQUEST)
     ]);
   }
+  initialize(config) {
+    this.configStore.initialize(config);
+  }
+  isInitialized() {
+    return this.configStore.isInitialized();
+  }
+  getApiKey() {
+    return this.configStore.get().apiKey;
+  }
+  setApiKey(apiKey) {
+    this.configStore.setApiKey(apiKey);
+  }
   cancelPendingRequests() {
     this.transport.cancelPendingRequests();
   }
@@ -2091,6 +2182,10 @@ var GroqHttpClient = class {
   streamChatCompletion(request, options) {
     return this.transport.streamChatCompletion(request, options);
   }
+  reset() {
+    this.transport.cancelPendingRequests();
+    this.configStore.reset();
+  }
 };
 function deduplicatorKey(request, apiKey) {
   const { model, messages, stream: _stream, ...rest } = request;
@@ -2967,6 +3062,278 @@ function createAI(config = {}) {
   return factory;
 }
 
+// src/providers/groq/groq-messages.ts
+var createUserMessage = (content) => ({
+  role: "user",
+  content
+});
+var createAssistantMessage = (content) => ({
+  role: "assistant",
+  content
+});
+var createSystemMessage = (content) => ({
+  role: "system",
+  content
+});
+var createMessage = (role, content) => ({
+  role,
+  content
+});
+var promptToMessages = (prompt) => [
+  createUserMessage(prompt)
+];
+var extractTextFromMessages = (messages) => messages.map((m) => m.content).join("\n\n");
+var formatMessagesForDisplay = (messages) => messages.map((m) => `[${m.role.toUpperCase()}]: ${m.content}`).join("\n\n---\n\n");
+
+// src/providers/groq/groq-structured-json.ts
+async function requestStructuredJson(client, request) {
+  const chatRequest = {
+    model: request.model,
+    messages: [createUserMessage(request.instructions)],
+    temperature: request.temperature,
+    max_tokens: request.maxTokens,
+    response_format: { type: "json_object" },
+    ...request.reasoningEffort !== void 0 ? { reasoning_effort: request.reasoningEffort } : {}
+  };
+  const response = await client.postChatCompletion(chatRequest);
+  const content = response.choices[0]?.message?.content ?? "";
+  const candidate = extractJsonCandidate(content);
+  try {
+    return JSON.parse(candidate);
+  } catch (err) {
+    throw new GroqError(
+      "UNKNOWN_ERROR" /* UNKNOWN_ERROR */,
+      "model response contained no parsable json object",
+      err instanceof Error ? err : void 0
+    );
+  }
+}
+
+// src/providers/groq/text-generation.service.ts
+var STRUCTURED_OUTPUT_TEMPERATURE = 0.1;
+var TextGenerationService = class {
+  constructor(httpClient, cache = new CacheManager(), defaultModel = DEFAULT_MODELS.TEXT) {
+    this.httpClient = httpClient;
+    this.cache = cache;
+    this.defaultModel = defaultModel;
+  }
+  httpClient;
+  cache;
+  defaultModel;
+  id = "groq-text-generation";
+  name = "Groq Text Generation";
+  capabilities = ["text"];
+  jsonParser = new JsonResponseParser();
+  generateCompletion(prompt, options = {}) {
+    const model = options.model ?? this.defaultModel;
+    return this.executeWithCache(
+      {
+        type: CACHE_KEY_TYPE_COMPLETION,
+        model,
+        prompt,
+        config: options.generationConfig
+      },
+      async () => {
+        const request = GroqChatRequestBuilder.for(model).withMessages([createUserMessage(prompt)]).withConfig(options.generationConfig).build();
+        const response = await this.httpClient.postChatCompletion(request);
+        return this.extractContent(response.choices[0]?.message?.content);
+      }
+    );
+  }
+  generateChatCompletion(messages, options = {}) {
+    const model = options.model ?? this.defaultModel;
+    return this.executeWithCache(
+      {
+        type: CACHE_KEY_TYPE_CHAT,
+        model,
+        messages: [...messages],
+        config: options.generationConfig
+      },
+      async () => {
+        const request = GroqChatRequestBuilder.for(model).withMessages(messages).withConfig(options.generationConfig).build();
+        const response = await this.httpClient.postChatCompletion(request);
+        return this.extractContent(response.choices[0]?.message?.content);
+      }
+    );
+  }
+  generateStructured(promptOrRequest, options) {
+    if (typeof promptOrRequest === "string") {
+      const model2 = options?.model ?? this.defaultModel;
+      const jsonPrompt = StructuredPromptBuilder.for(promptOrRequest).withSchema(options?.schema).build();
+      const request = GroqChatRequestBuilder.for(model2).withMessages([createUserMessage(jsonPrompt)]).withConfig({
+        ...options?.generationConfig,
+        temperature: STRUCTURED_OUTPUT_TEMPERATURE
+      }).build();
+      return this.httpClient.postChatCompletion(request).then((response) => {
+        const content = this.extractContent(response.choices[0]?.message?.content);
+        return this.jsonParser.parse(content);
+      });
+    }
+    const req = promptOrRequest;
+    const model = req.model ?? (req.tier ? this.resolveModel(req.tier) : this.defaultModel);
+    const schemaObj = req.schema;
+    const messages = [];
+    if (req.systemPrompt) {
+      messages.push({ role: "system", content: req.systemPrompt });
+    }
+    for (const m of req.messages) {
+      messages.push({ role: m.role, content: m.content });
+    }
+    const lastMsg = messages[messages.length - 1];
+    if (lastMsg) {
+      const structuredPrompt = StructuredPromptBuilder.for(lastMsg.content).withSchema(schemaObj).build();
+      messages[messages.length - 1] = { ...lastMsg, content: structuredPrompt };
+    }
+    const groqReq = GroqChatRequestBuilder.for(model).withMessages(messages).withConfig({
+      temperature: req.temperature ?? STRUCTURED_OUTPUT_TEMPERATURE,
+      maxTokens: req.maxTokens
+    }).build();
+    return this.httpClient.postChatCompletion(groqReq).then((response) => {
+      const content = this.extractContent(response.choices[0]?.message?.content);
+      const parsed = this.jsonParser.parse(content);
+      const genResult = {
+        text: content,
+        model: response.model,
+        providerId: this.id,
+        finishReason: response.choices[0]?.finish_reason ?? "stop",
+        usage: response.usage ? {
+          inputTokens: response.usage.prompt_tokens,
+          outputTokens: response.usage.completion_tokens,
+          totalTokens: response.usage.total_tokens
+        } : void 0
+      };
+      return { value: parsed, result: genResult };
+    });
+  }
+  streamCompletion(prompt, callbacks, options = {}) {
+    return this.streamChatCompletion([createUserMessage(prompt)], callbacks, options);
+  }
+  async streamChatCompletion(messages, callbacks, options = {}) {
+    const model = options.model ?? this.defaultModel;
+    const request = GroqChatRequestBuilder.for(model).withMessages(messages).withConfig(options.generationConfig).withStream(true).build();
+    const parser = new StreamingResponseParser(
+      {
+        onChunk: (chunk) => callbacks.onChunk?.(chunk)
+      },
+      extractContentFromSsePayload
+    );
+    try {
+      const stream = await this.httpClient.streamChatCompletion(request, {
+        signal: options.signal
+      });
+      const fullText = await parser.parse(stream);
+      callbacks.onComplete?.(fullText);
+      return fullText;
+    } catch (error) {
+      const err = error instanceof Error ? error : new Error(String(error));
+      callbacks.onError?.(err);
+      throw err;
+    }
+  }
+  clearCache() {
+    this.cache.clear();
+  }
+  // --- ITextProvider implementation ---
+  resolveModel(tier) {
+    return GROQ_TIERS[tier] ?? this.defaultModel;
+  }
+  async generate(request) {
+    const messages = [];
+    if (request.systemPrompt) {
+      messages.push({ role: "system", content: request.systemPrompt });
+    }
+    for (const m of request.messages) {
+      messages.push({ role: m.role, content: m.content });
+    }
+    const model = request.model ?? (request.tier ? this.resolveModel(request.tier) : this.defaultModel);
+    const groqReq = GroqChatRequestBuilder.for(model).withMessages(messages).withConfig({
+      temperature: request.temperature,
+      maxTokens: request.maxTokens
+    }).build();
+    const resp = await this.httpClient.postChatCompletion(groqReq);
+    const content = resp.choices[0]?.message?.content ?? "";
+    return {
+      text: content,
+      finishReason: resp.choices[0]?.finish_reason ?? "stop",
+      model: resp.model,
+      providerId: this.id,
+      usage: resp.usage ? {
+        inputTokens: resp.usage.prompt_tokens,
+        outputTokens: resp.usage.completion_tokens,
+        totalTokens: resp.usage.total_tokens
+      } : void 0
+    };
+  }
+  async stream(request, callbacks) {
+    const messages = [];
+    if (request.systemPrompt) {
+      messages.push({ role: "system", content: request.systemPrompt });
+    }
+    for (const m of request.messages) {
+      messages.push({ role: m.role, content: m.content });
+    }
+    const model = request.model ?? (request.tier ? this.resolveModel(request.tier) : this.defaultModel);
+    const text = await this.streamChatCompletion(
+      messages,
+      {
+        onChunk: callbacks.onChunk,
+        onError: callbacks.onError
+      },
+      {
+        model,
+        signal: request.signal,
+        generationConfig: {
+          temperature: request.temperature,
+          maxTokens: request.maxTokens
+        }
+      }
+    );
+    const result = {
+      text,
+      finishReason: "stop",
+      model,
+      providerId: this.id
+    };
+    callbacks.onComplete?.(result);
+    return result;
+  }
+  async countTokens(text) {
+    return Math.max(1, Math.ceil(text.length / 4));
+  }
+  async estimateCost(_request) {
+    return 0;
+  }
+  async healthCheck() {
+    try {
+      await this.generateCompletion("ping", {
+        generationConfig: { maxTokens: 1 }
+      });
+      return "healthy";
+    } catch {
+      return "unavailable";
+    }
+  }
+  async executeWithCache(params, factory) {
+    const key = this.cache.generateKey(params);
+    const cached = this.cache.get(key);
+    if (cached !== null) {
+      return cached;
+    }
+    const result = await factory();
+    this.cache.set(key, result);
+    return result;
+  }
+  extractContent(content) {
+    if (!content) {
+      throw new GroqError(
+        "UNKNOWN_ERROR" /* UNKNOWN_ERROR */,
+        "No content generated from Groq API"
+      );
+    }
+    return content;
+  }
+};
+
 // src/core/calculations/random-base36-suffix.ts
 var BASE_36_RADIX = 36;
 var DECIMAL_OFFSET = 2;
@@ -3194,12 +3561,37 @@ function readNumber(source, keys) {
   return null;
 }
 
+// src/providers/groq/groq-container.ts
+var container = null;
+function configureProvider(config) {
+  const configStore = new GroqConfigStore();
+  const transport = new GroqHttpTransport(configStore);
+  const httpClient = new GroqHttpClient({ configStore, transport });
+  httpClient.initialize(config);
+  const textGeneration = new TextGenerationService(
+    httpClient,
+    new CacheManager(),
+    config.textModel
+  );
+  const chat = new ChatService(textGeneration);
+  container = { groqHttpClient: httpClient, textGeneration, chat };
+  return container;
+}
+function getProvider() {
+  return container;
+}
+function resetProvider() {
+  container?.groqHttpClient.reset();
+  container = null;
+}
+
 exports.AIError = AIError;
 exports.AIErrorType = AIErrorType;
 exports.ANTHROPIC_API_VERSION = ANTHROPIC_API_VERSION;
 exports.ANTHROPIC_DEFAULT_MAX_TOKENS = ANTHROPIC_DEFAULT_MAX_TOKENS;
 exports.ANTHROPIC_MODELS = ANTHROPIC_MODELS;
 exports.ANTHROPIC_TIERS = ANTHROPIC_TIERS;
+exports.API_ENDPOINTS = API_ENDPOINTS;
 exports.AnthropicHttpTransport = AnthropicHttpTransport;
 exports.AnthropicProvider = AnthropicProvider;
 exports.AnthropicRequestBuilder = AnthropicRequestBuilder;
@@ -3209,8 +3601,11 @@ exports.ChatIdGenerator = ChatIdGenerator;
 exports.ChatService = ChatService;
 exports.CompositeExecutor = CompositeExecutor;
 exports.DEFAULT_API_KEY_RULE = DEFAULT_API_KEY_RULE;
+exports.DEFAULT_BASE_URL = DEFAULT_BASE_URL;
 exports.DEFAULT_ERROR_CLASSIFICATION_RULE = DEFAULT_ERROR_CLASSIFICATION_RULE;
+exports.DEFAULT_GENERATION_CONFIG = DEFAULT_GENERATION_CONFIG;
 exports.DEFAULT_HTTP_STATUS_RULE = DEFAULT_HTTP_STATUS_RULE;
+exports.DEFAULT_MODELS = DEFAULT_MODELS;
 exports.DEFAULT_PROMPT_STRATEGY = DEFAULT_PROMPT_STRATEGY;
 exports.DEFAULT_RETRY_BACKOFF_MULTIPLIER = DEFAULT_RETRY_BACKOFF_MULTIPLIER;
 exports.DEFAULT_RETRY_INITIAL_DELAY_MS = DEFAULT_RETRY_INITIAL_DELAY_MS;
@@ -3225,6 +3620,9 @@ exports.GROQ_MODELS = GROQ_MODELS;
 exports.GROQ_TIERS = GROQ_TIERS;
 exports.GeminiProvider = GeminiProvider;
 exports.GroqChatRequestBuilder = GroqChatRequestBuilder;
+exports.GroqConfigStore = GroqConfigStore;
+exports.GroqError = GroqError;
+exports.GroqErrorType = AIErrorType;
 exports.GroqHttpClient = GroqHttpClient;
 exports.GroqHttpTransport = GroqHttpTransport;
 exports.GroqProvider = GroqProvider;
@@ -3245,6 +3643,8 @@ exports.SSE_PREFIX = SSE_PREFIX;
 exports.StreamingResponseParser = StreamingResponseParser;
 exports.StructuredPromptBuilder = StructuredPromptBuilder;
 exports.TEXT_TIERS = TEXT_TIERS;
+exports.TIMEOUTS = TIMEOUTS;
+exports.TextGenerationService = TextGenerationService;
 exports.ThrottleManager = ThrottleManager;
 exports.USER_FRIENDLY_ERROR_MESSAGES = USER_FRIENDLY_ERROR_MESSAGES;
 exports.asObjectArray = asObjectArray;
@@ -3257,13 +3657,18 @@ exports.cacheManager = cacheManager;
 exports.chatIdGenerator = chatIdGenerator;
 exports.clampScore = clampScore;
 exports.classifyHttpStatus = classifyHttpStatus;
+exports.configureProvider = configureProvider;
 exports.consumeAnthropicStream = consumeAnthropicStream;
 exports.consumeGroqStream = consumeGroqStream;
 exports.createAI = createAI;
+exports.createAssistantMessage = createAssistantMessage;
 exports.createFalProvider = createFalProvider;
 exports.createGeminiProvider = createGeminiProvider;
+exports.createMessage = createMessage;
 exports.createPrunaProvider = createPrunaProvider;
 exports.createSseResponse = createSseResponse;
+exports.createSystemMessage = createSystemMessage;
+exports.createUserMessage = createUserMessage;
 exports.debounce = debounce;
 exports.debounceManager = debounceManager;
 exports.defaultEffortForModel = defaultEffortForModel;
@@ -3276,9 +3681,12 @@ exports.extractJsonCandidate = extractJsonCandidate;
 exports.extractJsonFromCodeFence = extractJsonFromCodeFence;
 exports.extractJsonObject = extractJsonObject;
 exports.extractSseDataPayload = extractSseDataPayload;
+exports.extractTextFromMessages = extractTextFromMessages;
 exports.extractTokenUsage = extractTokenUsage;
 exports.formatErrorForLogging = formatErrorForLogging;
+exports.formatMessagesForDisplay = formatMessagesForDisplay;
 exports.generateWithPruna = generateWithPruna;
+exports.getProvider = getProvider;
 exports.isAuthErrorType = isAuthErrorType;
 exports.isMediaProvider = isMediaProvider;
 exports.isRetryableErrorType = isRetryableErrorType;
@@ -3292,9 +3700,12 @@ exports.mergeTierOverrides = mergeTierOverrides;
 exports.messageFormatter = messageFormatter;
 exports.normalizeSseChunk = normalizeSseChunk;
 exports.oneOfCatalog = oneOfCatalog;
+exports.promptToMessages = promptToMessages;
 exports.repairCommonJsonDefects = repairCommonJsonDefects;
 exports.requestDeduplicator = requestDeduplicator;
 exports.requestQueue = requestQueue;
+exports.requestStructuredJson = requestStructuredJson;
+exports.resetProvider = resetProvider;
 exports.resolveModel = resolveModel;
 exports.resolveUserFriendlyErrorMessage = resolveUserFriendlyErrorMessage;
 exports.resolveUserFriendlyMessage = resolveUserFriendlyMessage;