universal-llm-client 4.5.0 → 4.5.1

package/CHANGELOG.md

@@ -5,6 +5,18 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [4.5.1] - 2026-06-14
+
+### Added
+
+- **Automatic vLLM tool-calling fallback (OpenAI-compatible provider)** — when a vLLM / OpenAI-compatible server rejects native tool calling (started without `--enable-auto-tool-choice` / `--tool-call-parser`, surfaced as a `"auto" tool choice requires …` 400), the client now transparently retries with a text-level tool protocol: it drops `tools` / `tool_choice`, injects a `<tool_call>name({…})</tool_call>` instruction, and recovers the emitted calls back into `message.tool_calls` (parsing `<tool_call>`, `<function=…>`, and `name(args)` forms). Works for both `chat()` and `chatStream()`, and emits a one-time warning pointing to the flags needed for native parsing. Tool calling now works against vLLM servers not started with tool-parser flags.
+
+### Changed
+
+- **Non-leading system messages normalized for OpenAI-compatible servers** — leading system messages are preserved, but a system message appearing *after* the conversation has started is rewritten to a `user` turn prefixed with `[SYSTEM MESSAGE]` (many OpenAI-compatible servers and chat templates reject mid-conversation system roles). Multimodal content is flattened to text for the prefix.
+- **`chat()` / `chatStream()` no longer auto-attach registered tools** — tools are sent only when explicitly passed via `options.tools` (the `chatWithTools` path already does this). Previously every tool added with `registerTool` was attached to plain `chat()` calls. Pass `{ tools }` explicitly (or use `chatWithTools`) if you relied on the old behavior.
+- **Minimal published package** — the npm tarball now ships only `dist/` (compiled JS + type declarations) plus `README` / `CHANGELOG` / `LICENSE`. `src/` (tests, demos, test-scripts) is no longer published, and the build no longer emits `.js.map` / `.d.ts.map` (they referenced sources that aren't shipped and triggered "missing source" warnings in consumer bundlers). Package size dropped ~78% (357 kB → 80 kB, 175 → 44 files). Because `src/` is no longer whitelisted, a local demo model cache under `src/` can never bloat the published package again.
+
 ## [4.5.0] - 2026-06-14
 
 ### Added

package/README.md

@@ -1,5 +1,7 @@
 # universal-llm-client
 
+[![npm version](https://img.shields.io/npm/v/universal-llm-client.svg)](https://www.npmjs.com/package/universal-llm-client) [![CI](https://github.com/igorls/universal-llm-client/actions/workflows/ci.yml/badge.svg)](https://github.com/igorls/universal-llm-client/actions/workflows/ci.yml) [![npm downloads](https://img.shields.io/npm/dm/universal-llm-client.svg)](https://www.npmjs.com/package/universal-llm-client) [![License: MIT](https://img.shields.io/npm/l/universal-llm-client.svg)](https://github.com/igorls/universal-llm-client/blob/main/LICENSE)
+
 A universal LLM client for JavaScript/TypeScript with **transparent provider failover** and a **provider-agnostic reasoning API** — one set of code across OpenAI, Anthropic, Google Gemini, Ollama, vLLM, and any OpenAI-compatible endpoint. Streaming tool execution, structured output, generation stats, and native observability included.
 
 ```typescript

package/dist/ai-model.d.ts

@@ -148,4 +148,3 @@ export declare class AIModel {
     private createClient;
     private normalizeType;
 }
-//# sourceMappingURL=ai-model.d.ts.map

package/dist/ai-model.js

@@ -290,4 +290,3 @@ export class AIModel {
         return type.toLowerCase();
     }
 }
-//# sourceMappingURL=ai-model.js.map

package/dist/auditor.d.ts

@@ -79,4 +79,3 @@ export declare class BufferedAuditor implements Auditor {
     /** Clear all buffered events without flushing */
     clear(): void;
 }
-//# sourceMappingURL=auditor.d.ts.map

package/dist/auditor.js

@@ -110,4 +110,3 @@ export class BufferedAuditor {
         this.events.length = 0;
     }
 }
-//# sourceMappingURL=auditor.js.map

package/dist/client.d.ts

@@ -86,4 +86,3 @@ export declare abstract class BaseLLMClient {
      */
     protected validateStructuredResponse(content: string, config: SchemaConfig<unknown>): void;
 }
-//# sourceMappingURL=client.d.ts.map

package/dist/client.js

@@ -297,4 +297,3 @@ export class BaseLLMClient {
         }
     }
 }
-//# sourceMappingURL=client.js.map

package/dist/gemma-channel.d.ts

@@ -11,4 +11,3 @@ export interface GemmaThoughtExtraction {
 export declare const GEMMA_THOUGHT_OPENERS: readonly ["<|channel>thought", "<|thought"];
 export declare function extractGemmaThoughtChannels(input: string): GemmaThoughtExtraction;
 export declare function normalizeGemmaThought(thought: string): string;
-//# sourceMappingURL=gemma-channel.d.ts.map

package/dist/gemma-channel.js

@@ -35,4 +35,3 @@ export function extractGemmaThoughtChannels(input) {
 export function normalizeGemmaThought(thought) {
     return thought.replace(/^\s+/, '').replace(/\s+$/, '');
 }
-//# sourceMappingURL=gemma-channel.js.map

package/dist/gemma-diffusion.d.ts

@@ -46,4 +46,3 @@ export declare function gemmaArgsToJson(body: string): string;
  * clean answer text.
  */
 export declare function parseGemmaDiffusionOutput(raw: string): GemmaDiffusionParsed;
-//# sourceMappingURL=gemma-diffusion.d.ts.map

package/dist/gemma-diffusion.js

@@ -144,4 +144,3 @@ export function parseGemmaDiffusionOutput(raw) {
         toolCalls,
     };
 }
-//# sourceMappingURL=gemma-diffusion.js.map

package/dist/http.d.ts

@@ -50,4 +50,3 @@ export declare function parseSSE(stream: AsyncGenerator<string>): AsyncGenerator
  * Respects authHeader / authPrefix from config for Azure-style or gateway auth.
  */
 export declare function buildHeaders(options: LLMClientOptions): Record<string, string>;
-//# sourceMappingURL=http.d.ts.map

package/dist/http.js

@@ -197,4 +197,3 @@ export function buildHeaders(options) {
     }
     return headers;
 }
-//# sourceMappingURL=http.js.map

package/dist/index.d.ts

@@ -15,4 +15,3 @@ export { httpRequest, httpStream, parseNDJSON, parseSSE, buildHeaders, type Http
 export { isGemmaDiffusionModel, parseGemmaDiffusionOutput, gemmaArgsToJson, type GemmaDiffusionParsed, type GemmaParsedToolCall, } from './gemma-diffusion.js';
 export { MCPToolBridge, type MCPBridgeConfig, type MCPServerConfig, type MCPTool, } from './mcp.js';
 export { StructuredOutputError, type StructuredOutputErrorOptions, type StructuredOutputOptions, type StructuredOutputResult, type StructuredOutputSuccess, type StructuredOutputFailure, type JSONSchema, type SchemaProvider, type ProviderSchema, type SchemaConfig, isStructuredOutputSuccess, isStructuredOutputFailure, normalizeJsonSchema, convertToProviderSchema, stripUnsupportedFeatures, getJsonSchema, getJsonSchemaFromConfig, parseStructured, tryParseStructured, validateStructuredOutput, stripJsonFences, StreamingJsonParser, type StreamingStructuredResult, } from './structured-output.js';
-//# sourceMappingURL=index.d.ts.map

package/dist/index.js

@@ -52,4 +52,3 @@ normalizeJsonSchema, convertToProviderSchema, stripUnsupportedFeatures, getJsonS
 parseStructured, tryParseStructured, validateStructuredOutput, stripJsonFences, 
 // Streaming parser
 StreamingJsonParser, } from './structured-output.js';
-//# sourceMappingURL=index.js.map

package/dist/interfaces.d.ts

@@ -677,4 +677,3 @@ export declare function hasImages(content: LLMMessageContent): boolean;
 export declare function audioContent(base64Data: string, mimeType: string): LLMAudioContent;
 /** Check if message content contains audio */
 export declare function hasAudio(content: LLMMessageContent): boolean;
-//# sourceMappingURL=interfaces.d.ts.map

package/dist/interfaces.js

@@ -74,4 +74,3 @@ export function hasAudio(content) {
         return false;
     return content.some(part => part.type === 'audio');
 }
-//# sourceMappingURL=interfaces.js.map

package/dist/mcp.d.ts

@@ -82,4 +82,3 @@ export declare class MCPToolBridge {
     private createToolHandler;
     private convertInputSchema;
 }
-//# sourceMappingURL=mcp.d.ts.map

package/dist/mcp.js

@@ -252,4 +252,3 @@ export class MCPToolBridge {
         };
     }
 }
-//# sourceMappingURL=mcp.js.map

package/dist/providers/anthropic.d.ts

@@ -53,4 +53,3 @@ export declare class AnthropicClient extends BaseLLMClient {
     /** Extract text from multimodal content */
     private extractText;
 }
-//# sourceMappingURL=anthropic.d.ts.map

package/dist/providers/anthropic.js

@@ -546,4 +546,3 @@ export class AnthropicClient extends BaseLLMClient {
             .join('');
     }
 }
-//# sourceMappingURL=anthropic.js.map

package/dist/providers/google.d.ts

@@ -56,4 +56,3 @@ export declare class GoogleClient extends BaseLLMClient {
      */
     private fetchWithFlexRetry;
 }
-//# sourceMappingURL=google.d.ts.map

package/dist/providers/google.js

@@ -712,4 +712,3 @@ export class GoogleClient extends BaseLLMClient {
         throw new Error('Unreachable');
     }
 }
-//# sourceMappingURL=google.js.map

package/dist/providers/index.d.ts

@@ -5,4 +5,3 @@ export { OllamaClient } from './ollama.js';
 export { OpenAICompatibleClient } from './openai.js';
 export { GoogleClient } from './google.js';
 export { AnthropicClient } from './anthropic.js';
-//# sourceMappingURL=index.d.ts.map

package/dist/providers/index.js

@@ -5,4 +5,3 @@ export { OllamaClient } from './ollama.js';
 export { OpenAICompatibleClient } from './openai.js';
 export { GoogleClient } from './google.js';
 export { AnthropicClient } from './anthropic.js';
-//# sourceMappingURL=index.js.map

package/dist/providers/ollama.d.ts

@@ -37,4 +37,3 @@ export declare class OllamaClient extends BaseLLMClient {
      */
     private buildFormatParameter;
 }
-//# sourceMappingURL=ollama.d.ts.map

package/dist/providers/ollama.js

@@ -426,4 +426,3 @@ export class OllamaClient extends BaseLLMClient {
         return 'json';
     }
 }
-//# sourceMappingURL=ollama.js.map

package/dist/providers/openai.d.ts

@@ -9,6 +9,7 @@ import type { LLMClientOptions, LLMChatMessage, LLMChatResponse, ChatOptions } f
 import type { DecodedEvent } from '../stream-decoder.js';
 import type { Auditor } from '../auditor.js';
 export declare class OpenAICompatibleClient extends BaseLLMClient {
+    private warnedVllmToolFallback;
     /**
      * DiffusionGemma on trimmed vLLM builds has no server-side reasoning or
      * tool-call parser — the native channel protocol is handled client-side
@@ -22,6 +23,7 @@ export declare class OpenAICompatibleClient extends BaseLLMClient {
      */
     private buildUrl;
     constructor(options: LLMClientOptions, auditor?: Auditor);
+    private warnVllmToolFallback;
     chat(messages: LLMChatMessage[], options?: ChatOptions): Promise<LLMChatResponse>;
     chatStream(messages: LLMChatMessage[], options?: ChatOptions): AsyncGenerator<DecodedEvent, LLMChatResponse | void, unknown>;
     private normalizeToolCall;
@@ -35,4 +37,3 @@ export declare class OpenAICompatibleClient extends BaseLLMClient {
      */
     private buildResponseFormat;
 }
-//# sourceMappingURL=openai.d.ts.map