@prismshadow/mmsp 0.5.0

package/dist/baseClient.js

@@ -0,0 +1,284 @@
+"use strict";
+// Copyright 2025 Prism Shadow. and/or its affiliates
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.LLMClient = void 0;
+const errors_1 = require("./errors");
+const legacy_1 = require("./legacy");
+const streamItems_1 = require("./streamItems");
+function deltaEvent(item) {
+    return {
+        role: "assistant",
+        event_type: "delta",
+        content_items: [item],
+        usage_metadata: null,
+        finish_reason: null,
+        created_at: Date.now(),
+    };
+}
+/**
+ * Turns the events a client yields into the public stream. A client yields deltas only; every
+ * one goes out as it arrives, in an event of its own, and `StreamItems` closes each item with its
+ * done item when the next item begins or the stream ends. The usage and finish reason of the
+ * client's stop events are merged into the one final stop.
+ */
+class PublicStream {
+    constructor(client) {
+        this.client = client;
+        this.usageMetadata = null;
+        this.finishReason = null;
+        this.doneItems = [];
+        this.items = new streamItems_1.StreamItems(client);
+    }
+    // a generator, so the deltas of an event reach the caller even when a later one fails
+    *push(event) {
+        if (event.event_type === "delta" &&
+            (event.usage_metadata != null || event.finish_reason != null)) {
+            throw new errors_1.StreamProtocolError({
+                client: this.client,
+                message: "a delta event carries usage_metadata or finish_reason",
+            });
+        }
+        for (const delta of event.content_items) {
+            yield* this.emit(this.items.delta(delta));
+        }
+        if (event.usage_metadata) {
+            const usage = this.usageMetadata ?? {
+                cached_tokens: null,
+                prompt_tokens: null,
+                thoughts_tokens: null,
+                response_tokens: null,
+            };
+            for (const field of [
+                "cached_tokens",
+                "prompt_tokens",
+                "thoughts_tokens",
+                "response_tokens",
+            ]) {
+                if (event.usage_metadata[field] != null) {
+                    usage[field] = event.usage_metadata[field];
+                }
+            }
+            this.usageMetadata = usage;
+        }
+        if (event.finish_reason) {
+            this.finishReason = event.finish_reason;
+        }
+    }
+    /**
+     * The client's stream ended: the item still streaming is done.
+     */
+    *end() {
+        yield* this.emit(this.items.end());
+    }
+    *emit(items) {
+        for (const item of items) {
+            if (item.type.endsWith(".done")) {
+                this.doneItems.push(item);
+            }
+            yield deltaEvent(item);
+        }
+    }
+    /**
+     * Build the stop event once the client's stream ended, rejecting a response that cannot be one.
+     */
+    stop() {
+        if (this.usageMetadata === null) {
+            throw new Error("Streaming response ended without usage_metadata");
+        }
+        if (this.finishReason === null) {
+            throw new Error("Streaming response ended without finish_reason");
+        }
+        // replaying a thinking-only assistant message on the next turn fails with a 400 error
+        const thinkingOnly = this.doneItems.every((item) => item.type === "thinking.done" || item.type === "inline_thinking.done");
+        if (thinkingOnly) {
+            throw new errors_1.EmptyResponseError({
+                client: this.client,
+                finishReason: this.finishReason,
+                usageMetadata: this.usageMetadata,
+            });
+        }
+        return {
+            role: "assistant",
+            event_type: "stop",
+            content_items: [],
+            usage_metadata: this.usageMetadata,
+            finish_reason: this.finishReason,
+            created_at: Date.now(),
+        };
+    }
+}
+/**
+ * Abstract base class for LLM clients.
+ *
+ * All model-specific clients must inherit from this class and implement
+ * the required abstract methods for complete SDK abstraction.
+ */
+class LLMClient {
+    constructor() {
+        this._model = "";
+        this._history = [];
+    }
+    /**
+     * Concatenate a stream of universal events into a single universal message.
+     *
+     * @param events - List of universal events from streaming response
+     * @returns Complete universal message object: every done item in stream order, with the
+     *   usage, finish reason and timestamp of the stop event
+     */
+    concatUniEventsToUniMessage(events) {
+        const contentItems = [];
+        let stopEvent = null;
+        for (const event of events) {
+            if (event.event_type === "stop") {
+                stopEvent = event;
+                continue;
+            }
+            for (const item of event.content_items) {
+                if (item.type.endsWith(".done")) {
+                    contentItems.push(item);
+                }
+            }
+        }
+        return {
+            role: "assistant",
+            content_items: contentItems,
+            usage_metadata: stopEvent?.usage_metadata ?? null,
+            finish_reason: stopEvent?.finish_reason ?? null,
+            created_at: stopEvent?.created_at,
+        };
+    }
+    /**
+     * Generate content in streaming mode (stateless).
+     *
+     * @param options - Object containing messages and config
+     * @yields Delta events, each carrying one delta or done item, then exactly one stop event
+     *   carrying the usage and the finish reason
+     */
+    async *streamingResponse(options) {
+        const { messages, config } = options;
+        // Stamp any messages that don't yet have a created_at timestamp
+        for (const msg of messages) {
+            if (msg.created_at == null) {
+                msg.created_at = Date.now();
+            }
+        }
+        const requestMessages = (0, legacy_1.normalizeLegacyMessages)(messages);
+        const output = new PublicStream(this.constructor.name);
+        for await (const event of this._streamingResponseInternal({
+            messages: requestMessages,
+            config,
+            signal: options.signal,
+        })) {
+            yield* output.push(event);
+        }
+        yield* output.end();
+        const stopEvent = output.stop();
+        // saved before the stop is yielded: a caller may stop iterating as soon as it sees it
+        if (config.trace_id) {
+            const { Tracer } = await Promise.resolve().then(() => __importStar(require("./integration/tracer")));
+            const assistantMessage = {
+                role: "assistant",
+                content_items: output.doneItems,
+                usage_metadata: stopEvent.usage_metadata,
+                finish_reason: stopEvent.finish_reason,
+                created_at: stopEvent.created_at,
+            };
+            const tracer = new Tracer();
+            tracer.saveHistory(this._model, [...requestMessages, assistantMessage], config.trace_id, config);
+        }
+        yield stopEvent;
+    }
+    /**
+     * Generate content in streaming mode (stateful).
+     *
+     * @param message - Latest universal message object to add to conversation
+     * @param config - Universal configuration object
+     * @yields Universal events from the streaming response
+     */
+    async *streamingResponseStateful(options) {
+        const { config } = options;
+        const [message] = (0, legacy_1.normalizeLegacyMessages)([options.message]);
+        const events = [];
+        for await (const event of this.streamingResponse({
+            messages: [...this._history, message],
+            config,
+            signal: options.signal,
+        })) {
+            events.push(event);
+            if (event.event_type === "stop") {
+                // recorded before the stop is yielded: a caller may stop iterating as soon as it sees it
+                this._history.push(message);
+                this._history.push(this.concatUniEventsToUniMessage(events));
+            }
+            yield event;
+        }
+    }
+    /**
+     * Clear the message history.
+     */
+    clearHistory() {
+        this._history = [];
+    }
+    /**
+     * Get the current message history.
+     *
+     * @returns Copy of the current message history
+     */
+    getHistory() {
+        return [...this._history];
+    }
+    /**
+     * Replace the message history with a copy of the provided history.
+     *
+     * @param history - List of universal message objects to set as the new history
+     */
+    setHistory(history) {
+        this._history = (0, legacy_1.normalizeLegacyMessages)(history);
+    }
+}
+exports.LLMClient = LLMClient;

package/dist/deepseek_official/client.d.ts

@@ -0,0 +1,60 @@
+import type { ResponseInputItem, ResponseStreamEvent } from "openai/resources/responses/responses";
+import { LLMClient } from "../baseClient";
+import { UniConfig, UniEvent, UniMessage } from "../types";
+/**
+ * DeepSeek V4-specific LLM client implementation using the OpenAI-compatible Responses API.
+ */
+export declare class DeepSeekOfficialClient extends LLMClient {
+    protected _model: string;
+    private _client;
+    /**
+     * Initialize DeepSeek client with model, API key, and base URL.
+     */
+    constructor(options: {
+        model: string;
+        apiKey?: string;
+        baseUrl?: string | null;
+        defaultHeaders?: Record<string, string>;
+    });
+    /**
+     * Convert ThinkingLevel enum to DeepSeek's reasoning effort.
+     *
+     * DeepSeek accepts low/high/max and maps medium and xhigh onto high server-side
+     * (llmsdk_docs/deepseek_v4/docs/thinking-mode.md), so this sends the value the server
+     * would settle on anyway. Effort "none" is what turns thinking off on this endpoint:
+     * the Chat Completions `thinking` toggle is ignored here (verified live 2026-08-21).
+     */
+    private _convertThinkingLevelToEffort;
+    /**
+     * Convert ToolChoice to DeepSeek's Responses-compatible tool_choice format.
+     */
+    private _convertToolChoice;
+    /**
+     * Transform universal configuration to DeepSeek-specific configuration.
+     */
+    transformUniConfigToModelConfig(config: UniConfig): any;
+    /**
+     * Transform universal message format to DeepSeek's Responses-compatible input format.
+     */
+    transformUniMessageToModelInput(messages: UniMessage[], _signal?: AbortSignal): ResponseInputItem[];
+    /**
+     * Transform one DeepSeek stream event into a universal event, identifying items by output
+     * item id. An item needs no done: it is done when the next one begins or the stream ends.
+     */
+    transformModelOutputToUniEvent(modelOutput: ResponseStreamEvent): UniEvent;
+    /**
+     * Stream generate using DeepSeek's OpenAI-compatible Responses API.
+     */
+    _streamingResponseInternal(options: {
+        messages: UniMessage[];
+        config: UniConfig;
+        signal?: AbortSignal;
+    }): AsyncGenerator<UniEvent>;
+    /**
+     * List the model ids the configured endpoint serves.
+     *
+     * @returns The model ids, in the order the endpoint returned them.
+     */
+    listModels(): Promise<string[]>;
+}
+//# sourceMappingURL=client.d.ts.map

package/dist/deepseek_official/client.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.d.ts","sourceRoot":"","sources":["../../src/deepseek_official/client.ts"],"names":[],"mappings":"AAeA,OAAO,KAAK,EACV,iBAAiB,EACjB,mBAAmB,EAEpB,MAAM,sCAAsC,CAAC;AAC9C,OAAO,EAAE,SAAS,EAAE,MAAM,eAAe,CAAC;AAE1C,OAAO,EAOL,SAAS,EACT,QAAQ,EACR,UAAU,EAEX,MAAM,UAAU,CAAC;AAYlB;;GAEG;AACH,qBAAa,sBAAuB,SAAQ,SAAS;IACnD,SAAS,CAAC,MAAM,EAAE,MAAM,CAAC;IACzB,OAAO,CAAC,OAAO,CAAS;IAExB;;OAEG;gBACS,OAAO,EAAE;QACnB,KAAK,EAAE,MAAM,CAAC;QACd,MAAM,CAAC,EAAE,MAAM,CAAC;QAChB,OAAO,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;QACxB,cAAc,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KACzC;IAsBD;;;;;;;OAOG;IACH,OAAO,CAAC,6BAA6B;IAYrC;;OAEG;IACH,OAAO,CAAC,kBAAkB;IAY1B;;OAEG;IAEH,+BAA+B,CAAC,MAAM,EAAE,SAAS,GAAG,GAAG;IAwEvD;;OAEG;IACH,+BAA+B,CAC7B,QAAQ,EAAE,UAAU,EAAE,EACtB,OAAO,CAAC,EAAE,WAAW,GACpB,iBAAiB,EAAE;IA2HtB;;;OAGG;IACH,8BAA8B,CAAC,WAAW,EAAE,mBAAmB,GAAG,QAAQ;IA8G1E;;OAEG;IACI,0BAA0B,CAAC,OAAO,EAAE;QACzC,QAAQ,EAAE,UAAU,EAAE,CAAC;QACvB,MAAM,EAAE,SAAS,CAAC;QAClB,MAAM,CAAC,EAAE,WAAW,CAAC;KACtB,GAAG,cAAc,CAAC,QAAQ,CAAC;IAqB5B;;;;OAIG;IACG,UAAU,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;CAQtC"}