@vybestack/llxprt-code-core 0.1.23 → 0.2.2

package/dist/src/services/history/HistoryService.js

@@ -0,0 +1,673 @@
+/**
+ * Copyright 2025 Vybestack LLC
+ *
+ * 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.
+ */
+import { ContentValidation, } from './IContent.js';
+import { EventEmitter } from 'events';
+import { OpenAITokenizer } from '../../providers/tokenizers/OpenAITokenizer.js';
+import { AnthropicTokenizer } from '../../providers/tokenizers/AnthropicTokenizer.js';
+import { DebugLogger } from '../../debug/index.js';
+import { randomUUID } from 'crypto';
+/**
+ * Service for managing conversation history in a provider-agnostic way.
+ * All history is stored as IContent. Providers are responsible for converting
+ * to/from their own formats.
+ */
+export class HistoryService extends EventEmitter {
+    history = [];
+    totalTokens = 0;
+    tokenizerCache = new Map();
+    tokenizerLock = Promise.resolve();
+    logger = new DebugLogger('llxprt:history:service');
+    // Compression state and queue
+    isCompressing = false;
+    pendingOperations = [];
+    /**
+     * Get or create tokenizer for a specific model
+     */
+    getTokenizerForModel(modelName) {
+        if (this.tokenizerCache.has(modelName)) {
+            return this.tokenizerCache.get(modelName);
+        }
+        let tokenizer;
+        if (modelName.includes('claude') || modelName.includes('anthropic')) {
+            tokenizer = new AnthropicTokenizer();
+        }
+        else if (modelName.includes('gpt') ||
+            modelName.includes('openai') ||
+            modelName.includes('o1') ||
+            modelName.includes('o3')) {
+            tokenizer = new OpenAITokenizer();
+        }
+        else {
+            // Default to OpenAI tokenizer for Gemini and other models (tiktoken is pretty universal)
+            tokenizer = new OpenAITokenizer();
+        }
+        this.tokenizerCache.set(modelName, tokenizer);
+        return tokenizer;
+    }
+    /**
+     * Generate a new normalized history tool ID.
+     * Format: hist_tool_<uuid-v4>
+     */
+    generateHistoryId() {
+        return `hist_tool_${randomUUID()}`;
+    }
+    /**
+     * Get a callback suitable for passing into converters
+     * which will generate normalized history IDs on demand.
+     */
+    getIdGeneratorCallback() {
+        return () => this.generateHistoryId();
+    }
+    /**
+     * Get the current total token count
+     */
+    getTotalTokens() {
+        return this.totalTokens;
+    }
+    /**
+     * Add content to the history
+     * Note: We accept all content including empty responses for comprehensive history.
+     * Filtering happens only when getting curated history.
+     */
+    add(content, modelName) {
+        // If compression is active, queue this operation
+        if (this.isCompressing) {
+            this.logger.debug('Queueing add operation during compression', {
+                speaker: content.speaker,
+                blockTypes: content.blocks?.map((b) => b.type),
+            });
+            this.pendingOperations.push(() => {
+                this.addInternal(content, modelName);
+            });
+            return;
+        }
+        // Otherwise, add immediately
+        this.addInternal(content, modelName);
+    }
+    addInternal(content, modelName) {
+        // Log content being added with any tool call/response IDs
+        this.logger.debug('Adding content to history:', {
+            speaker: content.speaker,
+            blockTypes: content.blocks?.map((b) => b.type),
+            toolCallIds: content.blocks
+                ?.filter((b) => b.type === 'tool_call')
+                .map((b) => b.id),
+            toolResponseIds: content.blocks
+                ?.filter((b) => b.type === 'tool_response')
+                .map((b) => ({
+                callId: b.callId,
+                toolName: b.toolName,
+            })),
+            contentId: content.metadata?.id,
+            modelName,
+        });
+        // Only do basic validation - must have valid speaker
+        if (content.speaker && ['human', 'ai', 'tool'].includes(content.speaker)) {
+            this.history.push(content);
+            this.logger.debug('Content added successfully, history length:', this.history.length);
+            // Update token count asynchronously but atomically
+            this.updateTokenCount(content, modelName);
+        }
+        else {
+            this.logger.debug('Content rejected - invalid speaker:', content.speaker);
+        }
+    }
+    /**
+     * Atomically update token count for new content
+     */
+    async updateTokenCount(content, modelName) {
+        // Use a lock to prevent race conditions
+        this.tokenizerLock = this.tokenizerLock.then(async () => {
+            let contentTokens = 0;
+            // First try to use usage data from the content metadata
+            if (content.metadata?.usage) {
+                contentTokens = content.metadata.usage.totalTokens;
+            }
+            else {
+                // Fall back to tokenizer estimation
+                // Default to gpt-4.1 tokenizer if no model name provided (most universal)
+                const defaultModel = modelName || 'gpt-4.1';
+                contentTokens = await this.estimateContentTokens(content, defaultModel);
+            }
+            // Atomically update the total
+            this.totalTokens += contentTokens;
+            // Emit event with updated count
+            const eventData = {
+                totalTokens: this.totalTokens,
+                addedTokens: contentTokens,
+                contentId: content.metadata?.id,
+            };
+            this.logger.debug('Emitting tokensUpdated:', eventData);
+            this.emit('tokensUpdated', eventData);
+        });
+        return this.tokenizerLock;
+    }
+    /**
+     * Estimate token count for content using tokenizer
+     */
+    async estimateContentTokens(content, modelName) {
+        const tokenizer = this.getTokenizerForModel(modelName);
+        let totalTokens = 0;
+        for (const block of content.blocks) {
+            let blockText = '';
+            switch (block.type) {
+                case 'text':
+                    blockText = block.text;
+                    break;
+                case 'tool_call':
+                    try {
+                        blockText = JSON.stringify({
+                            name: block.name,
+                            parameters: block.parameters,
+                        });
+                    }
+                    catch (error) {
+                        // Handle circular references or other JSON.stringify errors
+                        this.logger.debug('Error stringifying tool_call parameters, using fallback:', error);
+                        // Fallback to just the tool name for token estimation
+                        blockText = `tool_call: ${block.name}`;
+                    }
+                    break;
+                case 'tool_response':
+                    try {
+                        blockText = JSON.stringify(block.result || block.error || '');
+                    }
+                    catch (error) {
+                        // Handle circular references or other JSON.stringify errors
+                        this.logger.debug('Error stringifying tool_response result/error, using fallback:', error);
+                        // Fallback to just the tool name for token estimation
+                        blockText = `tool_response: ${block.toolName || 'unknown'}`;
+                    }
+                    break;
+                case 'thinking':
+                    blockText = block.thought;
+                    break;
+                case 'code':
+                    blockText = block.code;
+                    break;
+                case 'media':
+                    // For media, just count the caption if any
+                    blockText = block.caption || '';
+                    break;
+                default:
+                    // Unknown block type, skip
+                    break;
+            }
+            if (blockText) {
+                try {
+                    const blockTokens = await tokenizer.countTokens(blockText, modelName);
+                    totalTokens += blockTokens;
+                }
+                catch (error) {
+                    this.logger.debug('Error counting tokens for block, using fallback:', error);
+                    totalTokens += this.simpleTokenEstimateForText(blockText);
+                }
+            }
+        }
+        return totalTokens;
+    }
+    /**
+     * Simple token estimation for text
+     */
+    simpleTokenEstimateForText(text) {
+        if (!text)
+            return 0;
+        const wordCount = text.split(/\s+/).length;
+        const characterCount = text.length;
+        return Math.round(Math.max(wordCount * 1.3, characterCount / 4));
+    }
+    /**
+     * Add multiple contents to the history
+     */
+    addAll(contents, modelName) {
+        for (const content of contents) {
+            this.add(content, modelName);
+        }
+    }
+    /**
+     * Get all history
+     */
+    getAll() {
+        return [...this.history];
+    }
+    /**
+     * Clear all history
+     */
+    clear() {
+        // If compression is active, queue this operation
+        if (this.isCompressing) {
+            this.logger.debug('Queueing clear operation during compression');
+            this.pendingOperations.push(() => {
+                this.clearInternal();
+            });
+            return;
+        }
+        // Otherwise, clear immediately
+        this.clearInternal();
+    }
+    clearInternal() {
+        this.logger.debug('Clearing history', {
+            previousLength: this.history.length,
+        });
+        const previousTokens = this.totalTokens;
+        this.history = [];
+        this.totalTokens = 0;
+        // Emit event with reset count
+        this.emit('tokensUpdated', {
+            totalTokens: 0,
+            addedTokens: -previousTokens, // Negative to indicate removal
+            contentId: null,
+        });
+    }
+    /**
+     * Get the last N messages from history
+     */
+    getRecent(count) {
+        return this.history.slice(-count);
+    }
+    /**
+     * Get curated history (only valid, meaningful content)
+     * Matches the behavior of extractCuratedHistory in geminiChat.ts:
+     * - Always includes user/human messages
+     * - Always includes tool messages
+     * - Only includes AI messages if they are valid (have content)
+     */
+    getCurated() {
+        // Wait if compression is in progress
+        if (this.isCompressing) {
+            this.logger.debug('getCurated called during compression - returning snapshot');
+        }
+        // Build the curated list without modifying history
+        const curated = [];
+        let excludedCount = 0;
+        for (const content of this.history) {
+            if (content.speaker === 'human' || content.speaker === 'tool') {
+                // Always include user and tool messages
+                curated.push(content);
+            }
+            else if (content.speaker === 'ai') {
+                // Only include AI messages if they have valid content
+                if (ContentValidation.hasContent(content)) {
+                    curated.push(content);
+                }
+                else {
+                    excludedCount++;
+                    this.logger.debug('Excluding AI content without valid content:', {
+                        blocks: content.blocks?.map((b) => ({
+                            type: b.type,
+                            hasContent: b.type === 'text' ? !!b.text : true,
+                        })),
+                    });
+                }
+            }
+        }
+        this.logger.debug('Curated history summary:', {
+            totalHistory: this.history.length,
+            curatedCount: curated.length,
+            excludedAiCount: excludedCount,
+            toolCallsInCurated: curated.reduce((acc, c) => acc + c.blocks.filter((b) => b.type === 'tool_call').length, 0),
+            toolResponsesInCurated: curated.reduce((acc, c) => acc + c.blocks.filter((b) => b.type === 'tool_response').length, 0),
+            isCompressing: this.isCompressing,
+        });
+        return curated;
+    }
+    /**
+     * Get comprehensive history (all content including invalid/empty)
+     */
+    getComprehensive() {
+        return [...this.history];
+    }
+    /**
+     * Remove the last content if it matches the provided content
+     */
+    removeLastIfMatches(content) {
+        const last = this.history[this.history.length - 1];
+        if (last === content) {
+            this.history.pop();
+            return true;
+        }
+        return false;
+    }
+    /**
+     * Pop the last content from history
+     */
+    pop() {
+        const removed = this.history.pop();
+        if (removed) {
+            // Recalculate tokens since we removed content
+            // This is less efficient but ensures accuracy
+            this.recalculateTokens();
+        }
+        return removed;
+    }
+    /**
+     * Recalculate total tokens from scratch
+     * Use this when removing content or when token counts might be stale
+     */
+    async recalculateTokens(defaultModel = 'gpt-4.1') {
+        this.tokenizerLock = this.tokenizerLock.then(async () => {
+            let newTotal = 0;
+            for (const content of this.history) {
+                if (content.metadata?.usage) {
+                    newTotal += content.metadata.usage.totalTokens;
+                }
+                else {
+                    // Use the model from content metadata, or fall back to provided default
+                    const modelToUse = content.metadata?.model || defaultModel;
+                    newTotal += await this.estimateContentTokens(content, modelToUse);
+                }
+            }
+            const oldTotal = this.totalTokens;
+            this.totalTokens = newTotal;
+            // Emit event with updated count
+            this.emit('tokensUpdated', {
+                totalTokens: this.totalTokens,
+                addedTokens: this.totalTokens - oldTotal,
+                contentId: null,
+            });
+        });
+        return this.tokenizerLock;
+    }
+    /**
+     * Get the last user (human) content
+     */
+    getLastUserContent() {
+        for (let i = this.history.length - 1; i >= 0; i--) {
+            if (this.history[i].speaker === 'human') {
+                return this.history[i];
+            }
+        }
+        return undefined;
+    }
+    /**
+     * Get the last AI content
+     */
+    getLastAIContent() {
+        for (let i = this.history.length - 1; i >= 0; i--) {
+            if (this.history[i].speaker === 'ai') {
+                return this.history[i];
+            }
+        }
+        return undefined;
+    }
+    /**
+     * Record a complete turn (user input + AI response + optional tool interactions)
+     */
+    recordTurn(userInput, aiResponse, toolInteractions) {
+        this.add(userInput);
+        this.add(aiResponse);
+        if (toolInteractions) {
+            this.addAll(toolInteractions);
+        }
+    }
+    /**
+     * Get the number of messages in history
+     */
+    length() {
+        return this.history.length;
+    }
+    /**
+     * Check if history is empty
+     */
+    isEmpty() {
+        return this.history.length === 0;
+    }
+    /**
+     * Clone the history (deep copy)
+     */
+    clone() {
+        return JSON.parse(JSON.stringify(this.history));
+    }
+    /**
+     * Find unmatched tool calls (tool calls without responses)
+     */
+    findUnmatchedToolCalls() {
+        // With atomic tool call/response implementation, orphans are impossible by design
+        // Always return empty array since orphans cannot exist
+        this.logger.debug('No unmatched tool calls - atomic implementation prevents orphans');
+        return [];
+    }
+    /**
+     * Validate and fix the history to ensure proper tool call/response pairing
+     */
+    validateAndFix() {
+        // With atomic tool call/response implementation, the history is always valid by design
+        // No fixing needed since orphans cannot exist
+        this.logger.debug('History validation skipped - atomic implementation ensures validity');
+    }
+    /**
+     * Get curated history with circular references removed for providers.
+     * This ensures the history can be safely serialized and sent to providers.
+     */
+    getCuratedForProvider() {
+        // Get the curated history
+        const curated = this.getCurated();
+        // Deep clone to avoid circular references in tool call parameters
+        // We need a clean copy that can be serialized
+        return this.deepCloneWithoutCircularRefs(curated);
+    }
+    /**
+     * Deep clone content array, removing circular references
+     */
+    deepCloneWithoutCircularRefs(contents) {
+        return contents.map((content) => {
+            // Create a clean copy of the content
+            const cloned = {
+                speaker: content.speaker,
+                blocks: content.blocks.map((block) => {
+                    if (block.type === 'tool_call') {
+                        const toolCall = block;
+                        // For tool calls, sanitize the parameters to remove circular refs
+                        return {
+                            type: 'tool_call',
+                            id: toolCall.id,
+                            name: toolCall.name,
+                            parameters: this.sanitizeParams(toolCall.parameters),
+                        };
+                    }
+                    else if (block.type === 'tool_response') {
+                        const toolResponse = block;
+                        // For tool responses, sanitize the result to remove circular refs
+                        return {
+                            type: 'tool_response',
+                            callId: toolResponse.callId,
+                            toolName: toolResponse.toolName,
+                            result: this.sanitizeParams(toolResponse.result),
+                            error: toolResponse.error,
+                        };
+                    }
+                    else {
+                        // Other blocks should be safe to clone
+                        try {
+                            return JSON.parse(JSON.stringify(block));
+                        }
+                        catch {
+                            // If any block fails, return minimal version
+                            return { ...block };
+                        }
+                    }
+                }),
+                metadata: content.metadata ? { ...content.metadata } : {},
+            };
+            return cloned;
+        });
+    }
+    /**
+     * Sanitize parameters to remove circular references
+     */
+    sanitizeParams(params) {
+        const seen = new WeakSet();
+        const sanitize = (obj) => {
+            // Handle primitives
+            if (obj === null || typeof obj !== 'object') {
+                return obj;
+            }
+            // Check for circular reference
+            if (seen.has(obj)) {
+                return { _circular: true };
+            }
+            seen.add(obj);
+            // Handle arrays
+            if (Array.isArray(obj)) {
+                return obj.map((item) => sanitize(item));
+            }
+            // Handle objects
+            const result = {};
+            for (const [key, value] of Object.entries(obj)) {
+                result[key] = sanitize(value);
+            }
+            return result;
+        };
+        try {
+            return sanitize(params);
+        }
+        catch (error) {
+            this.logger.debug('Error sanitizing params:', error);
+            return {
+                _note: 'Parameters contained circular references and were sanitized',
+            };
+        }
+    }
+    /**
+     * Merge two histories, handling duplicates and conflicts
+     */
+    merge(other) {
+        // Simple append for now - could be made smarter to detect duplicates
+        this.addAll(other.getAll());
+    }
+    /**
+     * Get history within a token limit (for context window management)
+     */
+    getWithinTokenLimit(maxTokens, countTokensFn) {
+        const result = [];
+        let totalTokens = 0;
+        // Work backwards to keep most recent messages
+        for (let i = this.history.length - 1; i >= 0; i--) {
+            const content = this.history[i];
+            const tokens = countTokensFn(content);
+            if (totalTokens + tokens <= maxTokens) {
+                result.unshift(content);
+                totalTokens += tokens;
+            }
+            else {
+                break;
+            }
+        }
+        return result;
+    }
+    /**
+     * Summarize older history to fit within token limits
+     */
+    async summarizeOldHistory(keepRecentCount, summarizeFn) {
+        if (this.history.length <= keepRecentCount) {
+            return;
+        }
+        const toSummarize = this.history.slice(0, -keepRecentCount);
+        const toKeep = this.history.slice(-keepRecentCount);
+        const summary = await summarizeFn(toSummarize);
+        this.history = [summary, ...toKeep];
+    }
+    /**
+     * Export history to JSON
+     */
+    toJSON() {
+        return JSON.stringify(this.history, null, 2);
+    }
+    /**
+     * Import history from JSON
+     */
+    static fromJSON(json) {
+        const service = new HistoryService();
+        const history = JSON.parse(json);
+        service.addAll(history);
+        return service;
+    }
+    /**
+     * Mark compression as starting
+     * This will cause add() operations to queue until compression completes
+     */
+    startCompression() {
+        this.logger.debug('Starting compression - locking history');
+        this.isCompressing = true;
+    }
+    /**
+     * Mark compression as complete
+     * This will flush all queued operations
+     */
+    endCompression() {
+        this.logger.debug('Compression complete - unlocking history', {
+            pendingCount: this.pendingOperations.length,
+        });
+        this.isCompressing = false;
+        // Flush all pending operations
+        const operations = this.pendingOperations;
+        this.pendingOperations = [];
+        for (const operation of operations) {
+            operation();
+        }
+        this.logger.debug('Flushed pending operations', {
+            count: operations.length,
+        });
+    }
+    /**
+     * Wait for all pending operations to complete
+     * For synchronous operations, this is now a no-op but kept for API compatibility
+     */
+    async waitForPendingOperations() {
+        // Since operations are now synchronous, nothing to wait for
+        return Promise.resolve();
+    }
+    /**
+     * Get conversation statistics
+     */
+    getStatistics() {
+        let userMessages = 0;
+        let aiMessages = 0;
+        let toolCalls = 0;
+        let toolResponses = 0;
+        let totalTokens = 0;
+        let hasTokens = false;
+        for (const content of this.history) {
+            if (content.speaker === 'human') {
+                userMessages++;
+            }
+            else if (content.speaker === 'ai') {
+                aiMessages++;
+            }
+            for (const block of content.blocks) {
+                if (block.type === 'tool_call') {
+                    toolCalls++;
+                }
+                else if (block.type === 'tool_response') {
+                    toolResponses++;
+                }
+            }
+            if (content.metadata?.usage) {
+                totalTokens += content.metadata.usage.totalTokens;
+                hasTokens = true;
+            }
+        }
+        return {
+            totalMessages: this.history.length,
+            userMessages,
+            aiMessages,
+            toolCalls,
+            toolResponses,
+            totalTokens: hasTokens ? totalTokens : undefined,
+        };
+    }
+}
+//# sourceMappingURL=HistoryService.js.map