gsd-trae 1.0.0 → 1.0.2

package/refs/vbenchmark/tasks/ai-integration/agents/code-review-agent/PROMPT.md

@@ -1,64 +0,0 @@
-# Task: Automated Code Review Agent
-
-## Objective
-Build an LLM agent that reviews pull requests, identifies issues, and suggests improvements.
-
-## Requirements
-
-1. **PR Analysis**
-   - Fetch PR diff from GitHub
-   - Understand code context
-   - Identify changed files
-   - Parse code structure
-
-2. **Review Categories**
-   - Bugs and logic errors
-   - Security vulnerabilities
-   - Performance issues
-   - Code style violations
-   - Missing tests
-   - Documentation gaps
-
-3. **Review Output**
-   ```json
-   {
-     "summary": "string",
-     "overallScore": "number",
-     "issues": [{
-       "file": "string",
-       "line": "number",
-       "severity": "critical|major|minor|suggestion",
-       "category": "string",
-       "description": "string",
-       "suggestion": "string"
-     }],
-     "positives": ["string"]
-   }
-   ```
-
-4. **GitHub Integration**
-   - Post review comments on PR
-   - Add inline suggestions
-   - Request changes or approve
-   - Update on new commits
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- PyGithub
-- OpenAI/Anthropic
-
-## Files to Create
-- `agent/reviewer.py` - Review agent
-- `agent/analyzers/security.py` - Security checks
-- `agent/analyzers/style.py` - Style checks
-- `github/client.py` - GitHub API
-- `github/commenter.py` - PR commenting
-- `app/routers/webhooks.py` - Webhook handler
-
-## Success Criteria
-- [ ] PR diff analyzed correctly
-- [ ] Issues identified with line numbers
-- [ ] Suggestions are actionable
-- [ ] Comments posted to GitHub
-- [ ] Handles various languages

package/refs/vbenchmark/tasks/ai-integration/agents/code-review-agent/task.yaml

@@ -1,24 +0,0 @@
-name: Automated Code Review Agent
-category: ai-integration
-subcategory: agents
-description: Build LLM agent for automated PR code review
-difficulty: hard
-stack: python
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/test_review_agent.py
-
-tags:
-  - python
-  - agent
-  - code-review
-  - github
-  - llm
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/agents/research-agent/PROMPT.md

@@ -1,61 +0,0 @@
-# Task: Multi-Step Research Agent
-
-## Objective
-Build an LLM agent that performs multi-step research by searching, reading, and synthesizing information.
-
-## Requirements
-
-1. **Research Tools**
-   - Web search (Google/Bing API)
-   - Page content fetcher
-   - PDF reader
-   - Note taking
-   - Citation management
-
-2. **Agent Workflow**
-   - Decompose research question
-   - Search for relevant sources
-   - Read and extract key info
-   - Synthesize findings
-   - Generate report with citations
-
-3. **Output Format**
-   ```typescript
-   interface ResearchReport {
-     question: string;
-     summary: string;
-     keyFindings: string[];
-     sources: Array<{
-       url: string;
-       title: string;
-       relevantQuotes: string[];
-     }>;
-     confidence: number;
-   }
-   ```
-
-4. **API**
-   - `POST /research` - Start research
-   - `GET /research/{id}` - Get status
-   - `GET /research/{id}/report` - Get report
-   - Stream progress updates via SSE
-
-## Technical Stack
-- TypeScript/Node.js
-- LangChain or custom agent
-- Bing/Google Search API
-- OpenAI/Anthropic
-
-## Files to Create
-- `src/agent/researcher.ts` - Main agent
-- `src/agent/tools/search.ts` - Search tool
-- `src/agent/tools/reader.ts` - Content reader
-- `src/agent/synthesizer.ts` - Report generation
-- `src/routes/research.ts` - API endpoints
-
-## Success Criteria
-- [ ] Searches relevant sources
-- [ ] Extracts key information
-- [ ] Synthesizes coherent report
-- [ ] Citations are accurate
-- [ ] Handles multi-step queries

package/refs/vbenchmark/tasks/ai-integration/agents/research-agent/task.yaml

@@ -1,24 +0,0 @@
-name: Multi-Step Research Agent
-category: ai-integration
-subcategory: agents
-description: Build research automation agent in TypeScript
-difficulty: hard
-stack: nodejs
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/research.test.ts
-
-tags:
-  - typescript
-  - agent
-  - research
-  - llm
-  - automation
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/agents/web-scraper-agent/PROMPT.md

@@ -1,57 +0,0 @@
-# Task: Autonomous Web Scraper Agent
-
-## Objective
-Build an LLM-powered agent that autonomously navigates websites and extracts structured data.
-
-## Requirements
-
-1. **Agent Capabilities**
-   - Navigate to URLs
-   - Click elements by description
-   - Fill forms
-   - Extract data from pages
-   - Handle pagination
-
-2. **Tools**
-   - `navigate(url)` - Go to URL
-   - `click(selector/description)` - Click element
-   - `fill(selector, value)` - Fill input
-   - `extract(schema)` - Extract data
-   - `screenshot()` - Capture page
-
-3. **Task Execution**
-   - Accept natural language task
-   - Plan navigation steps
-   - Execute with error recovery
-   - Return structured results
-
-4. **API**
-   - `POST /tasks` - Create scraping task
-   - `GET /tasks/{id}` - Get task status
-   - `GET /tasks/{id}/results` - Get results
-
-## Example Task
-```
-"Go to Amazon, search for 'wireless headphones',
-extract the top 10 products with name, price, and rating"
-```
-
-## Technical Stack
-- Python 3.11+
-- Playwright for browser
-- LangChain or custom agent
-- OpenAI/Anthropic for planning
-
-## Files to Create
-- `agent/core.py` - Agent loop
-- `agent/tools.py` - Agent tools
-- `agent/planner.py` - Task planning
-- `agent/browser.py` - Browser control
-- `app/routers/tasks.py` - API endpoints
-
-## Success Criteria
-- [ ] Agent navigates to target pages
-- [ ] Forms filled correctly
-- [ ] Data extracted matches schema
-- [ ] Handles errors and retries
-- [ ] Pagination handled

package/refs/vbenchmark/tasks/ai-integration/agents/web-scraper-agent/task.yaml

@@ -1,24 +0,0 @@
-name: Autonomous Web Scraper Agent
-category: ai-integration
-subcategory: agents
-description: Build LLM agent for autonomous web data extraction
-difficulty: hard
-stack: python
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/test_scraper_agent.py
-
-tags:
-  - python
-  - agent
-  - scraping
-  - llm
-  - autonomous
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/embeddings/duplicate-detection/PROMPT.md

@@ -1,50 +0,0 @@
-# Task: Document Duplicate Detection
-
-## Objective
-Build a system to find near-duplicate documents using semantic embeddings.
-
-## Requirements
-
-1. **Document Processing**
-   - Accept various formats (text, PDF, HTML)
-   - Normalize text content
-   - Generate embeddings
-   - Store document fingerprints
-
-2. **Duplicate Detection**
-   - Exact duplicates (hash)
-   - Near duplicates (embedding similarity)
-   - Configurable threshold
-   - Cluster duplicates together
-
-3. **Batch Processing**
-   - Process large document sets
-   - Incremental updates
-   - Compare new vs existing
-   - Parallel processing
-
-4. **API**
-   - `POST /documents` - Add document
-   - `POST /check` - Check single doc for dupes
-   - `POST /batch/analyze` - Analyze batch
-   - `GET /clusters` - Get duplicate clusters
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- sentence-transformers
-- FAISS or pgvector
-
-## Files to Create
-- `app/services/processor.py` - Document processing
-- `app/services/embedder.py` - Embedding generation
-- `app/services/detector.py` - Duplicate detection
-- `app/services/clusterer.py` - Clustering logic
-- `app/routers/documents.py` - API endpoints
-
-## Success Criteria
-- [ ] Exact duplicates found
-- [ ] Near duplicates detected
-- [ ] Threshold configurable
-- [ ] Clusters make sense
-- [ ] Handles large batches

package/refs/vbenchmark/tasks/ai-integration/embeddings/duplicate-detection/task.yaml

@@ -1,24 +0,0 @@
-name: Document Duplicate Detection
-category: ai-integration
-subcategory: embeddings
-description: Find near-duplicate documents using embeddings
-difficulty: medium
-stack: python
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/test_duplicates.py
-
-tags:
-  - python
-  - embeddings
-  - deduplication
-  - similarity
-  - nlp
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/embeddings/recommendation-engine/PROMPT.md

@@ -1,51 +0,0 @@
-# Task: Content Recommendation Engine
-
-## Objective
-Build a recommendation engine using embeddings to suggest similar content to users.
-
-## Requirements
-
-1. **Content Indexing**
-   - Index content items (articles, products)
-   - Generate embeddings for content
-   - Update embeddings on content change
-   - Handle content metadata
-
-2. **User Modeling**
-   - Track user interactions
-   - Build user preference vector
-   - Combine viewed/liked item embeddings
-   - Decay old interactions
-
-3. **Recommendations**
-   - Similar to current item
-   - Based on user history
-   - Trending + personalized mix
-   - Filter already seen
-
-4. **API**
-   - `POST /items` - Index item
-   - `GET /recommend/similar/{id}` - Similar items
-   - `GET /recommend/user/{id}` - User recommendations
-   - `POST /interactions` - Record interaction
-
-## Technical Stack
-- Go 1.21+
-- OpenAI Embeddings API
-- Redis for user vectors
-- pgvector for content
-
-## Files to Create
-- `cmd/recommender/main.go` - Entry point
-- `internal/embeddings/client.go` - Embedding API
-- `internal/recommender/similar.go` - Similar items
-- `internal/recommender/personal.go` - Personalized
-- `internal/user/profile.go` - User modeling
-- `internal/handlers/api.go` - HTTP handlers
-
-## Success Criteria
-- [ ] Content indexed with embeddings
-- [ ] Similar items are relevant
-- [ ] User preferences captured
-- [ ] Personalized recs improve
-- [ ] Already seen items filtered

package/refs/vbenchmark/tasks/ai-integration/embeddings/recommendation-engine/task.yaml

@@ -1,24 +0,0 @@
-name: Content Recommendation Engine
-category: ai-integration
-subcategory: embeddings
-description: Build content recommendations with embeddings in Go
-difficulty: medium
-stack: go
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/recommend_test.go
-
-tags:
-  - go
-  - embeddings
-  - recommendations
-  - similarity
-  - personalization
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/embeddings/semantic-search/PROMPT.md

@@ -1,50 +0,0 @@
-# Task: Vector Similarity Search
-
-## Objective
-Build a semantic search system using vector embeddings for finding similar content.
-
-## Requirements
-
-1. **Embedding Generation**
-   - Text to embedding conversion
-   - Batch processing
-   - Caching embeddings
-   - Model selection (OpenAI, sentence-transformers)
-
-2. **Vector Storage**
-   - Store embeddings with metadata
-   - Efficient similarity search
-   - Support filtering
-   - Handle updates/deletes
-
-3. **Search Features**
-   - Semantic similarity search
-   - Hybrid (semantic + keyword)
-   - Top-K retrieval
-   - Similarity threshold
-
-4. **API**
-   - `POST /embed` - Generate embedding
-   - `POST /index` - Add to index
-   - `POST /search` - Similarity search
-   - `DELETE /index/{id}` - Remove item
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- OpenAI Embeddings or sentence-transformers
-- pgvector or FAISS
-
-## Files to Create
-- `app/services/embedder.py` - Embedding generation
-- `app/services/vector_store.py` - Vector storage
-- `app/services/searcher.py` - Search logic
-- `app/routers/search.py` - API endpoints
-- `app/models/document.py` - Data models
-
-## Success Criteria
-- [ ] Embeddings generated correctly
-- [ ] Similar items retrieved
-- [ ] Filters work properly
-- [ ] Performance acceptable
-- [ ] Handles large datasets

package/refs/vbenchmark/tasks/ai-integration/embeddings/semantic-search/task.yaml

@@ -1,24 +0,0 @@
-name: Vector Similarity Search
-category: ai-integration
-subcategory: embeddings
-description: Build semantic search with vector embeddings in Python
-difficulty: medium
-stack: python
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/test_semantic_search.py
-
-tags:
-  - python
-  - embeddings
-  - vector
-  - search
-  - similarity
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/fine-tuning/classification-model/PROMPT.md

@@ -1,50 +0,0 @@
-# Task: Text Classification Fine-Tuning
-
-## Objective
-Build a pipeline to fine-tune a language model for multi-class text classification.
-
-## Requirements
-
-1. **Data Preparation**
-   - Load training data (CSV/JSON)
-   - Train/validation split
-   - Text preprocessing
-   - Label encoding
-
-2. **Fine-Tuning**
-   - Use base model (BERT, DistilBERT)
-   - Add classification head
-   - Configure training params
-   - Track metrics (accuracy, F1)
-
-3. **Training Pipeline**
-   - Data loaders with batching
-   - Learning rate scheduling
-   - Early stopping
-   - Checkpointing
-
-4. **API**
-   - `POST /train` - Start training
-   - `GET /train/{id}/status` - Training status
-   - `POST /predict` - Run inference
-   - `GET /models` - List trained models
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- Hugging Face Transformers
-- PyTorch
-
-## Files to Create
-- `training/data.py` - Data loading
-- `training/model.py` - Model setup
-- `training/trainer.py` - Training loop
-- `inference/predictor.py` - Inference
-- `app/routers/train.py` - API endpoints
-
-## Success Criteria
-- [ ] Model fine-tunes successfully
-- [ ] Validation metrics tracked
-- [ ] Checkpoints saved
-- [ ] Inference works correctly
-- [ ] Multiple labels supported

package/refs/vbenchmark/tasks/ai-integration/fine-tuning/classification-model/task.yaml

@@ -1,24 +0,0 @@
-name: Text Classification Fine-Tuning
-category: ai-integration
-subcategory: fine-tuning
-description: Fine-tune model for text classification in Python
-difficulty: hard
-stack: python
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/test_classification.py
-
-tags:
-  - python
-  - fine-tuning
-  - classification
-  - transformers
-  - ml
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/function-calling/api-orchestrator/PROMPT.md

@@ -1,60 +0,0 @@
-# Task: LLM-Driven API Orchestrator
-
-## Objective
-Build a system where LLM orchestrates multiple API calls to complete complex tasks using function calling.
-
-## Requirements
-
-1. **API Registry**
-   - Register available APIs
-   - Define function schemas
-   - Authentication handling
-   - Rate limit awareness
-
-2. **Function Definitions**
-   ```typescript
-   const functions = [
-     {
-       name: "get_user",
-       description: "Get user by ID",
-       parameters: { userId: "string" }
-     },
-     {
-       name: "send_email",
-       description: "Send email to user",
-       parameters: { to: "string", subject: "string", body: "string" }
-     },
-     // More functions...
-   ];
-   ```
-
-3. **Orchestration**
-   - Accept natural language request
-   - LLM determines needed API calls
-   - Execute calls in correct order
-   - Handle dependencies between calls
-   - Aggregate results
-
-4. **API**
-   - `POST /execute` - Execute task
-   - `GET /functions` - List available functions
-   - `POST /functions` - Register new function
-
-## Technical Stack
-- TypeScript/Node.js
-- OpenAI function calling
-- Express or Hono
-
-## Files to Create
-- `src/orchestrator/engine.ts` - Main orchestrator
-- `src/orchestrator/executor.ts` - Function executor
-- `src/registry/functions.ts` - Function registry
-- `src/registry/schemas.ts` - JSON schemas
-- `src/routes/execute.ts` - API endpoints
-
-## Success Criteria
-- [ ] LLM selects correct functions
-- [ ] Functions called in right order
-- [ ] Dependencies handled
-- [ ] Results aggregated correctly
-- [ ] Errors handled gracefully

package/refs/vbenchmark/tasks/ai-integration/function-calling/api-orchestrator/task.yaml

@@ -1,24 +0,0 @@
-name: LLM-Driven API Orchestrator
-category: ai-integration
-subcategory: function-calling
-description: Build API orchestration with LLM function calling
-difficulty: medium
-stack: nodejs
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/orchestrator.test.ts
-
-tags:
-  - typescript
-  - function-calling
-  - api
-  - llm
-  - orchestration
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/function-calling/calendar-assistant/PROMPT.md

@@ -1,50 +0,0 @@
-# Task: AI Calendar Assistant
-
-## Objective
-Build an AI assistant that manages calendar events using natural language and function calling.
-
-## Requirements
-
-1. **Calendar Functions**
-   - `list_events(start, end)` - Get events
-   - `create_event(title, start, end, attendees)` - Create event
-   - `update_event(id, changes)` - Modify event
-   - `delete_event(id)` - Remove event
-   - `find_free_time(duration, attendees)` - Find availability
-
-2. **Natural Language**
-   - "Schedule a meeting with John tomorrow at 2pm"
-   - "What's on my calendar next week?"
-   - "Move my 3pm to 4pm"
-   - "Find a 30-minute slot with Sarah this week"
-
-3. **Smart Features**
-   - Timezone handling
-   - Conflict detection
-   - Recurring events
-   - Attendee availability
-
-4. **API**
-   - `POST /assistant` - Send message
-   - `GET /events` - List events
-   - `POST /events` - Create event
-   - Maintain conversation context
-
-## Technical Stack
-- TypeScript/Node.js
-- Google Calendar API (or mock)
-- OpenAI function calling
-
-## Files to Create
-- `src/assistant/agent.ts` - Main assistant
-- `src/assistant/functions.ts` - Function definitions
-- `src/calendar/client.ts` - Calendar API client
-- `src/calendar/availability.ts` - Availability logic
-- `src/routes/assistant.ts` - API endpoints
-
-## Success Criteria
-- [ ] Events created correctly
-- [ ] Natural language understood
-- [ ] Conflicts detected
-- [ ] Free time found correctly
-- [ ] Timezone handled