gsd-trae 1.0.1 → 1.0.2

package/refs/vbenchmark/tasks/ai-integration/function-calling/calendar-assistant/task.yaml

@@ -1,24 +0,0 @@
-name: AI Calendar Assistant
-category: ai-integration
-subcategory: function-calling
-description: Build calendar management assistant with TypeScript
-difficulty: medium
-stack: nodejs
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/calendar.test.ts
-
-tags:
-  - typescript
-  - calendar
-  - function-calling
-  - llm
-  - scheduling
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/function-calling/database-query/PROMPT.md

@@ -1,62 +0,0 @@
-# Task: Natural Language to SQL
-
-## Objective
-Build a system that converts natural language questions into SQL queries and executes them safely.
-
-## Requirements
-
-1. **Schema Awareness**
-   - Ingest database schema
-   - Understand table relationships
-   - Know column types
-   - Maintain schema cache
-
-2. **Query Generation**
-   - Natural language to SQL
-   - Support SELECT queries
-   - Handle JOINs
-   - Aggregations (COUNT, SUM, AVG)
-   - Filters and sorting
-
-3. **Safety**
-   - Read-only queries only
-   - Query validation
-   - Injection prevention
-   - Result size limits
-
-4. **API**
-   - `POST /schema` - Upload schema
-   - `POST /query` - Ask question
-   - Response includes SQL and results
-
-## Example
-```
-User: "How many orders did each customer place last month?"
-
-SQL: SELECT c.name, COUNT(o.id) as order_count
-     FROM customers c
-     JOIN orders o ON c.id = o.customer_id
-     WHERE o.created_at >= '2025-12-01'
-     GROUP BY c.id, c.name
-     ORDER BY order_count DESC
-```
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- SQLAlchemy
-- OpenAI/Anthropic
-
-## Files to Create
-- `app/services/schema_parser.py` - Schema ingestion
-- `app/services/query_generator.py` - NL to SQL
-- `app/services/query_executor.py` - Safe execution
-- `app/services/validator.py` - Query validation
-- `app/routers/query.py` - API endpoints
-
-## Success Criteria
-- [ ] Simple queries generated correctly
-- [ ] JOINs work properly
-- [ ] Aggregations correct
-- [ ] Only SELECT allowed
-- [ ] Results returned accurately

package/refs/vbenchmark/tasks/ai-integration/function-calling/database-query/task.yaml

@@ -1,24 +0,0 @@
-name: Natural Language to SQL
-category: ai-integration
-subcategory: function-calling
-description: Convert natural language to SQL queries with Python
-difficulty: medium
-stack: python
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/test_nl2sql.py
-
-tags:
-  - python
-  - sql
-  - llm
-  - nlp
-  - database
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/multimodal/chart-interpreter/PROMPT.md

@@ -1,60 +0,0 @@
-# Task: Chart Data Extractor
-
-## Objective
-Build a service that extracts underlying data from chart images using vision LLMs.
-
-## Requirements
-
-1. **Chart Types**
-   - Bar charts
-   - Line charts
-   - Pie charts
-   - Scatter plots
-   - Tables
-
-2. **Data Extraction**
-   ```typescript
-   interface ChartData {
-     title: string;
-     type: 'bar' | 'line' | 'pie' | 'scatter' | 'table';
-     xAxis?: { label: string; values: string[] };
-     yAxis?: { label: string; unit?: string };
-     series: Array<{
-       name: string;
-       data: Array<{ x: string | number; y: number }>;
-     }>;
-     insights: string[];
-   }
-   ```
-
-3. **Features**
-   - Identify chart type
-   - Extract axis labels
-   - Extract data points
-   - Generate insights
-   - Export to CSV/JSON
-
-4. **API**
-   - `POST /extract` - Extract chart data
-   - `POST /analyze` - Analyze and describe
-   - `POST /export` - Export as CSV
-
-## Technical Stack
-- TypeScript/Node.js
-- OpenAI Vision API
-- Express or Hono
-- Sharp for preprocessing
-
-## Files to Create
-- `src/services/vision.ts` - Vision API client
-- `src/services/extractor.ts` - Data extraction
-- `src/services/analyzer.ts` - Chart analysis
-- `src/exporters/csv.ts` - CSV export
-- `src/routes/charts.ts` - API endpoints
-
-## Success Criteria
-- [ ] Chart type identified
-- [ ] Data points extracted accurately
-- [ ] Axis labels captured
-- [ ] CSV export works
-- [ ] Handles various chart styles

package/refs/vbenchmark/tasks/ai-integration/multimodal/chart-interpreter/task.yaml

@@ -1,24 +0,0 @@
-name: Chart Data Extractor
-category: ai-integration
-subcategory: multimodal
-description: Extract data from chart images using vision LLM
-difficulty: hard
-stack: nodejs
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/chart.test.ts
-
-tags:
-  - typescript
-  - vision
-  - charts
-  - data-extraction
-  - multimodal
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/multimodal/image-captioning/PROMPT.md

@@ -1,49 +0,0 @@
-# Task: Image Caption Generator
-
-## Objective
-Build an image captioning service using vision-language models.
-
-## Requirements
-
-1. **Image Input**
-   - Accept image uploads
-   - Accept image URLs
-   - Support JPEG, PNG, WebP
-   - Handle various sizes
-
-2. **Caption Generation**
-   - Descriptive captions
-   - Configurable style (brief, detailed)
-   - Multiple captions per image
-   - Language selection
-
-3. **Additional Features**
-   - Object detection/listing
-   - Scene classification
-   - Text extraction (OCR)
-   - Alt-text optimization
-
-4. **API**
-   - `POST /caption` - Generate caption
-   - `POST /describe` - Detailed description
-   - `POST /objects` - List objects
-   - `POST /batch` - Batch processing
-
-## Technical Stack
-- Python 3.11+
-- FastAPI
-- OpenAI Vision API or BLIP-2
-- Pillow for image processing
-
-## Files to Create
-- `app/services/vision.py` - Vision model client
-- `app/services/captioner.py` - Caption generation
-- `app/services/analyzer.py` - Image analysis
-- `app/routers/images.py` - API endpoints
-
-## Success Criteria
-- [ ] Captions are accurate
-- [ ] Style variations work
-- [ ] Objects detected correctly
-- [ ] Various formats handled
-- [ ] Batch processing works

package/refs/vbenchmark/tasks/ai-integration/multimodal/image-captioning/task.yaml

@@ -1,24 +0,0 @@
-name: Image Caption Generator
-category: ai-integration
-subcategory: multimodal
-description: Generate image descriptions using vision models
-difficulty: medium
-stack: python
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/test_captioning.py
-
-tags:
-  - python
-  - vision
-  - multimodal
-  - llm
-  - images
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/code-assistant/PROMPT.md

@@ -1,51 +0,0 @@
-# Task: Codebase Q&A Assistant
-
-## Objective
-Build a RAG-powered assistant that answers questions about a codebase using semantic code search.
-
-## Requirements
-
-1. **Code Indexing**
-   - Parse source files (TS, JS, Python, Go)
-   - Chunk by functions/classes
-   - Extract docstrings and comments
-   - Store code embeddings
-
-2. **Semantic Search**
-   - Natural language queries
-   - Find relevant code snippets
-   - Rank by relevance
-   - Filter by file type/path
-
-3. **Q&A Interface**
-   - Answer questions about code
-   - Explain function behavior
-   - Find usage examples
-   - Suggest improvements
-
-4. **API**
-   - `POST /index` - Index a repository
-   - `POST /query` - Ask question
-   - `GET /search` - Semantic search
-   - `DELETE /index/{repo}` - Clear index
-
-## Technical Stack
-- TypeScript/Node.js
-- OpenAI Embeddings
-- Pinecone or Qdrant
-- Tree-sitter for parsing
-
-## Files to Create
-- `src/indexer/parser.ts` - Code parsing
-- `src/indexer/chunker.ts` - Code chunking
-- `src/indexer/embedder.ts` - Embedding generation
-- `src/search/retriever.ts` - Retrieval logic
-- `src/qa/assistant.ts` - Q&A generation
-- `src/routes/api.ts` - API endpoints
-
-## Success Criteria
-- [ ] Code indexed correctly
-- [ ] Relevant snippets retrieved
-- [ ] Questions answered accurately
-- [ ] Context includes related code
-- [ ] Handles large codebases

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/code-assistant/task.yaml

@@ -1,24 +0,0 @@
-name: Codebase Q&A Assistant
-category: ai-integration
-subcategory: rag-chatbot
-description: Build RAG-powered codebase Q&A assistant in TypeScript
-difficulty: hard
-stack: nodejs
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/codeqa.test.ts
-
-tags:
-  - typescript
-  - rag
-  - code
-  - embeddings
-  - assistant
-
-evaluation:
-  functional: 0.5
-  quality: 0.25
-  cost: 0.1
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/doc-search/PROMPT.md

@@ -1,51 +0,0 @@
-# Task: Multi-Document Semantic Search
-
-## Objective
-Build a semantic search engine in Go that searches across multiple documents using embeddings.
-
-## Requirements
-
-1. **Document Ingestion**
-   - Accept PDF, TXT, MD files
-   - Chunk documents intelligently
-   - Generate embeddings per chunk
-   - Store metadata (source, page, etc.)
-
-2. **Search Features**
-   - Natural language queries
-   - Hybrid search (semantic + keyword)
-   - Filters (date, source, type)
-   - Pagination and ranking
-
-3. **Results**
-   - Return relevant passages
-   - Highlight matching content
-   - Source attribution
-   - Relevance scores
-
-4. **API**
-   - `POST /documents` - Upload document
-   - `GET /search?q=query` - Search
-   - `DELETE /documents/{id}` - Remove doc
-   - `GET /documents` - List documents
-
-## Technical Stack
-- Go 1.21+
-- OpenAI Embeddings API
-- pgvector or Milvus
-- Chi or Gin router
-
-## Files to Create
-- `cmd/search/main.go` - Entry point
-- `internal/ingestion/processor.go` - Doc processing
-- `internal/ingestion/chunker.go` - Text chunking
-- `internal/embeddings/client.go` - Embedding API
-- `internal/search/engine.go` - Search logic
-- `internal/handlers/api.go` - HTTP handlers
-
-## Success Criteria
-- [ ] Documents indexed correctly
-- [ ] Semantic search returns relevant results
-- [ ] Filters work correctly
-- [ ] Results include source info
-- [ ] Handles large document sets

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/doc-search/task.yaml

@@ -1,24 +0,0 @@
-name: Multi-Document Semantic Search
-category: ai-integration
-subcategory: rag-chatbot
-description: Build semantic search across multiple documents in Go
-difficulty: medium
-stack: go
-timeout: 600
-tokenLimit: 150000
-
-tests:
-  functional: tests/search_test.go
-
-tags:
-  - go
-  - embeddings
-  - search
-  - semantic
-  - vector-db
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/pdf-qa/PROMPT.md

@@ -1,76 +0,0 @@
-# PDF Q&A RAG Chatbot
-
-## Context
-
-Build a Retrieval-Augmented Generation (RAG) chatbot that can answer questions about uploaded PDF documents. The system should extract, chunk, embed, and retrieve relevant content to answer user queries.
-
-## Requirements
-
-### 1. PDF Processing
-
-- Accept PDF file uploads via API
-- Extract text from PDF using pypdf or similar
-- Handle multi-page documents
-- Preserve document structure where possible
-
-### 2. Text Chunking & Embedding
-
-- Split documents into semantic chunks (500-1000 tokens)
-- Generate embeddings using OpenAI text-embedding-ada-002 or similar
-- Store embeddings in a vector database (Chroma, Pinecone, or pgvector)
-
-### 3. Query Processing
-
-- Accept natural language questions via API
-- Retrieve top-k relevant chunks (k=3-5)
-- Generate answers using retrieved context
-- Include source citations in responses
-
-### 4. API Endpoints
-
-```
-POST /documents
-  - Upload PDF file
-  - Returns: { documentId, pageCount, status }
-
-GET /documents/{id}
-  - Get document metadata
-  - Returns: { documentId, name, pageCount, chunkCount }
-
-POST /chat
-  - Body: { documentId, question }
-  - Returns: { answer, sources: [{ page, text }] }
-```
-
-## Technical Stack
-
-- FastAPI for API
-- LangChain for orchestration
-- ChromaDB for vector storage
-- OpenAI for embeddings and completion
-- PyPDF2 for PDF extraction
-
-## Acceptance Criteria
-
-- [ ] PDF upload processes successfully
-- [ ] Questions return relevant, grounded answers
-- [ ] Sources are correctly cited
-- [ ] Handles documents up to 100 pages
-- [ ] Response time < 5 seconds for queries
-
-## Test Document
-
-The fixture `handbook.pdf` is a sample employee handbook. Test questions:
-
-1. "What is the vacation policy?"
-2. "How do I submit expenses?"
-3. "What are the working hours?"
-
-Answers must be grounded in the document content, not hallucinated.
-
-## Environment Variables
-
-```
-OPENAI_API_KEY=your-key
-CHROMA_PERSIST_DIR=./chroma_data
-```

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/pdf-qa/docker-compose.yaml

@@ -1,30 +0,0 @@
-version: '3.8'
-
-services:
-  app:
-    build:
-      context: .
-      dockerfile: Dockerfile
-    ports:
-      - "8000:8000"
-    environment:
-      - OPENAI_API_KEY=${OPENAI_API_KEY}
-      - CHROMA_PERSIST_DIR=/data/chroma
-    volumes:
-      - chroma_data:/data/chroma
-      - ./fixtures:/app/fixtures:ro
-
-  test:
-    build:
-      context: .
-      dockerfile: Dockerfile.test
-    depends_on:
-      - app
-    environment:
-      - BASE_URL=http://app:8000
-    volumes:
-      - ./tests:/app/tests:ro
-      - ./fixtures:/app/fixtures:ro
-
-volumes:
-  chroma_data:

package/refs/vbenchmark/tasks/ai-integration/rag-chatbot/pdf-qa/task.yaml

@@ -1,30 +0,0 @@
-name: PDF Q&A RAG Chatbot
-category: ai-integration
-subcategory: rag-chatbot
-description: Build a RAG chatbot that answers questions from PDF documents
-difficulty: hard
-stack: python
-timeout: 900
-tokenLimit: 200000
-
-tests:
-  functional: tests/functional/qa.test.py
-
-docker:
-  compose: docker-compose.yaml
-
-weight: 2.0
-
-tags:
-  - rag
-  - chatbot
-  - pdf
-  - langchain
-  - embeddings
-  - vector-db
-
-evaluation:
-  functional: 0.5
-  quality: 0.2
-  cost: 0.15
-  speed: 0.15