CSV智能分析套件

cd CSV_MCP

# Start HTTP server
python server.py --http --host 0.0.0.0 --port 8000

# Server now available at:
# http://localhost:8000
# SSE endpoint: http://localhost:8000/sse

cd CSV_MCP
npx @modelcontextprotocol/inspector python server.py

cd CSV_Agent
python main.py "What is the average loan amount?"

# Place CSV files in uploads/ folder
cp mydata.csv uploads/

# Run query (auto-processes new files)
python main.py "Show me all records from mydata"

# Continuously monitor uploads/ folder
python watch_data_folder.py

# Add files to uploads/ - they'll be processed automatically

# Using CSV_MCP tools in Claude or other MCP client

# 1. Upload CSV
upload_csv(
    file_path="d:/data/sales_2024.csv",
    user_id="john_doe",
    table_name="sales"
)

# 2. Query the data
query_data(
    query="What were the total sales by region?",
    user_id="john_doe",
    table_name="sales"
)

# 3. List all tables
list_tables(user_id="john_doe")

# 4. Delete when done
delete_data(
    user_id="john_doe",
    table_name="sales",
    confirm=true
)

from core.orchestrator import Orchestrator

# Initialize
config_path = "CSV_Agent/config.json"
orchestrator = Orchestrator(config_path)

# Upload CSV
result = orchestrator.upload_csv(
    file_path="sales.csv",
    user_id="analyst_1",
    table_name="monthly_sales"
)

# Query
response = orchestrator.process_query(
    query="Show me monthly trends",
    user_id="analyst_1",
    table_name="monthly_sales"
)

print(response['natural_language_response'])
print(response['sql_query'])

┌─────────────────────────────────────────────────────────────┐
│                        MCP Clients                          │
│              (Claude Desktop, Custom Apps)                  │
└────────────────────┬────────────────────────────────────────┘
                     │
                     ▼
┌─────────────────────────────────────────────────────────────┐
│                      CSV_MCP Server                         │
│                  (FastMCP Protocol)                         │
└────────────────────┬────────────────────────────────────────┘
                     │
                     ▼
┌─────────────────────────────────────────────────────────────┐
│                    CSV_Agent System                         │
│                    (Orchestrator)                           │
└─────┬───────────────────────────────────────────────────┬───┘
      │                                                   │
      ▼                                                   ▼
┌─────────────────┐                           ┌──────────────────┐
│  Agent Pipeline │                           │  Data Stores     │
├─────────────────┤                           ├──────────────────┤
│ • Intent        │                           │ • PostgreSQL     │
│ • Schema        │◄──────────────────────────│   (CSV Data)     │
│ • SQL Gen       │                           │                  │
│ • Validation    │                           │ • ChromaDB       │
│ • Execution     │◄──────────────────────────│   (Metadata)     │
│ • Formatting    │                           │                  │
│ • Visualization │                           │ • Redis          │
│ • Caching       │◄──────────────────────────│   (Query Cache)  │
└─────────────────┘                           └──────────────────┘

CSV Upload Flow:
┌──────────┐    ┌──────────────┐    ┌──────────────┐    ┌───────────┐
│   CSV    │───▶│  Validation  │───▶│   Metadata   │───▶│PostgreSQL │
│   File   │    │  & Parsing   │    │  Extraction  │    │  Storage  │
└──────────┘    └──────────────┘    └──────┬───────┘    └───────────┘
                                            │
                                            ▼
                                     ┌──────────────┐
                                     │  ChromaDB    │
                                     │  Indexing    │
                                     └──────────────┘

Query Processing Flow:
┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐
│ Natural  │───▶│  Intent  │───▶│  Schema  │───▶│   SQL    │───▶│  Query   │
│ Language │    │Classifier│    │  Filter  │    │Generator │    │Execution │
└──────────┘    └──────────┘    └──────────┘    └──────────┘    └────┬─────┘
                                                                       │
                   ┌───────────────────────────────────────────────────┘
                   │
                   ▼
┌──────────┐    ┌──────────┐    ┌──────────┐
│ Response │◄───│  Format  │◄───│   Cache  │
│ to User  │    │  Results │    │  Check   │
└──────────┘    └──────────┘    └──────────┘

{
  "agents": {
    "postgres_handler": {
      "params": {
        "db_url": "postgresql://user:pass@host:5432/dbname",
        "schema": "public"
      }
    },
    "query_cache": {
      "params": {
        "redis_host": "localhost",
        "redis_port": 6379,
        "ttl_seconds": 86400,
        "use_fallback": true
      }
    },
    "metadata_indexer": {
      "params": {
        "llm_model": "PetrosStav/gemma3-tools:4b",
        "api_base": "http://localhost:11434",
        "chroma_persist_dir": "../data/db_storage"
      }
    }
  }
}

User: john_doe
├── PostgreSQL Tables: sales_john_doe, customers_john_doe
├── ChromaDB Collection: john_doe_metadata
└── Metadata Files: db_storage/john_doe/

cd CSV_Agent
python -m pytest tests/

# Test SQL generation
python simplified_query.py "show me all customers"

# Test cache
python inspect_cache.py

# Clear databases
python clear_databases.py

{
  "logging": {
    "level": "DEBUG",
    "file": "textsql.log"
  }
}

# Check PostgreSQL is running
pg_isready

# Verify credentials
psql -U username -d parseqri

# Update config.json with correct credentials

# Start Ollama service
ollama serve

# Verify models are available
ollama list

# Pull missing models
ollama pull PetrosStav/gemma3-tools:4b

# Ensure write permissions
chmod -R 755 data/db_storage

# Or on Windows, check folder permissions in Properties