# API Reference
Complete API reference for the Combined MCP Server.
## Redshift Tools
### run_query
Execute SQL queries on Amazon Redshift with IAM-based authentication.
**Parameters:**
- `sql` (string, required): The SQL query to execute
- `db_user` (string, required): Database user for authentication
- `db_group` (string, optional): Database group for permissions
**Returns:**
- `row_count`: Total number of rows in result
- `columns`: List of column names
- `rows`: Result rows (sample of 20 if > 100 total)
- `is_sample`: Whether result is a sample
- `s3_path`: S3 location of full results (if sampled)
- `execution_time_ms`: Query execution time
**Example:**
```json
{
"sql": "SELECT customer_id, name FROM customers WHERE region = 'US'",
"db_user": "analyst",
"db_group": "sales_team"
}
```
### list_schemas
List all schemas in the Redshift database.
### list_tables
List all tables in a specific schema.
### describe_table
Get detailed column information for a table.
---
## Knowledgebase Tools
### build_vectorstore
Build or rebuild the vector store from S3 markdown files.
**Process:**
1. Downloads all `.md` files from configured S3 bucket
2. Chunks documents into ~1000 character segments
3. Generates embeddings using AWS Bedrock Titan
4. Stores in PostgreSQL with pgvector
**Returns:**
- `status`: Build status (ready, failed)
- `document_count`: Number of document chunks indexed
- `build_time_seconds`: Time taken to build
### query_vectorstore
Search the knowledge base with hybrid search.
**Parameters:**
- `query` (string, required): Search query text
- `top_k` (integer, default: 10): Maximum results to return
- `search_type` (enum, default: "hybrid"): Search mode
- `semantic`: Vector similarity only
- `keyword`: Full-text search only
- `hybrid`: Combined with RRF reranking
**Returns:**
- `results`: List of matching documents
- `search_type`: Type of search performed
- `query_time_ms`: Search duration
- `cached`: Whether result was from cache
### get_vectorstore_status
Get current vector store status and cache statistics.
---
## Error Handling
All tools return a consistent error format:
```json
{
"success": false,
"error": "Error message description",
"error_type": "ErrorClassName"
}
```
## Rate Limits
- Bedrock embeddings: 10 requests/second
- Redshift queries: Limited by cluster capacity
- Query cache: 30 items (FIFO eviction)
