← Back to skills
extension
Category: Content & MediaAPI key requirement unconfirmed

ai-tools

Reference for all AI tools available in DBX Studio's AI chat system. Use when adding, modifying, or debugging AI tool definitions, tool execution, or provider integrations.

personAuthor: jakexiaohubgithub

DBX Studio AI Tools Reference

Tool Definitions Location

apps/api/src/orpc/routers/ai/tools.ts

Tool Executor Location

apps/api/src/orpc/routers/ai/toolExecutor.ts

Available Tools (AI_TOOLS array)

| Tool Name | Purpose | Required Params | |-----------|---------|-----------------| | read_schema | Get all tables and structure from schema store | schema_name | | get_table_data | Preview rows from a table (generates SELECT) | table_name | | execute_query | Run a SELECT/WITH query | sql | | generate_chart | Create chart config (bar/line/pie/scatter/area/histogram) | chart_type, title, data_query | | describe_table | Get table metadata from schemaTables DB table | table_name | | get_table_stats | Get statistics summary for a table | table_name |

AI Providers

| Service ID | Provider | File | |------------|----------|------| | 1 | AWS Bedrock | ai-stream.ts (streaming) + providersWithTools.ts | | 2 | OpenAI | providersWithTools.ts → callOpenAIWithTools | | 3 | Anthropic | providersWithTools.ts → callAnthropicWithTools | | 4 | Qwen | index.ts | | 5 | Ollama | index.ts | | 6 | Gemini | index.ts | | 7 | Groq | index.ts | | 8 | DBX Agent | External → MAIN_SERVER_URL |

Tool Format Conversions

// Anthropic format
getAnthropicTools() → { name, description, input_schema }

// OpenAI format
getOpenAITools() → { type: 'function', function: { name, description, parameters } }

// Bedrock format (in ai-stream.ts)
{ toolSpec: { name, description, inputSchema: { json: input_schema } } }

Adding a New Tool

  1. Add to AI_TOOLS array in tools.ts
  2. Add execution logic in toolExecutor.ts → executeTool switch
  3. Implement the handler function executeMyNewTool(input, context)
  4. Both getAnthropicTools() and getOpenAITools() will pick it up automatically
  5. For Bedrock, the conversion in ai-stream.ts is also automatic

System Prompt Location

Main streaming system prompt: apps/api/src/routes/ai-stream.ts around line 132–172

oRPC chat system prompt: apps/api/src/orpc/routers/ai/providersWithTools.ts — SYSTEM_PROMPT_WITH_TOOLS constant