emiber/utdt

How emiber/utdt works

UTDT is a Vue-and-Fastify text-to-SQL chatbot. A user supplies the shared access key, explores the available data, then submits a question; the backend classifies it, generates and executes read-only SQL, builds narrative and chart/table blocks, and returns the answer. Follow-up context and LLM-call records persist in the app database; the UI also offers a usage report. The graph preserves the documented catalog, multi-result answers, conversational replies, and usage features. Some subsystem links use README evidence because their implementations were not excerpted.

Main entry points: pipeline.ts Answer pipeline · api.ts API client · app.ts Route assembly · llm.ts LLM adapter

Components 32

Web interface

App shell frontend/src/App.vue
Key entry frontend/src/components/AccessKeyForm.vue
Key state frontend/src/auth.ts
Chat view frontend/src/components/ChatView.vue
Answer blocks frontend/src/components/ResponseBlocks.vue
Chart display frontend/src/components/ChartBlock.vue
Data catalog view frontend/src/components/DataCatalog.vue
Usage view frontend/src/components/UsageView.vue
API client frontend/src/api.ts

API and state

HTTP server backend/src/server.ts
Route assembly backend/src/app.ts
Chat endpoint backend/src/chat.ts
Request authentication backend/src/auth.ts
Conversation history backend/src/conversations.ts
Catalog endpoint backend/src/catalog.ts
Usage endpoint backend/src/usage-route.ts
Data metadata backend/src/metadata.ts
Dataset database backend/src/db.ts

Question answering

Answer pipeline backend/src/pipeline.ts
Question classification backend/src/classify.ts
Schema context backend/src/schema-context.ts
SQL generation backend/src/generate-sql.ts
SQL execution backend/src/execute-sql.ts
Text replies backend/src/reply.ts
Result insights backend/src/insights.ts
Chart construction backend/src/chart.ts

LLM and reporting

Usage reporting backend/src/usage.ts
LLM adapter backend/src/llm.ts
Prompt rendering backend/src/prompts.ts
Call timing backend/src/timing.ts

Other components

User
OpenAI APIs

Connections 40

User → Key entry: enters key
Key entry → Key state: saves key frontend/src/components/AccessKeyForm.vue
User → Chat view: asks question
App shell → Chat view: routes to
App shell → Data catalog view: shows catalog frontend/src/App.vue
App shell → Usage view: routes to
Chat view → API client: requests chat frontend/src/components/ChatView.vue
Data catalog view → API client: loads catalog frontend/src/components/DataCatalog.vue
Usage view → API client: loads report frontend/src/components/UsageView.vue
API client → Key state: uses key frontend/src/api.ts
API client → Chat endpoint: posts question
API client → Catalog endpoint: requests catalog
API client → Usage endpoint: requests report
Answer blocks → Chart display: renders chart frontend/src/components/ResponseBlocks.vue
Chat view → Answer blocks: renders answer frontend/src/components/ChatView.vue
HTTP server → Route assembly: builds app backend/src/server.ts
Route assembly → Request authentication: checks key backend/src/app.ts
Route assembly → Chat endpoint: registers route backend/src/app.ts
Route assembly → Usage endpoint: registers route backend/src/app.ts
Route assembly → Catalog endpoint: serves catalog backend/src/app.ts
Chat endpoint → Answer pipeline: runs pipeline backend/src/chat.ts
Chat endpoint → Conversation history: loads and saves backend/src/chat.ts
Catalog endpoint → Data metadata: reads metadata
Catalog endpoint → Dataset database: reads dataset
Answer pipeline → Schema context: builds context backend/src/pipeline.ts
Answer pipeline → Question classification: classifies question backend/src/pipeline.ts
Answer pipeline → Text replies: builds text reply backend/src/pipeline.ts
Answer pipeline → SQL generation: generates queries backend/src/pipeline.ts
Answer pipeline → Result insights: builds insights backend/src/pipeline.ts
Answer pipeline → Chart construction: builds charts backend/src/pipeline.ts
SQL generation → SQL execution: executes SQL backend/src/generate-sql.ts
SQL generation → LLM adapter: requests SQL backend/src/generate-sql.ts
Question classification → LLM adapter: requests classification backend/src/classify.ts
Text replies → LLM adapter: requests reply backend/src/reply.ts
Result insights → LLM adapter: requests insights backend/src/insights.ts
SQL generation → Prompt rendering: renders prompts backend/src/generate-sql.ts
Answer pipeline → Conversation history: uses history types backend/src/pipeline.ts
LLM adapter → OpenAI APIs: calls APIs backend/src/llm.ts
Usage endpoint → Usage reporting: builds report
Usage reporting → Conversation history: reads call records

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·