csallahbakash/enterprise-data-agent

How csallahbakash/enterprise-data-agent works

This is a database-focused AI agent application: users request schema analysis, natural-language querying, Airflow DAG generation, or data-quality checks through a Streamlit UI or REST API. The API dispatches each request to a LangGraph agent; those agents use database introspection, query execution, and, where applicable, Claude to produce results. The README also documents Prometheus monitoring and an evaluation framework. The schema-discovery implementation was not sampled, so its internal wiring is not shown.

Main entry points: routes.py Agent Routes · database.py Database Access · main.py FastAPI App · llm.py LLM Client

Components 25

User Interface

Streamlit UI frontend/app.py
FastAPI App src/api/main.py
Agent Routes src/api/routes.py

Agent Workflows

Schema Discovery src/agents/schema_discovery.py
NL-to-SQL src/agents/nl_to_sql.py
Pipeline Generator src/agents/pipeline_generator.py
Data Quality src/agents/data_quality.py
Gather Schema Context src/agents/pipeline_generator.py
Generate DAG src/agents/pipeline_generator.py
Explain DAG src/agents/pipeline_generator.py
Load Schema Context src/agents/nl_to_sql.py
Generate SQL src/agents/nl_to_sql.py
Execute Query src/agents/nl_to_sql.py
Format Answer src/agents/nl_to_sql.py
Null Checks src/agents/data_quality.py
Duplicate Checks src/agents/data_quality.py
Quality Summary src/agents/data_quality.py

Data and LLM

Database Access src/core/database.py
PostgreSQL
LLM Client src/core/llm.py
Anthropic Claude

Evaluation and Monitoring

Agent Evaluation src/evaluation/evals.py
Prometheus Metrics src/evaluation/metrics.py
Prometheus monitoring/prometheus.yml

Other components

User

Connections 34

User → Streamlit UI: uses
User → FastAPI App: requests
Streamlit UI → FastAPI App: HTTP requests
FastAPI App → Agent Routes: mounts routes
Agent Routes → Schema Discovery: invokes
Agent Routes → NL-to-SQL: invokes
Agent Routes → Pipeline Generator: invokes
Agent Routes → Data Quality: invokes
NL-to-SQL → Load Schema Context: starts
Load Schema Context → Generate SQL: passes context
Generate SQL → Execute Query: passes SQL
Execute Query → Format Answer: passes results
Pipeline Generator → Gather Schema Context: starts
Gather Schema Context → Generate DAG: passes schema
Generate DAG → Explain DAG: passes DAG
Data Quality → Null Checks: starts
Null Checks → Duplicate Checks: passes checks
Duplicate Checks → Quality Summary: passes checks
Load Schema Context → Database Access: introspects
Execute Query → Database Access: executes query
Gather Schema Context → Database Access: introspects
Null Checks → Database Access: checks data
Duplicate Checks → Database Access: checks data
Database Access → PostgreSQL: reads
Generate SQL → LLM Client: requests SQL
Format Answer → LLM Client: requests answer
Generate DAG → LLM Client: requests DAG
Explain DAG → LLM Client: requests explanation
LLM Client → Anthropic Claude: invokes
Agent Routes → Prometheus Metrics: records requests
FastAPI App → Prometheus: exposes metrics
Prometheus → FastAPI App: scrapes
Agent Evaluation → NL-to-SQL: invokes graph
Agent Routes → Streamlit UI: returns results

More Python diagrams: lumen-labs/brainapi2 · nvidia-bionemo/proteina-complexa · opensqz/minicpm-v-cookbook · test-time-training/ttt-lm-jax · drdroidlab/playbooks

Drawn by GitDiagram on September 27, 2026 from the repository's file tree, README and source files. Read as Markdown ·