How jawadafransaneem-31/dagger-leader-ai works
Dagger Leader is a multi-agent data-intelligence engine: a user asks a question, the workflow coordinates schema, memory, semantic, and SQL agents, validates and executes governed SQL, retries failures through healing, then explains the results. The README documents this pipeline and additional warehouse, Kafka, and dbt capabilities. Sampled code confirms the API/SDK entry points, shared state, LLM client, Snowflake connector, governance guard, and ChromaDB/Neo4j clients; some agent and workflow internals were not sampled, so uncertain wiring is omitted.
Main entry points: langgraph_workflow.py LangGraph workflow · llm_client.py LLM client · orchestrator.py Orchestrator · main.py SDK and CLI
Components 23
User interfaces
- REST API
api/app.py - SDK and CLI
main.py
Query pipeline
- LangGraph workflow
core/langgraph_workflow.py - Shared agent state
config/core/state.py - Orchestrator
agents/orchestrator.py - Schema agent
agents/schema_agent.py - Memory agent
agents/memory_agent.py - Semantic agent
agents/semantic_agent.py - SQL agent
agents/sql_agent.py - SQL validator
agents/validator.py - SQL healer
agents/healer_agent.py - Result explainer
agents/explainer_agent.py
Data context
- Schema and memory store
vector_db/chroma_client.py - Business-term graph
knowledge_graph/neo4j_client.py
Data and services
- LLM provider
- LLM client
core/llm_client.py - SQL guardrails
governance/sql_guard.py - Snowflake connector
core/data_connector.py - Snowflake warehouse
- BigQuery connector
data_sources/bigquery_connector.py - Kafka consumer
data_sources/kafka_consumer.py - dbt runner
data_sources/dbt_runner.py
Other components
- Question asker
Connections 28
- Question asker → REST API: submits question
- Question asker → SDK and CLI: asks question
- REST API → SDK and CLI: calls query
- SDK and CLI → LangGraph workflow: invokes workflow
- LangGraph workflow → Orchestrator: dispatches
- LangGraph workflow → Schema agent: dispatches
- LangGraph workflow → Memory agent: dispatches
- LangGraph workflow → Semantic agent: dispatches
- LangGraph workflow → SQL agent: dispatches
- LangGraph workflow → SQL validator: dispatches
- LangGraph workflow → SQL healer: dispatches
- LangGraph workflow → Result explainer: dispatches
- LangGraph workflow → Shared agent state: passes state
- Orchestrator → LLM client: requests completion
- Schema agent → Schema and memory store: searches schemas
- Memory agent → Schema and memory store: retrieves memory
- Semantic agent → Business-term graph: looks up terms
- Orchestrator → LLM client: uses
- SQL agent → LLM client: uses
- SQL healer → LLM client: uses
- Result explainer → LLM client: uses
- LLM client → LLM provider: calls provider
- SQL validator → SQL guardrails: checks SQL
- LangGraph workflow → Snowflake connector: executes query
- Snowflake connector → Snowflake warehouse: queries
- LangGraph workflow → BigQuery connector: supports source
- LangGraph workflow → Kafka consumer: supports streaming
- LangGraph workflow → dbt runner: supports models