How deeep095/auto_ml works
AutoML is a Telegram-guided tabular machine-learning service. Users upload or reuse a dataset, receive a classification or regression plan, optionally customize model parameters, and get evaluated models plus visualizations, reports, and a saved best model. The documented service boundary is n8n to FastAPI; the sampled engine code shows the training pipeline’s internal stages. The n8n workflow exports and some artifact-delivery details were not sampled, so those relationships rely on README-level documentation.
Main entry points: pipeline.py Run coordinator · app.py FastAPI routes · reports.py Reports and summaries · utils.py Artifact utilities
Components 16
Telegram workflows
- Chat router
n8n_workflows/main_workflow.json - AI customizer
n8n_workflows/ai_customize_workflow.json - Training workflow
n8n_workflows/analyze_and_train_workflow.json
ML service
- FastAPI routes
ml_engine/app.py - Plan and model schemas
ml_engine/app.py - Runtime configuration
ml_engine/config.py
Training pipeline
- Run coordinator
ml_engine/pipeline.py - Dataset preparation
ml_engine/preprocessing.py - Model registry
ml_engine/models.py - Evaluation metrics
ml_engine/metrics.py
Results and artifacts
- Visualization generation
ml_engine/plots.py - Reports and summaries
ml_engine/reports.py - Artifact utilities
ml_engine/utils.py - Shared artifacts
shared/artifacts
Other components
- Telegram user
- Telegram Bot API
Connections 19
- Telegram user → Chat router: sends dataset
README.md - Chat router → AI customizer: requests planning
- AI customizer → Training workflow: customizes plan
- Training workflow → FastAPI routes: submits plan
- FastAPI routes → Plan and model schemas: validates request
ml_engine/app.py - FastAPI routes → Run coordinator: runs training
ml_engine/app.py - Run coordinator → Dataset preparation: prepares data
ml_engine/pipeline.py - Run coordinator → Model registry: builds estimators
ml_engine/pipeline.py - Run coordinator → Evaluation metrics: evaluates predictions
ml_engine/pipeline.py - Run coordinator → Visualization generation: generates plots
ml_engine/pipeline.py - Run coordinator → Reports and summaries: generates reports
ml_engine/pipeline.py - Run coordinator → Artifact utilities: saves model
ml_engine/pipeline.py - Dataset preparation → Runtime configuration: reads defaults
ml_engine/preprocessing.py - Reports and summaries → Runtime configuration: reads report settings
ml_engine/reports.py - Artifact utilities → Runtime configuration: reads storage settings
ml_engine/utils.py - Artifact utilities → Shared artifacts: writes outputs
- Reports and summaries → Shared artifacts: writes reports
- FastAPI routes → Telegram Bot API: returns results
README.md - Telegram Bot API → Telegram user: delivers results
README.md