How guruvaidev/avaloka works
Avaloka turns a natural-language goal and dataset into a checked analysis, reproducible artifacts, and optionally a trained and served model. Users can enter through the CLI or the web/API platform; the documented analysis path profiles data, plans and validates code, executes it, then reports findings with lineage. Training and deployment, Auto Insights, collaboration, scheduled and distributed runs, and data connectors are major product branches. The README documents the workflow; source excerpts and the index give only partial implementation evidence, so unverified internal links are omitted.
Main entry points: cli.py Avaloka CLI · server.py Web and API · execution_agent.py Analysis execution · result_narrator.py Results and lineage
Components 23
User interfaces
- Avaloka CLI
avaloka/cli.py - Web and API
app/api/server.py - Web dashboard
ui/src/components/dashboard - Analysis service
app/interfaces/service.py
Analysis workflow
- Goal planner
app/agents/planner.py - Data profiling
app/agents/profiling_agent.py - Data preparation
app/agents/preparation_pipeline.py - Analysis coding
app/agents/coder.py - Code validation
app/agents/validator.py - Analysis execution
app/agents/execution_agent.py - Visualization
app/agents/visualization_agent.py - Results and lineage
app/agents/result_narrator.py
Data and artifacts
- File connectors
file_handler/handler.py - Data sampling
app/agents/sampling_agent_daft.py - Run artifacts
app/interfaces/artifacts.py
Model lifecycle
- Model training
app/agents/mta_v2/agent.py - Model inference
app/agents/mta_v2/inference.py - Cloud deployment
app/agents/mta_v2/inference_service_manager.py
Platform services
- Object storage
app/core/storage.py - Ray execution
app/infra/ray_manager.py - Scheduled runs
app/agents/scheduler.py - Layered memory
app/services/memory_runtime.py
Other components
- Analyst
Connections 18
- Analyst → Avaloka CLI: asks goal
README.md - Analyst → Web and API: submits analysis
- Web dashboard → Web and API: uses platform
- Avaloka CLI → Data profiling: profiles data
README.md - Avaloka CLI → Goal planner: plans analysis
README.md - Avaloka CLI → Analysis coding: writes code
README.md - Avaloka CLI → Code validation: validates code
README.md - Avaloka CLI → Analysis execution: runs analysis
README.md - Avaloka CLI → Results and lineage: reports findings
README.md - File connectors → Data sampling: provides rows
- Analysis execution → Run artifacts: records outputs
- Run artifacts → Object storage: persists artifacts
- Avaloka CLI → Model training: trains model
README.md - Model training → Model inference: enables inference
- Model inference → Cloud deployment: deploys service
app/agents/mta_v2/inference_service_manager.py - Avaloka CLI → Ray execution: distributes runs
README.md - Web and API → Scheduled runs: schedules runs
- Results and lineage → Layered memory: records results