How ankit-144/mft works
MFT is a market-data and trading platform. Its main live workflow ingests broker ticks into closed candles, which inference reads to build features and produce signals; inference can submit orders to the execution service, which applies risk controls and manages broker order state. Separate branches provide historical backfill and strategy research/backtesting. The graph preserves documented capabilities where implementation excerpts are absent; those links are omitted or marked unverified rather than inferred from filenames alone.
Main entry points: engine.go Order lifecycle · engine.py Research simulation · pipeline.go Tick ingestion · candles.py Candle reader
Components 30
Market data
- Tick ingestion
services/ingestion/pipeline.go - Candle rollup
services/ingestion/candles.go - Historical backfill
services/jobs/worker.go - Backfill store
services/jobs/store.go
Inference
- Candle reader
services/inference/app/candles.py - Feature builder
services/inference/app/features.py - Model predictor
services/inference/app/predictor.py - Model registry
services/inference/model/base.py - Algorithm baselines
services/inference/model/algorithms.py - Heuristic model
services/inference/model/heuristic_model.py - TabFM model
services/inference/model/tabfm_model.py - Minute scheduler
services/inference/app/loop.py - Inference API
services/inference/app/main.py - Signal execution
services/inference/app/execution.py
Execution
- Order lifecycle
services/execution/engine.go - Risk controls
services/execution/risk/risk.go - Order journal
services/execution/journal.go - Execution API
services/execution/http.go
Research and backtesting
- Research simulation
services/inference/app/research/engine.py - Research adapters
services/inference/app/research/adapters.py - Chronological splits
services/inference/app/research/splits.py - Candidate evaluators
services/inference/app/research/evaluators.py - Backtest runner
services/inference/app/backtest.py
Shared platform
- Parquet storage
core/storage/storage.go - Runtime configuration
core/config/config.go - Domain contracts
core/contracts/contracts.go - KV cache
core/fluxkv/fluxkv.go
Other components
- Kite broker
core/broker/kite.go - Inference client
- Broker service
Connections 38
- Kite broker → Tick ingestion: streams ticks
- Tick ingestion → Parquet storage: writes ticks
services/ingestion/pipeline.go - Tick ingestion → Candle rollup: rolls ticks
- Candle rollup → Parquet storage: writes candles
- Historical backfill → Kite broker: fetches history
- Historical backfill → Backfill store: writes segments
- Parquet storage → Candle reader: reads candles
- Backfill store → Candle reader: historical data
- Inference API → Candle reader: reads context
services/inference/app/main.py - Inference API → Feature builder: builds features
services/inference/app/main.py - Inference API → Model predictor: scores context
services/inference/app/main.py - Model predictor → Model registry: selects model
services/inference/app/predictor.py - Model registry → Heuristic model: registers model
services/inference/model/base.py - Model registry → TabFM model: registers model
services/inference/model/base.py - Model registry → Algorithm baselines: registers algorithms
services/inference/model/__init__.py - Minute scheduler → Candle reader: reads candles
services/inference/app/runtime.py - Minute scheduler → Feature builder: builds features
- Minute scheduler → Model predictor: requests scores
services/inference/app/runtime.py - Minute scheduler → Signal execution: dispatches signals
- Signal execution → Execution API: submits orders
- Inference client → Inference API: requests prediction
services/inference/app/main.py - Execution API → Order lifecycle: routes orders
- Order lifecycle → Risk controls: checks risk
services/execution/engine.go - Order lifecycle → Order journal: persists state
- Order lifecycle → Broker service: reconciles orders
services/execution/engine.go - Research simulation → Research adapters: loads data
services/inference/app/research/engine.py - Research simulation → Chronological splits: splits chronology
services/inference/app/research/engine.py - Research simulation → Candidate evaluators: evaluates candidates
services/inference/app/research/engine.py - Research simulation → Feature builder: builds features
services/inference/app/research/engine.py - Research simulation → Algorithm baselines: compares algorithms
services/inference/app/research/engine.py - Backtest runner → Candle reader: loads candles
- Backtest runner → Model predictor: scores model
- Runtime configuration → Tick ingestion: configures pipeline
- Runtime configuration → Order lifecycle: configures engine
- Domain contracts → Feature builder: defines inputs
- Domain contracts → Parquet storage: maps records
core/storage/storage.go - KV cache → Tick ingestion: caches state
- KV cache → Order lifecycle: tracks claims