How divyanshgarg25/symantix works
Symantix is an outage-diagnosis service: logs enter through RabbitMQ, are validated and sanitized, then indexed semantically and as service topology. An authenticated engineer request queries vector matches, uses their trace IDs for graph traversal, and streams a diagnosis response. Training utilities prepare triplets and fine-tune embeddings, but they are separate from the runtime ingestion and diagnosis path. The request handler currently streams a simulated model response; the README describes local model inference as the intended capability. Edges cite excerpts or explicit README statements; unsampled model-training subsystems are included without speculative runtime connections.
Main entry points: worker.py Log worker · api.py Diagnosis router · vector_store.py Vector store · graph_store.py Topology store
Components 17
Log ingestion
- Log simulator
backend/logger_simulator.py - RabbitMQ
- Log worker
backend/worker.py - Log event model
backend/models.py - Entity extraction
backend/graph_extractor.py - Runtime settings
backend/config.py
Hybrid retrieval
- Vector store
backend/vector_store.py - Topology store
backend/graph_store.py - Infrastructure errors
backend/exceptions.py
Diagnosis API
- Diagnosis router
backend/api.py - Diagnosis stream
backend/api.py
Embedding training
- SRE instruction data
core-ml/sre_instructions_generator.py - Triplet generator
core-ml/dataset_generator.py - Embedding trainer
core-ml/train.py - QLoRA fine-tuning
core-ml/qlora_finetune.py
Other components
- Log producers
- SRE engineer
Connections 19
- Log producers → RabbitMQ: sends logs
README.md - Log simulator → RabbitMQ: publishes events
backend/logger_simulator.py - Log simulator → Log event model: creates events
backend/logger_simulator.py - RabbitMQ → Log worker: delivers messages
README.md - Log worker → Log event model: validates payload
backend/worker.py - Log worker → Entity extraction: extracts entities
backend/worker.py - Log worker → Vector store: indexes logs
backend/worker.py - Log worker → Topology store: stores topology
backend/worker.py - SRE engineer → Diagnosis router: requests diagnosis
README.md - Diagnosis router → Vector store: searches semantics
backend/api.py - Diagnosis router → Topology store: traverses topology
backend/api.py - Diagnosis router → Diagnosis stream: streams response
backend/api.py - Diagnosis stream → SRE engineer: returns diagnosis
README.md - Embedding trainer → Triplet generator: generates training data
core-ml/train.py - Log worker → Runtime settings: reads settings
backend/worker.py - Diagnosis router → Infrastructure errors: handles failures
backend/api.py - Vector store → Infrastructure errors: raises errors
backend/vector_store.py - Topology store → Infrastructure errors: raises errors
backend/graph_store.py - Diagnosis router → Diagnosis stream: local model inference
backend/api.py