Project Type Detection & Adaptation Track

Step 0 Pre-Check: Architecture & Track Classification

This document records the formal codebase inspection and classification of 14-tracing-backend.

---

Detection Evidence

A comprehensive inspection of the repository /Users/avishjain/avishjain9-repos/14-tracing-backend was conducted:

Artifact Checked	Findings / Content	Classification Evidence
Directory Name	14-tracing-backend	Distributed tracing observability platform.
Backend Manifest	backend/alembic.ini, backend/src/api/app.py	Python 3.13, FastAPI, SQLAlchemy, Alembic, ClickHouse/Async SQLite, Pydantic.
Frontend Manifest	frontend/package.json, frontend/vite.config.ts	React 18, Vite 6, TypeScript 5.7, Tailwind CSS, Lucide icons.
ML Engine	ml/ (11 Python modules)	Outlier latency anomaly detectors, dependency anomaly scorers, and LLM incident summary engine.
Test Suite	backend/tests/ (141 test files)	313 passing unit/integration tests running via pytest.

---

Formal Classification

• Primary Classification: Track A (Full Stack Polyglot / Systems Observability Platform)
• Sub-tracks:
  • High-throughput asynchronous backend (FastAPI, asyncio, ClickHouse columnar storage)
  • Interactive data-visualization frontend (React 18, Vite, Span DAG Explorer)
  • Production ML & AI anomaly detection engine (statistical heuristics, causal graphs, pre-trained LLM summarization API)

---

Adaptation Rules for This Project

1. Backend Sections: Focus on asynchronous ingestion pipelines, tail-sampling ring buffers, span DAG reconstruction algorithms, critical path graph traversal, and columnar ClickHouse indexing.
2. Frontend Sections: Focus on React component tree reconciliation, SVG waterfall timelines, canvas dependency graphs, virtualized span trees, and sub-millisecond telemetry search.
3. Database Sections: Compare ClickHouse columnar database vs PostgreSQL vs Elasticsearch vs Cassandra for append-only span storage.
4. AI & ML Sections: Document the 11 ML models (latency outliers, causal root cause graphs, vector semantic trace search, and pre-trained LLM incident summarizer).
5. Interview Focus: Distributed tracing standards (W3C TraceContext, OpenTelemetry OTLP), sampling strategies (head vs tail vs adaptive), and scaling to 100,000 spans/sec.