fact-telemetry 0.1.0a2

An asynchronous telemetry framework for FastAPI with pluggable storage backends.

FACT

FACT (FastAPI Advanced Collection Telemetry) is a lightweight, asynchronous telemetry framework for FastAPI applications.

FACT captures request telemetry through middleware, validates telemetry records, buffers them in an asynchronous in-memory queue, and persists them to configurable storage backends using batched writes. The framework is designed to minimize request overhead while providing reliable telemetry collection through retry mechanisms, dead-letter queue (DLQ) support, and runtime metrics.

FACT currently supports PostgreSQL and ClickHouse storage backends and provides a configurable architecture suitable for high-throughput, multi-tenant applications.

Status: Alpha Development

---

Features

• Middleware-based automatic telemetry collection
• Asynchronous in-memory queue
• Configurable batch processing
• PostgreSQL storage backend
• ClickHouse storage backend
• Multi-tenant request support
• Dynamic metadata support
• Payload validation
• Exponential backoff retry mechanism
• Local file-based Dead Letter Queue (DLQ)
• Runtime telemetry metrics
• Centralized configuration using FactConfig
• Environment variable support
• Automated test suite
• Benchmark and soak testing utilities

---

Architecture

                Incoming Request
                       │
                       ▼
        ClickHouseTelemetryMiddleware
                       │
                       ▼
            Payload Validation
                       │
                       ▼
           Async In-Memory Queue
                       │
                       ▼
              Batch Collection
                       │
                       ▼
           Retry with Backoff
                       │
             ┌─────────┴─────────┐
             │                   │
             ▼                   ▼
      Storage Backend        Dead Letter Queue
             │
             ▼
 PostgreSQL / ClickHouse

FACT separates request processing from persistence by using an asynchronous producer-consumer architecture. Incoming requests enqueue telemetry records immediately, while a background worker batches and persists records independently. This minimizes request latency and provides resilience during temporary storage failures.

---

Documentation

Detailed documentation is available in the docs/ directory.

• Installation
• Quick Start
• Architecture
• Configuration
• Storage Backends
• Middleware
• Benchmarking
• Roadmap

---

Project Status

FACT is currently in Alpha Development.

Implemented capabilities include:

• Middleware-based telemetry ingestion
• Batch processing engine
• PostgreSQL backend
• ClickHouse backend
• Retry with exponential backoff
• Dead Letter Queue (DLQ)
• Payload validation
• Runtime metrics
• Benchmarking utilities
• Automated tests

---

License

This project is licensed under the Apache License 2.0.

See the LICENSE file for details.