How saebyn/glowing-telegram works
Project type and purpose
• This is a backend‐focused, serverless microservices application (the “glowing-telegram” backend) that ingests, processes, transcribes, summarizes, and manages video streams and associated chat data. It exposes REST and WebSocket APIs, stores metadata in DynamoDB, files in S3, uses AWS Batch for heavy jobs, and integrates with OpenAI, Twitch, YouTube, and FFmpeg.
File structure
• Top-level service directories (each corresponds to a container or Lambda function):
• ai_chat_lambda (Rust)
• audio_transcriber (Python + Whisper)
• chat_processor_lambda (Rust)
• crud_api (Rust)
• embedding_service (Rust)
• media_lambda (Python)
• render_job (Rust)
• summarize_transcription (Rust)
• twitch_lambda (Rust)
• upload_video (Rust)
• video_ingestor (Rust)
• youtube_lambda (Rust)
• youtube_uploader_lambda (Python)
• plus shared libraries: gt_ffmpeg, gt_axum, gt_secrets, types
• Infrastructure as Code:
• cdk/ – AWS CDK (TypeScript + Python Lambdas) defining API Gateway (REST & WebSocket), Lambdas, DynamoDB tables, S3 buckets, AWS Batch jobs, Cognito, monitoring, CI/CD pipelines.
• Scripts: build & test orchestration, data migration, versioning.
README
• Confirms microservices architecture, AWS-based deployment via GitHub Actions, Docker images per service, automated CDK deploy, integration & unit test strategies.
• Describes major features (transcription, summarization, chat ingestion, highlights, archiving, OTIO generation).
Detailed capture of essential elements
• Related Lambdas into logical areas (Ingestion, Processing, Summarization, Delivery).
• Event sources/triggers (S3 PUT, DynamoDB Stream, HTTP POST).
• Interactions with off-AWS dependencies (OpenAI Chat API, Whisper inference).
• CI/CD flow: GitHub Actions → Docker Build → ECR → CDK Deploy → updated services.