⚡ Text Summariser

About this demo

Drop a plain-text file and get exactly five bullet-point takeaways in a few seconds. Everything runs on a fully serverless stack—there are no EC2 instances to patch or keep warm.

• AWS Lambda – one Python function that calls the LLM and writes results to S3.
• Amazon API Gateway (HTTP API) – HTTPS endpoint the browser hits.
• Amazon Bedrock – choose the model on the fly: Titan-Lite (default), Cohere Command-Light, or Claude 3 Haiku.
• Amazon S3 – stores your original file under raw/ and the summary under summary/.
• AWS SAM – one template deploys the whole lot.

What it’s good for

• Quick content previews before you dive into a long article.
• Batch-summarising many files by pushing them to the raw/ folder.
• Plug-and-play demo of how to wire Bedrock into a serverless workflow.

Current limitations

• Text only (<200 KB per file). No PDFs or HTML yet.
• One language at a time—auto-detect works but bullets come back in English.
• Your Bedrock account is billed for tokens; Haiku costs more than Titan/Cohere.

Future potential: swap in larger models, extend to PDFs via Textract, or trigger downstream workflows (Slack alerts, vector search, etc.).

Source code on GitHub.