meet-transcription
Automatic Google Meet transcription pipeline: capture, queue, transcription, persistence and TXT download.
Stack
PythonFastAPIDeepgramWhisperRedis+2
Google DriveworkerqueueDokployGHCR
Projects
Full flow, not an empty showcase.
These projects show input, processing, queue, database, automation, useful output and real operations. No invented metrics, no inflated status.
6 selected projectsGitHub auditedno inflated statusfull flow
Jump to gridFilter projects
Filter projects
6 visible projects
Shows every project selected for this phase.
Transcription pipelineFlagship Automation APIOpen-source notebooklm-api
HTTP-first API to operate NotebookLM flows programmatically, with CLI, web UI and per-account isolation.
StackFastAPICLIWeb UIMulti-accountPlaywright/storage stateNotebookLMHTTP-firstsession isolationBearer authsafe errorsCPU transcription APIOpen-source transcricao
Lightweight transcription API designed to run on CPU and cheap VPS, without requiring GPU.
StackFastAPIfaster-whisperCTranslate2DockerTraefikCPU-onlyVPSPortainerhealthcheckOperational dashboardOpen-source dashboard-sheets
Dashboard wired to Google Sheets, packaged as a multi-arch container and ready for Portainer.
StackStreamlitGoogle Sheets APIDocker multi-archGitHub ActionsGHCRnon-roothealthcheckPortainerTechnical publicationOpen-source interface-atlas
Visual documentation about interface architecture, frontend systems, SEO and modern UI patterns.
StackAstroMDXTypeScriptReact islandsSEOcontenti18nsearchsitemapRSSOwn productIn progress OrkestraAI
Own product to automate Instagram replies, comments and DMs with AI agents. Landing live; app in development.
StackReact 19HonoNestJSBullMQRedis+3InstagramagentsRAGmulti-tenantbilling
Reading guide
How I evaluate a project
Naming the stack is not enough. The project needs to show the data path and how it stays alive.
- Real problem: which manual work, risk or bottleneck it reduces.
- Architecture: input, processing, persistence/queue and output.
- Operations: Docker, healthcheck, logs, env and deploy when documented.
- Handled failure: limits, expired session, external provider or stopped queue show up in the design.
- Maintenance: honest next steps, without calling a lab project production.
Next process
Want to turn a manual process into a running system?
Show me the input, the steps and the output you need. I answer with a possible architecture.