transcript-summarizer

Transcript summarization converts meeting recordings into structured summaries highlighting decisions, action items, and key topics. Place your VTT, SRT, or plain-text transcript in the input directory, configure environment variables for source/destination paths and LLM backend, then run the Reflex plugin to generate markdown summaries ready for indexing.

Yes, transcript summarization works with VTT/SRT captions, plain transcripts, and pasted notes. The skill extracts text from caption formats and processes it alongside raw transcripts, making it flexible for different meeting capture methods without reformatting.

The skill produces structured markdown summaries with executive overview, topic lists, decisions with reasoning, action-item tables, and metadata. Outputs are formatted for RAG storage and knowledge-base indexing in vector stores like Qdrant, enabling semantic search across meetings.

Configure REFLEX_TRANSCRIPT_SRC_DIR and REFLEX_TRANSCRIPT_DST_DIR environment variables to point to input and output directories, optionally set REFLEX_TRANSCRIPT_LLM and REFLEX_TRANSCRIPT_MODEL for your backend, then run the Reflex plugin after uploading transcripts to the source folder.

Yes, the skill uses optional LLM backends to parse transcripts and automatically extract decisions with reasoning and action items into structured tables. This enables teams to capture outcomes without manual meeting notes.

Yes, outputs are specifically designed for RAG indexing and vector-store storage. Structured markdown and metadata make summaries directly queryable, allowing teams to search and retrieve meeting outcomes semantically alongside other knowledge-base content.