Agyeman Enterprises
Time-synchronized audio and visual ingestion. Ask in plain language, get the answer bound to its exact visual source.
Watches Claude Code JSONL files and Cursor local SQLite databases. Captures coding sessions in real time before they expire.
Captures shared screen slides and transcribes system/microphone audio, linking them to solve slides decoupling.
Powers the Lektr student device. Captures physical room audio and whiteboard feeds with glare-resistant OCR.
Uses nomic-embed-text via Mneme or Ollama with prefixes for search optimization.
Encrypts transcripts and slide text at rest, keeping vectors readable for similarity search.
Guarantees speech segments are permanently bound to the visual frame active during duration.
Runs perceptual image hash comparison to keep only distinct slide/board states.