video1.0<iframe src="https://www.loom.com/embed/dc0401aa5e2541b1a7cb7d5ad98dc925" frameborder="0" width="1912" height="1434" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14341912Loomhttps://www.loom.com14341912https://cdn.loom.com/os/loom-media-content/loom/d7213939-ee11-4aa7-8bf9-d0d26f7b6948/b162abfa-df15-4cec-af84-d12ce6a0653c/sessions-thumbnails-dc0401aa5e2541b1a7cb7d5ad98dc925-4b40d9960f9903d9.gif/binary?authToken=eyJraWQiOiJtaWNyb3MtZ3JvdXAvdGRwLW9zLWFyY2hldHlwZS9iNmJibGw0OGo3cm9zbDByIiwiYWxnIjoiUlMyNTYifQ.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.atOynnqlagqAqpUvghiHEiSQAiLYhHySCFdCZuoDkVvA2Dhfrzq-SbBm5dp-Jj07daUoA3rGNsSipz1GToTl9vWNvkq0PXLTdhNMuRLtoCXT_l6ngj7gqOuuTcpC4yjOUuT5-KsHH2ABZiWEdBX4t2L-c6NCFgcGQJg-9fGbJRPob8BWoVnbOl_2zNwjeVJoh55wGVsBciQgye9GMcS4FH_Q2xW1XD_TGXNFT5BjIuEZelP-NgSJc-VgPehi1Must9W9JFxePGYaDaAAQgxRnqFO-hvXzW6sqwU7pi3l7rKtYj0fnNioR7ueD4ATIzM6JBzsdzUm2k794H_hQasS8g&Expires=1792154249&Signature=IieUzY31QY9qMoSbnbTYTwCQhpJm6qlYiYXlobo6X9UVhBRcbgsoyVFSioO2YsAWQOmJMRugIkz-gqSe6%7EzF9OuX0DhKPCjtP3AJSbFKFLL8y8XGMkHQdYAM6%7EqMSRdOr-RTBPSDUUM9EYCLwPzgCs8HIHg1z0JrtVJq-AZ9hhxPuv4--gC0wrBv-bonJOrffjQnGAtemOpzmF4kN0lQqK28cY-TggMr7SoDePdgnkPs8xkshS3PjoLYsvFSbU0B6eLg54D4YzzpBiBFDBadgJ2EQUiD1iqKG2C-Mg6%7E11bXIIA9Ub%7En2CFzJXwgwFPY3YLzFvbDq-PHAehviNSpTg__&Key-Pair-Id=KQOSYIR44AIC0140.689Internal Knowledge Assistant WorkflowThis Loom explains an Internal Knowledge Assistant built on a RAC-based system. The workflow triggers on changes in a Dropbox folder, verifies incoming requests with a crypto node, and uses cursor information to a Superbase so only new or changed files are processed. It downloads text or PDF files, extracts their content, and uses Gemini embedding with a default data loader containing metadata like file name, Dropbox, file ID, and timestamp to store vectors in Superbase. Part two covers the user side where a Slack-based node triggers processing, checks whether the request is from a user or a bot, runs an AI agent using an OpenRouter model with Gemini 2.5 flash, and retrieves relevant info from the same embedding store using simple memory.