{"ref":"content:cap-in-the-age-of-ai-7-vector-embeddings-wrap-up","page":"/content/cap-in-the-age-of-ai-7-vector-embeddings-wrap-up","json":"/content/cap-in-the-age-of-ai-7-vector-embeddings-wrap-up.json","generated":"2026-10-26","drafted_with_ai":true,"entity":{"id":"cap-in-the-age-of-ai-7-vector-embeddings-wrap-up","content_type":"video","url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up","published_on":"source:youtube","published":"2026-10-16","part_of":"content:cap-in-the-age-of-ai","position":7,"authors":["actor:dj-adams","actor:akatuoro"],"summary":"The same embeddings model taken from SQLite to [SAP HANA Cloud](artifact:btp-hana-cloud), then handed to an agent. Locally it is [@cap-js/ai](artifact:cap-js-ai) that adds the vector functions SQLite lacks and downloads an all-MiniLM L6v2 ONNX model on first start; on HANA natural language processing has to be switched on, the model name is a mandatory argument, the vectors it returns are 768 dimensions, and dimensions must match on both sides of a comparison. HANA itself can be given [SAP AI Core](artifact:btp-ai-core) as a remote source for other embedding models, which a `CDS.ql` expression has to inject because the postfix notation only takes what the CDS model holds, and the hour ends with a `search` function over cosine similarity exposed over [@cap-js/mcp](artifact:cap-js-mcp), so an agent asked about low output from solar panels calls semantic search instead of writing the query itself.","short":"CAP embeddings moved from SQLite to SAP HANA Cloud with @cap-js/ai","explains":["artifact:cap-js-ai","artifact:cap-js-hana","artifact:cap-js-mcp","artifact:btp-hana-cloud","artifact:btp-ai-core","artifact:sap-cds-dk","capability:retrieval-augmented-generation","capability:mcp-servers"],"sources":[{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the package behind the SQLite functions","checked":"2026-10-18","quote":"So this package provides custom functions in SQLite to uh calculate embeddings uh with a local embedding model."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — which package it is","checked":"2026-10-18","quote":"So we um did install Capture AI."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the model it downloads","checked":"2026-10-18","quote":"But here then we have the all-MiniLM L6v2. Uh this is the embedding model for ONNX here."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — downloaded on first start","checked":"2026-10-18","quote":"for the first when you start for the first time, it will automatically download a model, right?"},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — cds bind --exec for the REPL","checked":"2026-10-18","quote":"this like CDS bind excel as well. So that's a way with which we can start any command, right? Using like the environment variable with that get provided by the hybrid profile."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — adding the HANA package","checked":"2026-10-18","quote":"I did already pre like previously do like CDS at Hana, right? So, that's just install the CDS Hana package with all the defaults."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — natural language processing on HANA","checked":"2026-10-18","quote":"This is something where I need to um enable here natural language processing down here."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the model name is mandatory","checked":"2026-10-18","quote":"on Hana this is mandatory right so like if I write this yeah, I do get an error because it does expect these arguments."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — 768 dimensions","checked":"2026-10-18","quote":"these models have like vector embedding dimension of 768"},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — dimensions have to match","checked":"2026-10-18","quote":"why is it important is that when we now do cosine similarity this needs to match actually. So so the two vectors need to have the same dimensions."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — HANA connected to SAP AI Core","checked":"2026-10-18","quote":"we're not connecting this the CAP server now to AI Core, we're connecting Hana itself to AI Core."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the remote source","checked":"2026-10-18","quote":"That's the remote source that is created that I created on Hana when connecting AI Core."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — why an expression is needed","checked":"2026-10-18","quote":"this post fix notation we expect things to be in our CDS model. That's not the case here."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the search function","checked":"2026-10-18","quote":"add another function here, right? Just to make the search easily right callable."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — exposed over MCP","checked":"2026-10-18","quote":"uh and um uh so yeah, we have our processor service now connected via MCP."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — the agent calls the search tool","checked":"2026-10-18","quote":"It does call the um right simply a search with a query, low energy solar panels."},{"url":"https://www.youtube.com/watch?v=JGXE952l4JM","title":"CAP in the age of AI #7 - vector embeddings wrap-up — why give an agent semantic search","checked":"2026-10-18","quote":"if you uh give embedding functionality also to agents, right? So, like if you give something like a semantic search to agents"}],"provenance":{"discovered_via":[{"source":"source:youtube"}],"first_seen":"2026-10-18"},"drafted_with_ai":true},"readings":{"liveness":[{"label":"Checked against the source","tier":"unknown","evidence":"nobody has checked it"},{"label":"Source reachable","tier":"unknown","evidence":"not crawled"},{"label":"Age","tier":"good","evidence":"published this month"}],"compatibility":[{"label":"Version","tier":"unknown","evidence":"no version declared"},{"label":"Moved since?","tier":"good","evidence":"no recorded change after 2026-10-16"},{"label":"Findings","tier":"good","evidence":"none flagged"}],"reach":[{"value":null,"unit":"views"},{"value":null,"unit":"comments"},{"value":null,"unit":"likes"}]},"signals":null,"series":{"head":{"id":"cap-in-the-age-of-ai","title":"CAP in the age of AI","content_type":"series"},"container":true,"of":7,"at":7,"parts":[{"position":1,"title":"CAP in the age of AI #1 - the MCP adapter","url":"https://www.youtube.com/watch?v=VPJhkR9pHVM","published":"2026-08-07","state":"indexed","id":"cap-in-the-age-of-ai-1-the-mcp-adapter"},{"position":2,"title":"CAP in the age of AI #2 - more on the MCP adapter","url":"https://www.youtube.com/watch?v=5RBFxl5IqKs","published":"2026-08-14","state":"indexed","id":"cap-in-the-age-of-ai-2-more-on-the-mcp-adapter"},{"position":3,"title":"CAP in the age of AI #3 - from MCP to A2A","url":"https://www.youtube.com/watch?v=Nu_bEVFM7DM","published":"2026-08-21","state":"indexed","id":"cap-in-the-age-of-ai-3-from-mcp-to-a2a"},{"position":4,"title":"CAP in the age of AI #4 - more on agents","url":"https://www.youtube.com/watch?v=2nWIwzoNw-M","published":"2026-09-11","state":"indexed","id":"cap-in-the-age-of-ai-4-more-on-agents"},{"position":5,"title":"CAP in the age of AI #5 - agents wrap-up","url":"https://www.youtube.com/watch?v=cMqSxjtN1hs","published":"2026-10-02","state":"indexed","id":"cap-in-the-age-of-ai-5-agents-wrap-up"},{"position":6,"title":"CAP in the age of AI #6 - vector embeddings","url":"https://www.youtube.com/watch?v=tdtRgPy4B40","published":"2026-10-09","state":"indexed","id":"cap-in-the-age-of-ai-6-vector-embeddings"},{"position":7,"title":"CAP in the age of AI #7 - vector embeddings wrap-up","url":"https://www.youtube.com/watch?v=JGXE952l4JM","published":"2026-10-16","state":"indexed","id":"cap-in-the-age-of-ai-7-vector-embeddings-wrap-up"}]},"edges":{"authors":[{"ref":"actor:dj-adams","name":"DJ Adams","json":"/actors/dj-adams.json"},{"ref":"actor:akatuoro","name":"Johannes Vogt","json":"/actors/akatuoro.json"}]}}