--- title: [Hands-on] Build Semantic Search Inside Your Database... description: A newsletter covering three topics: (1) MongoDB Atlas auto-embedding with Voyage AI that eliminates the need for a separate embedding pipeline — you configure... canonical: https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: [Hands-on] Build Semantic Search Inside Your Database Without an Embedding Pipeline | daily.dev og:description: A newsletter covering three topics: (1) MongoDB Atlas auto-embedding with Voyage AI that eliminates the need for a separate embedding pipeline — you configure... og:url: https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g og:image: https://api.daily.dev/og/posts/1ifRdsT4G.png og:image:alt: [Hands-on] Build Semantic Search Inside Your Database Without an Embedding Pipeline og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # [Hands-on] Build Semantic Search Inside Your Database Without an Embedding Pipeline **[Daily Dose of Data Science \| Avi Chawla \| Substack](https://daily.dev/sources/dailydoseofds)** · 9 min read · 1 upvotes · 0 comments ## Summary A newsletter covering three topics: (1) MongoDB Atlas auto-embedding with Voyage AI that eliminates the need for a separate embedding pipeline — you configure a vector search index pointing at a text field and Atlas handles embedding generation and updates automatically, demonstrated with 21,000 movie plots; (2) a detailed breakdown of 8 LLM precision formats (FP32, TF32, BF16, FP16, FP8, INT8, INT4, NF4) explaining their bit layouts, trade-offs, and use cases including why 4-bit became the default in tools like Ollama and vLLM; (3) an overview of memory optimization techniques for agentic systems using LangGraph and CrewAI, covering short-term, long-term, entity, contextual, and user memory types. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I set up semantic search in MongoDB Atlas without building my own embedding pipeline? MongoDB Atlas supports auto-embedding, where you create a vector search index with an autoEmbed field type pointing to a text field, specifying a Voyage AI model such as voyage-4. Atlas generates and stores the vectors automatically and re-embeds documents when they change, removing the need for an external embedding service, separate vector store, or sync logic. _Teams wiring up semantic search track database feature rollouts like this on daily.dev._ ### Why does a 7B parameter LLM run on a 12GB consumer GPU when FP32 weights alone would need 28GB? Tools like Ollama do not load models in FP32; by default Ollama pulls a Q4_K_M GGUF quantization averaging about 4.8 bits per weight, shrinking a 7B model to roughly 4.1GB. This is why running a 7B model with a command like ollama run llama3 works fine on hardware that could never hold the full-precision 28GB weights. _Developers sizing GPUs for local LLMs follow quantization tradeoffs like these on daily.dev._ ### What is the difference between GPTQ and AWQ quantization methods for INT4 weights? GPTQ quantizes a block one weight at a time, adjusting the not-yet-quantized weights to cancel out error already introduced, while AWQ instead finds the channels carrying the most signal and scales them up before quantizing so more of the 16 available INT4 values are spent where they matter. Both require a few hundred real calibration inputs to identify which weights matter most. _Engineers choosing a quantization method for deployment compare tradeoffs like these on daily.dev._ ## Similar posts on daily.dev - [MongoDB Introduces Embedding and Reranking API on Atlas](https://daily.dev/posts/mongodb-introduces-embedding-and-reranking-api-on-atlas-ow0bpzewc) · InfoQ · 0 upvotes · 0 comments - [MongoDB targets AI’s retrieval problem](https://daily.dev/posts/mongodb-targets-ai-s-retrieval-problem-fdiya16um) · InfoWorld · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#mongodb](https://daily.dev/tags/mongodb), [#vector-search](https://daily.dev/tags/vector-search) [View this post on daily.dev](https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"[Hands-on] Build Semantic Search Inside Your Database Without an Embedding Pipeline","url":"https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g"},"datePublished":"2026-08-07T20:46:47.707Z","dateModified":"2026-09-13T19:55:58.948Z","description":"A newsletter covering three topics: (1) MongoDB Atlas auto-embedding with Voyage AI that eliminates the need for a separate embedding pipeline — you configure...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/e3f3a9cae7f2bf5b17d399056db45025?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/e3f3a9cae7f2bf5b17d399056db45025?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Daily Dose of Data Science | Avi Chawla | Substack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Daily Dose of Data Science | Avi Chawla | Substack","logo":"https://media.daily.dev/image/upload/s--4IHQgTOw--/f_auto/v1710503712/logos/dailydoseofds","url":"https://daily.dev/sources/dailydoseofds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,mongodb,vector-search","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Daily Dose of Data Science | Avi Chawla | Substack","item":"https://daily.dev/sources/dailydoseofds"},{"@type":"ListItem","position":3,"name":"[Hands-on] Build Semantic Search Inside Your Database Without an Embedding Pipeline"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/hands-on-build-semantic-search-inside-your-database-without-an-embedding-pipeline-1ifrdst4g#faq","mainEntity":[{"@type":"Question","name":"How do I set up semantic search in MongoDB Atlas without building my own embedding pipeline?","acceptedAnswer":{"@type":"Answer","text":"MongoDB Atlas supports auto-embedding, where you create a vector search index with an autoEmbed field type pointing to a text field, specifying a Voyage AI model such as voyage-4. Atlas generates and stores the vectors automatically and re-embeds documents when they change, removing the need for an external embedding service, separate vector store, or sync logic. Teams wiring up semantic search track database feature rollouts like this on daily.dev."}},{"@type":"Question","name":"Why does a 7B parameter LLM run on a 12GB consumer GPU when FP32 weights alone would need 28GB?","acceptedAnswer":{"@type":"Answer","text":"Tools like Ollama do not load models in FP32; by default Ollama pulls a Q4_K_M GGUF quantization averaging about 4.8 bits per weight, shrinking a 7B model to roughly 4.1GB. This is why running a 7B model with a command like ollama run llama3 works fine on hardware that could never hold the full-precision 28GB weights. Developers sizing GPUs for local LLMs follow quantization tradeoffs like these on daily.dev."}},{"@type":"Question","name":"What is the difference between GPTQ and AWQ quantization methods for INT4 weights?","acceptedAnswer":{"@type":"Answer","text":"GPTQ quantizes a block one weight at a time, adjusting the not-yet-quantized weights to cancel out error already introduced, while AWQ instead finds the channels carrying the most signal and scales them up before quantizing so more of the 16 available INT4 values are spent where they matter. Both require a few hundred real calibration inputs to identify which weights matter most. Engineers choosing a quantization method for deployment compare tradeoffs like these on daily.dev."}}]} ```