NEWVectors or files. Pick a path.Start →

    Guides

    Vendor-neutral, engineer-written guides to the concepts behind multimodal AI: perception, retrieval, embeddings, and the infrastructure agents use to see, hear, and search unstructured data. Learn the idea first; then see how Mixpeek applies it.

    121 guides across 18 topics

    Architecture
    10 min read

    What Is a Semantic Contract? The Interface Between Extraction and Retrieval

    A semantic contract is what an application binds to instead of an index: what one result unit is, which model produced it, what the score means, which fields are filterable and at what path, how fresh it is, and what happens when a stage fails. Includes what counts as a breaking change and a worked declaration.

    Semantic Layer
    Architecture
    Retrieval Architecture
    Aug 2026Read guide
    Architecture
    11 min read

    What Does a Semantic Layer Compile To When the Target Is Not SQL?

    A structured semantic layer compiles a named concept into SQL. Over unstructured data it compiles a named capability into a retrieval plan. The operator set, the guarantees a retrieval plan gives up, why the cost model is inverted, why the optimizer has no statistics, and why text-to-SQL does not cover this.

    Semantic Layer
    Query Planning
    Architecture
    Aug 2026Read guide
    Agent Architecture
    9 min read

    How Should an AI Agent Choose Between Retrieval Capabilities?

    Capability-aware retrieval gives an agent a short roster of named, described, costed retrieval contracts instead of raw index access. What each declaration has to carry, how raw vector-search tools degrade silently, the three ways capability selection goes wrong, and how to make a bad choice observable.

    Agent Perception
    Agentic Retrieval
    Semantic Layer
    Aug 2026Read guide
    Architecture
    10 min read

    Should Semantic Features Be Materialized or Computed at Query Time?

    The latency arithmetic that decides it, what the query-time path can actually afford, why staleness and re-extraction are the price of materializing, the hybrid split every production system lands on, and the three problems live semantic federation across arbitrary sources still has to solve.

    Semantic Layer
    Architecture
    Feature Extraction
    Aug 2026Read guide
    Data Infrastructure
    11 min read

    How Do You Model Unstructured Data From Raw Objects to Serving Contracts?

    Five layers, each rebuildable from the one below: raw objects, units, features, entities, contracts. What each owns, why stable unit ids are the whole trick, a blast-radius table for every kind of change, how to keep a re-run from re-paying full extraction, and the layer almost everyone skips.

    Semantic Layer
    Data Modeling
    Architecture
    Aug 2026Read guide
    Architecture
    11 min read

    What a Semantic Layer Looks Like in Production

    The semantic-layer architecture for unstructured data, mapped onto one shipped video-search deployment layer by layer: what each layer owns, what it prevented, and the three places this deployment does not prove the thesis. Uses Cutsio's RED R3D and ARRI RAW pipeline, with their published numbers.

    Semantic Layer
    Architecture
    Retrieval Architecture
    Aug 2026Read guide
    Copyright & IP
    17 min read

    How to Check if a Video Is Copyrighted

    Five ways to check whether a video is copyrighted before you publish it, compared on what each one catches, what it costs and where it fails: YouTube Content ID, manual rights clearance, frame-by-frame reverse image search, audio fingerprinting, and automated multimodal scanning.

    Copyright
    Video
    IP Safety
    Aug 2026Read guide
    Infrastructure
    12 min read

    Compute Pushdown: Why Your Vector Search Over Object Storage Is Slow

    Searching data that lives in S3 usually means downloading it first. Compute pushdown inverts that: send the predicate to the bytes instead of the bytes to the predicate. Covers what can and cannot be pushed down, why the network is the bottleneck rather than the math, the prior art from Parquet and S3 Select to Aurora, and how to tell whether pushdown would actually help your workload.

    Compute Pushdown
    Object Storage
    Vector Search
    Aug 2026Read guide
    Infrastructure
    14 min read

    How to Merge Data Sources With Different Schemas Into One Searchable Index

    Your content lives in three systems and none of them agree on field names. One calls it headline, one calls it name, one calls it asset_title. Covers the four ways teams reconcile that for search, how to design a canonical field set you will not regret, what to do when two sources claim the same target field, how to keep lineage back to the origin record after the merge, and when merging is the wrong answer.

    Data Integration
    Schema Mapping
    Ingestion
    Aug 2026Read guide
    Infrastructure
    15 min read

    Write-Ahead Logs on Object Storage: How S3 Becomes a Database's Durability Layer

    A write-ahead log acknowledges a write the moment it is durable, before any index work happens. Move that log onto S3 or GCS and the bucket becomes the durability layer: recovery, read replicas and change feeds all become readers of one chain of sequenced objects. Covers what a WAL is, why object storage changes the design, the two coordination problems it creates (writer fencing and garbage collection that could delete the only copy of a write), how conditional writes solve them, and the shipped write path in Mixpeek MVS.

    Write-Ahead Log
    Object Storage
    S3
    Aug 2026Read guide
    Retrieval
    16 min read

    Query Planning for Multimodal Retrieval: Stage Order, Filter Pushdown, and Candidate Budgets

    Why the same five retrieval stages can differ by two orders of magnitude in cost depending only on their order and their candidate budgets. Covers pre-filter versus post-filter and why naive filtering destroys recall on a graph index, how to measure selectivity when your vector store keeps no statistics, how to size the reranker budget from a recall curve instead of a guess, deduplicating before the expensive stage, and what changes when the rows are video scenes rather than text chunks.

    Query Planning
    Vector Search
    Filtering
    Aug 2026Read guide
    Retrieval
    14 min read

    Multi-Stage Retrieval: Retrieve, Filter, Rerank

    How a production search pipeline chains retrieval, filtering and reranking, what each stage costs, how many candidates to pass between them, and whether to filter before or after the vector search.

    Multi-Stage Retrieval
    Reranking
    Filtering
    Aug 2026Read guide

    All guides

    Every guide in one place, 121 in total.

    From concept to production

    These guides explain how multimodal perception and retrieval actually work. Mixpeek is the platform that runs them: point it at your storage and get back relevant, timestamped results.