Skip to content
J·BJon BishopFractional growth marketer

Data platforms · llms.txt file breakdown

Milvus llms.txt

A directly retrieved Markdown index for Milvus's documentation, with 456 links across 95 detected headings.

The actual file

Source Markdown

The saved llms.txt text, displayed as source. The file belongs to its publisher; the live endpoint may change.

Open saved text ↗

Showing the first 18000 characters. Load the complete file to read the remainder.

# Milvus

> Milvus is an open-source, high-performance vector database designed for similarity search and AI applications. It supports billion-scale vector storage and search across deployment modes: Milvus Lite (embedded, for prototyping), Milvus Standalone (single-node, for small-scale production), and Milvus Distributed (Kubernetes, for enterprise scale). The primary SDK is PyMilvus for Python; Java, Go, Node.js, and RESTful SDKs are also available.

This index is generated from the Milvus v3.0.x documentation tree. Every URL below resolves to a published page.

## URL rules

Read this before citing any milvus.io URL.

Documentation URLs take exactly one of these four forms:

    https://milvus.io/docs/<page-id>                    latest version (v3.0.x), English
    https://milvus.io/docs/<version>/<page-id>          pinned version
    https://milvus.io/docs/<lang>/<page-id>             translated, latest version
    https://milvus.io/docs/<lang>/<version>/<page-id>   translated, pinned version

- `<page-id>` always ends in `.md` — for example `overview.md`. The suffix is
  part of the route, not a file extension. `https://milvus.io/docs/overview` is a 404;
  `https://milvus.io/docs/overview.md` is the page.
- `<version>` is one of: v2.6.x, v2.5.x, v2.4.x. Omit the segment entirely for
  the latest version (v3.0.x) — there is no `/docs/v3.0.x/` path.
- `<lang>` is one of: zh, zh-hant, ja, ko, de, fr, es, it, pt, ru, ar, id. Omit the segment for English.
- API reference URLs are `https://milvus.io/api-reference/<sdk>/<version>/<page>.md` and
  are versioned separately from the docs; see the API Reference section.
- Blog, AI Quick Reference, and Learn pages do not use a `.md` suffix.

**Unknown paths return a hard 404 — they do not redirect and they do not fall
back to a search page.** A URL that looks right but was never published is a
dead link for the reader.

Therefore: cite only page ids that appear in this file. If the page you need is
not listed, say it is not in the index rather than assembling a plausible slug
from the topic name. `https://milvus.io/sitemap.xml` enumerates every published URL and
is the correct place to resolve anything this file does not cover.

## Rules for generating Milvus code

- Use the `MilvusClient` interface introduced in v2.4+. Do not use the legacy ORM API (`connections.connect()`, `Collection()`, `utility.list_collections()`) — it is deprecated and will be removed. If you find existing ORM code, advise upgrading the SDK and rewriting against `MilvusClient`.
- Check PyPI (`pip install --upgrade pymilvus`) or npm for the current SDK version rather than relying on a memorized version number.
- To change existing entities use `client.upsert()`. There is no `client.update()`. `upsert()` inserts when the primary key is absent and replaces the whole entity when it is present; use `client.insert()` only for data known not to collide.
- A collection schema is immutable in v2.5.x and earlier — to change it, drop and recreate the collection. From v2.6+ you may add scalar fields with `add_collection_field()`, but vector fields cannot be added and existing fields cannot be modified or removed. Changing a field's data type is not supported in any version. Check the server version before suggesting a schema change.
- Primary keys must be `DataType.INT64` or `DataType.VARCHAR`, must be unique across the whole collection including partitions, and cannot be composite.
- For BM25 full-text search, the BM25 function and its analyzer must be declared when the collection is created; they cannot be added later.
- A vector field must be indexed and the collection loaded before it can be searched.
- In `client.hybrid_search()`, each `AnnSearchRequest` takes exactly one query vector — use one request per vector field — and each search accepts exactly one ranker. Rankers cannot be chained.
- Search iterators (`with-iterators.md`) support basic ANN search only, not hybrid search.
- Both scalar and vector fields support `nullable=True`; only Array of Structs fields (and any field nested inside one) cannot be nullable. A nullable field cannot be a partition key, nullability is fixed at field creation, and a nullable vector field cannot be filtered with `IS NULL` / `IS NOT NULL`. See `nullable-and-default.md`, which is authoritative here.

## About Milvus

- [What is Milvus](https://milvus.io/docs/overview.md): Milvus is a high-performance, highly scalable vector database that runs efficiently across a wide range of environments, from a laptop to large-scale…
- [Milvus Adopters](https://milvus.io/docs/milvus_adopters.md): Learn about companies that have adopted Milvus in production applications.
- [Milvus Roadmap](https://milvus.io/docs/roadmap.md): Explore the Milvus roadmap for search, analytics, schema evolution, lake integration, and performance across the 3.0 and 3.1 releases.
- [Releases](https://milvus.io/docs/release_notes.md): Milvus Release Notes
- [Comparison](https://milvus.io/docs/comparison.md): This article compares Milvus with other vector search solutions.

## Get Started

Start here for first-time setup: run the Quickstart against Milvus Lite, then pick a deployment mode and install an SDK.

- [Quickstart](https://milvus.io/docs/quickstart.md): Get started with Milvus.
- [Connect to Milvus Server](https://milvus.io/docs/connect-to-milvus-server.md): This topic describes how to establish a client connection to a Milvus server and configure common connection options.

### Install Milvus

- [Overview](https://milvus.io/docs/install-overview.md): Milvus is a highly performant, scalable vector database.
- [Run Milvus Lite](https://milvus.io/docs/milvus_lite.md): Get started with Milvus Lite.

#### Run Milvus Standalone

- [Requirements](https://milvus.io/docs/prerequisite-docker.md): Learn the necessary preparations before installing Milvus Standalone.
- [RPM/DEB](https://milvus.io/docs/install_standalone-binary.md): Learn how to install Milvus standalone with a pre-built RPM/DEB package.
- [Docker (Linux)](https://milvus.io/docs/install_standalone-docker.md): Learn how to install Milvus standalone with Docker.
- [Docker Compose (Linux)](https://milvus.io/docs/install_standalone-docker-compose.md): Learn how to install Milvus standalone with Docker Compose.
- [Docker Desktop (Windows)](https://milvus.io/docs/install_standalone-windows.md): Learn how to install Milvus standalone with Docker Desktop for Windows.

#### Run Milvus Distributed

- [Requirements](https://milvus.io/docs/prerequisite-helm.md): Learn the necessary preparations before installing Milvus with Helm.
- [Milvus Operator](https://milvus.io/docs/install_cluster-milvusoperator.md): Learn how to install Milvus cluster on Kubernetes using Milvus Operator
- [Helm Chart](https://milvus.io/docs/install_cluster-helm.md): Learn how to install Milvus cluster on Kubernetes.

#### Run Milvus with GPU

- [Requirements](https://milvus.io/docs/prerequisite-gpu.md): Learn the necessary preparations before installing Milvus with GPU.
- [Helm Chart](https://milvus.io/docs/install_cluster-helm-gpu.md): Learn how to install Milvus cluster on Kubernetes.
- [Docker Compose](https://milvus.io/docs/install_standalone-docker-compose-gpu.md): Learn how to install Milvus cluster on Kubernetes.

### Install SDKs

- [PyMilvus](https://milvus.io/docs/install-pymilvus.md): Learn how to install the Python SDK of Milvus.
- [Java SDK](https://milvus.io/docs/install-java.md): Learn how to install the Java SDK of Milvus.
- [Go SDK](https://milvus.io/docs/install-go.md): Learn how to install the GO SDK of Milvus.
- [Node.js SDK](https://milvus.io/docs/install-node.md): Learn how to install the Node.js SDK of Milvus.

## Concepts

- [Bitset](https://milvus.io/docs/bitset.md): Learn about bitsets in Milvus.
- [Multi-tenancy](https://milvus.io/docs/multi_tenancy.md): In Milvus, multi-tenancy means multiple customers or teams—referred to as tenants— share the same cluster while maintaining isolated data environments.
- [Timestamp](https://milvus.io/docs/timestamp.md): Learn about the concept of timestamp and the four main timestamp-related parameters in the Milvus vector database.
- [Time Synchronization](https://milvus.io/docs/time_sync.md): Learn about the time synchronization system in Milvus.
- [Metric Types](https://milvus.io/docs/metric.md): Similarity metrics are used to measure similarities among vectors.
- [In-memory Replica](https://milvus.io/docs/replica.md): Learn about in-memory replica in Milvus.
- [Terminology](https://milvus.io/docs/glossary.md): Glossary of Milvus-specific terms and definitions.

### Architecture

- [Overview](https://milvus.io/docs/architecture_overview.md): Milvus provides a fast, reliable, and stable vector database built specifically for similarity search and artificial intelligence.
- [Main Components](https://milvus.io/docs/main_components.md): Learn about the main components in Milvus standalone and cluster.
- [Streaming Service](https://milvus.io/docs/streaming_service.md): The Streaming Service is a concept for Milvus internal streaming system module, built around the Write-Ahead Log (WAL) to support various streaming-related…
- [Data Processing](https://milvus.io/docs/data_processing.md): Learn about the data processing procedure in Milvus.
- [Storage V3](https://milvus.io/docs/storage-v3.md): Learn which Milvus 3.0 features require Storage V3, how to enable it, and which compatibility limits apply.
- [Woodpecker](https://milvus.io/docs/woodpecker_architecture.md): Woodpecker is a cloud-native WAL system in Milvus 2.6.
- [Knowhere](https://milvus.io/docs/knowhere.md): Learn about Knowhere in Milvus.

## User Guide

- [Database](https://milvus.io/docs/manage_databases.md): Milvus introduces a database layer above collections, providing a more efficient way to manage and organize your data while supporting multi-tenancy.

### Collections

- [Collection Explained](https://milvus.io/docs/manage-collections.md): On Milvus, you can create multiple collections to manage your data, and insert your data as entities into the collections.
- [Create Collection](https://milvus.io/docs/create-collection.md): You can create a collection by defining its schema, index parameters, metric type, and whether to load it upon creation.
- [Create an External Collection](https://milvus.io/docs/create-an-external-collection.md): An external collection is a type of data collection in Milvus that accesses data from external storage systems or database tables such as AWS S3 and Iceberg…
- [View Collections](https://milvus.io/docs/view-collections.md): You can obtain the name list of all the collections in the currently connected database, and check the details of a specific collection.
- [Modify Collection](https://milvus.io/docs/modify-collection.md): You can rename a collection or change its settings. This page focuses on how to modify a collection.
- [Load & Release](https://milvus.io/docs/load-and-release.md): Loading a collection is the prerequisite to conducting similarity searches and queries in collections.
- [Set Collection TTL](https://milvus.io/docs/set-collection-ttl.md): Configure collection-level or entity-level TTL policies to expire stale data automatically in Milvus.
- [Set Consistency Level](https://milvus.io/docs/tune_consistency.md): As a distributed vector database, Milvus offers multiple levels of consistency to ensure that each node or replica can access the same data during read and…
- [Manage Partitions](https://milvus.io/docs/manage-partitions.md): A partition is a subset of a collection.
- [Manage Aliases](https://milvus.io/docs/manage-aliases.md): Milvus provides alias management capabilities. This page demonstrates the procedures to create, list, alter, and drop aliases.
- [Drop Collection](https://milvus.io/docs/drop-collection.md): You can drop a collection if it is no longer needed.

### Schema & Data Fields

Schema changes are heavily constrained — see the schema and primary-key rules above before designing a collection.

- [Schema Explained](https://milvus.io/docs/schema.md): A schema defines the data structure of a collection. Before creating a collection, you need to work out a design of its schema.
- [Primary Field & AutoID](https://milvus.io/docs/primary-field.md): Every collection in Milvus must have a primary field to uniquely identify each entity.
- [Dense Vector](https://milvus.io/docs/dense-vector.md): Dense vectors are numerical data representations widely used in machine learning and data analysis.
- [Binary Vector](https://milvus.io/docs/binary-vector.md): Binary vectors are a special form of data representation that convert traditional high-dimensional floating-point vectors into binary vectors containing only…
- [Sparse Vector](https://milvus.io/docs/sparse_vector.md): Sparse vectors are an important method of capturing surface-level term matching in information retrieval and natural language processing.
- [VarChar Field](https://milvus.io/docs/string.md): VARCHAR is a scalar field type for storing short, bounded string metadata in Milvus.
- [Text Field](https://milvus.io/docs/text.md): TEXT is a scalar field type for storing document text, passages, and other long text content in Milvus.
- [Number Field](https://milvus.io/docs/number.md): A number field is a scalar field that stores numeric values.
- [Array Field](https://milvus.io/docs/array_data_type.md): An ARRAY field stores an ordered set of elements of the same data type. Here's an example of how ARRAY fields store data:
- [Geometry Field](https://milvus.io/docs/geometry-field.md): When building applications like Geographic Information Systems (GIS), mapping tools, or location-based services, you often need to store and query geometric…
- [TIMESTAMPTZ Field](https://milvus.io/docs/timestamptz-field.md): Applications that track time across regions, such as e-commerce systems, collaboration tools, or distributed logging, need precise handling of timestamps with…
- [Dynamic Field](https://milvus.io/docs/enable-dynamic-field.md): Milvus allows you to insert entities with flexible, evolving structures through a special feature called the dynamic field.
- [Nullable Fields](https://milvus.io/docs/nullable-and-default.md): Configure nullable fields and default values, including schema, insert, index, search, and filter behavior.
- [Default Values](https://milvus.io/docs/default-values.md): Set default values for scalar fields so Milvus fills missing values during entity insertion.
- [Alter Collection Field](https://milvus.io/docs/alter-collection-field.md): You can alter the properties of a collection field to change column constraints or enforce stricter data integrity rules.
- [Alter Collection Schema](https://milvus.io/docs/add-fields-to-an-existing-collection.md): Alter an existing collection schema by adding or dropping user-defined fields or Functions with their generated vector fields.
- [Alter External Collection Schema](https://milvus.io/docs/alter-external-collection-schema.md): Learn how to expose an additional field from an external data source in an existing external collection.

#### JSON Field

- [JSON Field Overview](https://milvus.io/docs/json-field-overview.md): When building applications like product catalogs, content management systems, or user preference engines, you often need to store flexible metadata alongside…
- [JSON Indexing](https://milvus.io/docs/json-indexing.md): JSON fields provide a flexible way to store structured metadata in Milvus.
- [JSON Shredding](https://milvus.io/docs/json-shredding.md): JSON shredding accelerates JSON queries by converting traditional row-based storage into optimized columnar storage.

#### StructArray

- [StructArray Overview](https://milvus.io/docs/array-of-structs.md): Use StructArray when one entity needs to store an ordered list of structured elements, such as one document with many chunks, one page with many visual…
- [Create a StructArray Field](https://milvus.io/docs/create-structarray-field.md): Create a StructArray field when one entity needs to contain an ordered list of structured elements.
- [Insert Data into StructArray Fields](https://milvus.io/docs/insert-data-into-structarray-fields.md): Insert data into a StructArray field when each entity contains an ordered list of structured elements.
- [Index StructArray Fields](https://milvus.io/docs/index-structarray-fields.md): Create indexes on StructArray subfields before you run vector search or accelerate scalar filtering.
- [StructArray Limits](https://milvus.io/docs/structarray-limits.md): StructArray support spans schema definition, insert payloads, indexing, search modes, and StructArray-specific filters.

#### Analyzer

- [Analyzer Overview](https://milvus.io/docs/analyzer-overview.md): In text processing, an analyzer is a crucial component that converts raw text into a structured, searchable format.
- [Multi-language Analyzers](https://milvus.io/docs/multi-language-analyzers.md): When Milvus performs text analysis, it typically applies a single analyzer across an entire text field in a collection.
- [Choose the Right Analyzer for Your Use Case](https://milvus.io/docs/choose-the-right-analyzer-for-your-use-case.md): Notes
- [Manage File Resources](https://milvus.io/docs/manage-file-resources.md): Register and manage external dictionary files that Milvus text analyzers can load at runtime.

##### Built-in Analyzers

- [Standard](https://milvus.io/docs/standard-analyzer.md): The standard analyzer is the default analyzer in Milvus, which is automatically applied to text fields if no analyzer is specified.
- [English](https://milvus.io/docs/english-analyzer.md): The english analyzer in Milvus is designed to process English text, applying language-specific rules for tokenization and filtering.
- [Chinese](https://milvus.io/docs/chinese-analyzer.md): The chinese analyzer is designed specifically to handle Chinese text, providing effective segmentation and tokenization.
- [A

Measured from the saved file

Linked document

UTF-8 text size
80.5 KB82465 bytes
Token count
19243o200k_base encoding
Markdown links
456456 distinct URLs
Fenced code blocks
0799 lines in the file

Size measures the captured UTF-8 text, not compressed transfer size. Tokens are counted with a named tokenizer; other models can count differently. Neither measure establishes context compatibility or answer quality.

Project H1
Milvus

Opening blockquote

Milvus is an open-source, high-performance vector database designed for similarity search and AI applications. It supports billion-scale vector storage and search across deployment modes: Milvus Lite (embedded, for prototyping), Milvus Standalone (single-node, for small-scale production), and Milvus Distributed (Kubernetes, for enterprise scale). The primary SDK is PyMilvus for Python; Java, Go, Node.js, and RESTful SDKs are also available.

Context architecture

How the content is divided

18 H2 sections organize 456 parsed Markdown links. 439 links point to .md or .markdown paths; 2 link occurrences use another hostname. Linked destinations have not been recursively validated.

H2 sectionLinksUTF-8 bytesTokens
URL rulesLine 701634438
Rules for generating Milvus codeLine 3702088454
About MilvusLine 505735173
Get StartedLine 58192623630
ConceptsLine 97141955454
User GuideLine 117173323037115
Data ImportLine 382230563
AI ToolsLine 38771218282
Administration GuideLine 4025983832061
ToolsLine 5274065641586
IntegrationsLine 59683156093810
TutorialsLine 715244025933
FAQsLine 7455730155
API ReferenceLine 75381002306
Interactive LearningLine 7665705190
Sizing ToolsLine 7763361101
Show 2 more sections
H2 sectionLinksUTF-8 bytesTokens
External ToolsLine 782231172
Other Site SectionsLine 78971041241

Link counts exclude parsed fenced and inline code and resolve relative destinations against the final endpoint. Section sizes cover the text below each H2 until the next H2.

Embedded material

Code, installation and API examples

Fenced blocks
0 closed
Declared languages
No fenced blocks detected
Package-install commands
1 detected in fenced or inline code
Authentication examples
0 fenced lines match API-key, Authorization, Bearer, OAuth or access-token patterns
HTTP request examples
0 fenced lines match request-method patterns
Authentication links
0 links have authentication-related labels

Embedded example text and links to further documentation are counted separately. A code block does not establish that a sample runs or that an authentication flow succeeds.

Detected installation lines

Source line 40 · inline

pip install --upgrade pymilvus

Publisher prioritization

The Optional section

No H2 section named “Optional” was detected in this file.

The community proposal uses Optional by convention for secondary links that can be skipped for a shorter context. This label does not establish that everything outside it is critical, or that the material inside it is deprecated.

Machine-readable routes

MCP references and configuration text

MCP-related text was detected in the captured file. 2 links have MCP-related labels or paths. 0 fenced lines match mcpServers, mcp_servers or an mcp add command.

These are text and link observations. No MCP server connection, authentication or tool call was tested.

Index and bundle

llms-full.txt companion

Not found at checked URLs

A verified text response establishes a companion at the checked URL. It does not prove the bundle contains all product documentation, that generation is automated, or that it fits an assistant’s context window.

Change evidence

What can be said about freshness

This capture establishes a baseline. Update frequency cannot be determined from one observation.

Linked release and migration material

2 links have release, changelog, deprecation or migration terms in their labels or paths. These provide places to investigate product changes separately.

File changes can reveal documentation edits or navigation changes. They do not by themselves measure product release velocity, accuracy, agent adoption or AI visibility.

Capture identity and measurement method
Requested endpoint
https://milvus.io/llms.txt
Final endpoint
https://milvus.io/llms.txt
HTTP status
200
Content type
text/plain; charset=UTF-8
Text retrieval
Firecrawl literal text capture
Tokenizer
tiktoken 0.14.0 · o200k_base
SHA-256 of saved UTF-8 text
350034783a4794af59f745596294a5af27c0cc000a4f64793e31e1d7d3aafd9a

Counts use a deterministic parser, not a spec-compliance grade. Unusual Markdown and indented code can fall outside its recognized patterns. See the llms.txt proposal and tokenizer implementation.

Related tools and resources

Contact Jon