From 2ff192129aebd523a5e38d681985f2edffdcc6d8 Mon Sep 17 00:00:00 2001
From: Mendon Kissling <59585235+mendonk@users.noreply.github.com>
Date: Fri, 19 Jun 2026 13:11:25 -0400
Subject: [PATCH] docs: add lfx-nextplaid bundle (#13679)
---
docs/docs/Components/bundles-nextplaid.mdx | 121 +++++++++++++
docs/docs/Components/bundles-vllm.mdx | 11 ++
docs/docs/Support/release-notes.mdx | 201 +--------------------
docs/sidebars.js | 1 +
4 files changed, 140 insertions(+), 194 deletions(-)
create mode 100644 docs/docs/Components/bundles-nextplaid.mdx
diff --git a/docs/docs/Components/bundles-nextplaid.mdx b/docs/docs/Components/bundles-nextplaid.mdx
new file mode 100644
index 0000000000..e9699ca79e
--- /dev/null
+++ b/docs/docs/Components/bundles-nextplaid.mdx
@@ -0,0 +1,121 @@
+---
+title: NextPlaid
+slug: /bundles-nextplaid
+---
+
+import Icon from "@site/src/components/icon";
+import PartialParams from '@site/docs/_partial-hidden-params.mdx';
+import PartialVectorSearchResults from '@site/docs/_partial-vector-search-results.mdx';
+import PartialVectorStoreInstance from '@site/docs/_partial-vector-store-instance.mdx';
+
+ [**Bundles**](/components-bundle-components) contain custom components that support specific third-party integrations with Langflow.
+
+The **NextPlaid** bundle provides multi-vector ColBERT-style retrieval for Langflow through a running [NextPlaid](https://github.com/meetdoshi90/next-plaid) server.
+NextPlaid stores each document as a matrix of token embeddings, which enables higher retrieval quality on semantic search tasks through ColBERT-style late interaction with MaxSim scoring.
+For more information on multi-vector embeddings, see the [langchain-plaid](https://github.com/meetdoshi90/langchain-plaid) package that this bundle is built on.
+
+## Install the NextPlaid bundle
+
+The bundle includes two components:
+
+* **NextPlaid**: vector store component backed by a running NextPlaid server.
+* **vLLM Multivector Embeddings**: generates the multi-vector token embeddings required by NextPlaid.
+
+The **NextPlaid** bundle is included in the `lfx-nextplaid` Extension bundle, which is installed automatically as part of `uv pip install langflow`.
+
+If you need to install it separately, run:
+
+```bash
+uv pip install lfx-nextplaid
+uv run langflow run
+```
+
+To verify the bundle is loaded in your environment:
+
+```bash
+lfx extension list
+```
+
+## NextPlaid
+
+The **NextPlaid** component reads from and writes to a [NextPlaid](https://github.com/meetdoshi90/next-plaid) multi-vector search server.
+
+NextPlaid is a Rust-based server that implements ColBERT-style late interaction retrieval with MaxSim scoring.
+Each document is stored as a matrix of token embeddings rather than a single dense vector, giving better retrieval quality on semantic search tasks.
+The component supports both text retrieval with ColBERT models and image retrieval with ColPali models.
+
+
+About vector store instances
+
+
+
+
+
+
+
+### Use the NextPlaid component in a flow
+
+Connect a **vLLM Multivector Embeddings** component to the **Embedding (Multivector)** input. Standard single-vector embedding components are not compatible with NextPlaid.
+
+The **NextPlaid** component can be used for both writes and reads:
+
+* When writing, it ingests documents from an attached data source, computes multi-vector embeddings with the connected **vLLM Multivector Embeddings** component, and then loads them into the NextPlaid index.
+ To trigger writes, click **Run component** on the **NextPlaid** component.
+
+* When reading, the **NextPlaid** component uses chat input to perform a MaxSim similarity search against the index and returns the top results.
+
+### NextPlaid parameters
+
+
+
+| Name | Type | Description |
+|------|------|-------------|
+| **Server URL** (`url`) | String | Input parameter. Base URL of the running NextPlaid server. Default: `http://localhost:8080`. |
+| **Index Name** (`index_name`) | String | Input parameter. Name of the index to create or connect to. Default: `langflow`. |
+| **Ingest Data** (`ingest_data`) | Data | Input parameter. Documents or images to write to the vector store. Only relevant for writes. |
+| **Search Query** (`search_query`) | String | Input parameter. The query string to use for similarity search. Only relevant for reads. |
+| **Embedding (Multivector)** (`embedding`) | Embeddings | Input parameter. Connect a **vLLM Multivector Embeddings** component to generate token-level embeddings. Required for both reads and writes. |
+| **Index Batch Size** (`index_batch_size`) | Integer | Input parameter. Number of documents per indexing request. PLAID builds its initial cluster centroids from the first batch — larger batches produce better retrieval quality. Default: `500`. |
+| **Number of Results** (`number_of_results`) | Integer | Input parameter. Number of results to return from similarity search. Default: `4`. |
+| **Quantization Bits** (`nbits`) | Dropdown | Input parameter. Bit-width for PLAID quantization. `4` gives better quality; `2` uses less memory. Options: `2`, `4`. Default: `4`. |
+| **Create Index If Not Exists** (`create_index_if_not_exists`) | Boolean | Input parameter. If `true`, creates the index on the NextPlaid server if it does not already exist. Default: `true`. |
+| **Write Timeout** (`write_timeout`) | Float | Input parameter. Seconds to wait for each indexing batch to finish. Set to `0` for async indexing (search may return empty results on the first run). Recommended: `30` or higher when ingesting and searching in the same flow run. Default: `30.0`. |
+
+## vLLM Multivector Embeddings
+
+The **vLLM Multivector Embeddings** component generates multi-vector token embeddings by calling vLLM's `/pooling` endpoint with `task: token_embed`.
+
+The output is a multi-vector `Embeddings` object that returns a matrix of token embeddings per document, which is required by the **NextPlaid** vector store. It is not compatible with standard single-vector stores.
+
+For more information about using embedding model components in flows, see [Embedding model components](/components-embedding-models).
+
+Your vLLM server must be started with a ColBERT- or ColPali-compatible model and the pooling runner enabled:
+
+```bash
+vllm serve --runner pooling --pooler-config '{"task": "token_embed"}'
+```
+
+Compatible models include:
+* **Text (ColBERT):** `answerdotai/answerai-colbert-small-v1`
+* **Text + Images (ColPali):** `ModernVBERT/colmodernvbert`
+
+For more information on the vLLM server, see the [vLLM documentation](https://docs.vllm.ai/).
+
+### vLLM Multivector Embeddings parameters
+
+
+
+| Name | Type | Description |
+|------|------|-------------|
+| **Model Name** (`model_name`) | String | Input parameter. The multi-vector model name served by vLLM. Default: `answerdotai/answerai-colbert-small-v1`. |
+| **vLLM API Base** (`api_base`) | String | Input parameter. Base URL of the vLLM server (without the `/v1` suffix). Default: `http://localhost:8000`. |
+| **API Key** (`api_key`) | SecretString | Input parameter. API key for the vLLM server. Leave empty for local servers. Optional. |
+| **Request Timeout** (`request_timeout`) | Float | Input parameter. Timeout in seconds for each request to the vLLM API. Default: `60.0`. Advanced. |
+| **Max Retries** (`max_retries`) | Integer | Input parameter. Number of times to retry a failed request before raising an error. Default: `3`. Advanced. |
+
+## See also
+
+* [**Embedding model** components](/components-embedding-models)
+* [NextPlaid server](https://github.com/meetdoshi90/next-plaid)
+* [langchain-plaid](https://github.com/meetdoshi90/langchain-plaid)
+* [vLLM documentation](https://docs.vllm.ai/)
diff --git a/docs/docs/Components/bundles-vllm.mdx b/docs/docs/Components/bundles-vllm.mdx
index 41b6bc92c4..d7f7adf4c6 100644
--- a/docs/docs/Components/bundles-vllm.mdx
+++ b/docs/docs/Components/bundles-vllm.mdx
@@ -56,8 +56,19 @@ To use the vLLM component, you need to have a vLLM server running. Here are the
For more detailed setup instructions, see the [vLLM documentation](https://docs.vllm.ai/en/latest/getting_started/quickstart.html).
+## vLLM multi-vector embeddings
+
+For multi-vector ColBERT or ColPali-style retrieval with vLLM, use the **vLLM Multivector Embeddings** component.
+
+The **vLLM Multivector Embeddings** component ships with the **NextPlaid** bundle (`lfx-nextplaid`), since it is purpose-built to feed that vector store.
+
+For more information on using the component in a flow, see [vLLM Multi-vector Embeddings](./bundles-nextplaid#vllm-multivector-embeddings).
+
+For more information on multi-vector embeddings, see the [langchain-plaid repository](https://github.com/meetdoshi90/langchain-plaid).
+
## See also
* [**Agent** component](/components-agents)
* [**Language Model** component](/components-models)
+* [**NextPlaid** bundle](./bundles-nextplaid)
* [vLLM GitHub repository](https://github.com/vllm-project/vllm)
diff --git a/docs/docs/Support/release-notes.mdx b/docs/docs/Support/release-notes.mdx
index c25058be49..56764e607c 100644
--- a/docs/docs/Support/release-notes.mdx
+++ b/docs/docs/Support/release-notes.mdx
@@ -46,211 +46,24 @@ To avoid the impact of potential breaking changes and test new versions, the Lan
If you made changes to your flows in the isolated installation, you might want to export and import those flows back to your upgraded primary installation so you don't have to repeat the component upgrade process.
-## 1.10.x
+## 1.11.x
Highlights of this release include the following changes.
For all changes, see the [Changelog](https://github.com/langflow-ai/langflow/releases).
-
-### Breaking changes
-
-- Upcoming breaking change: bundle separation
-
- Langflow 1.10.x introduces the Extension bundles model.
- In future releases, bundle separation will extend to more component providers.
-
- Saved flows will continue to open normally, but the following may require manual updates before upgrading:
-
- - Code or scripts that reference component class names or `lfx.components.` import paths.
- - External tooling that reads component type identifiers from raw flow JSON.
- - Custom deployment scripts that install or pin individual component packages.
-
- For more information, see [Langflow Extensions overview](../Develop/extensions-overview.mdx).
-
### New features and enhancements
-- **Langflow Assistant**: build complete flows
+- NextPlaid multi-vector bundle
- **Langflow Assistant** can now build entire flows, and not only individual components.
+ The **NextPlaid** bundle adds two new components for ColBERT-style multi-vector retrieval.
- For more information, see [Build flows and components with Langflow Assistant](/langflow-assistant).
+ The **NextPlaid** vector store, backed by a running [NextPlaid](https://github.com/meetdoshi90/next-plaid) server, stores each document as a matrix of token embeddings, and the **vLLM Multivector Embeddings** component generates the token-level multi-vector embeddings required by NextPlaid.
-- Redis-backed job queue for multi-worker deployments
+ For more information, see [NextPlaid bundle](./bundles-nextplaid).
- Langflow now supports a Redis-backed job queue that allows flow build events to be shared across multiple Gunicorn/Uvicorn workers and across multiple replicas behind a load balancer.
+## 1.10.x
- Single-process deployments are unaffected. The default `asyncio` in-memory queue is unchanged.
-
- For setup instructions and configuration details, see [Deploy Langflow with multiple workers](./deployment-multi-worker).
-
-- Memory bases
-
- Memory bases are per-flow vector stores that automatically ingest conversation messages.
- The **Memory Base** component retrieves context from the vector store, offering long-term semantic memory for flows.
- For more information, see [Manage memory bases](../Develop/memory-bases.mdx).
-
-- Python 3.14 support
-
- Langflow now supports Python 3.10 through 3.14 on macOS, Linux, and Windows.
- The Langflow Docker images now use Python 3.14.
-
-
- Optional integrations not yet available on Python 3.14
-
- The following optional integrations are excluded from installs on Python 3.14:
-
- - [IBM watsonx](/bundles-ibm)
- - [IBM watsonx Orchestrate](/deployment-wxo)
- - [ALTK](/bundles-altk)
- - [CUGA](/bundles-cuga)
- - [LangWatch](/integrations-langwatch)
- - [Pinecone](/bundles-pinecone)
- - OpenDsStar
-
-
-
-- **Agent** component: structured response output
-
- The **Agent** component now includes a **Structured Response** (`structured_response`) output that returns the agent's reply as structured data according to an **Output Schema** you define.
-
- For more information, see [Agent component output](/agents#agent-component-output).
-
-- **Agent** component: default system prompt
-
- The **Agent** component now uses a structured default system prompt that includes the current date and the active model name.
- Existing flows are not affected.
-
- For more information, see [Agent instructions and input](/agents#agent-instructions-and-input).
-
-- Extension bundles
-
- Langflow 1.10 introduces Extension bundles: component providers that are packaged and versioned as standalone pip packages, independent of the core Langflow server.
- All bundles are still included in `uv pip install langflow`, so nothing changes for existing users.
-
- The following Extension bundles ship with Langflow 1.10:
-
- | Package | Bundle | Components |
- |---------|--------|------------|
- | `lfx-arxiv` | [arXiv](/bundles-arxiv) | arXiv search |
- | `lfx-docling` | [Docling](/bundles-docling) | Document parsing and chunking |
- | `lfx-duckduckgo` | [DuckDuckGo](/bundles-duckduckgo) | Web search |
- | `lfx-ibm` | [IBM](/bundles-ibm) | IBM watsonx.ai LLM and embeddings, IBM Db2 Vector Store |
-
- To view all Extension bundles currently loaded in your environment:
-
- ```bash
- lfx extension list
- ```
-
- If a bundle is not installed, its components are absent from the canvas.
- To install a missing bundle, install its standalone package and restart Langflow:
-
- ```bash
- uv pip install lfx-duckduckgo
- uv run langflow run
- ```
-
- Components in Extension bundles use a namespaced identifier instead of a bare class name.
- For example, `DuckDuckGoSearchComponent` is now identified as:
-
- ```text
- ext:duckduckgo:DuckDuckGoSearchComponent@official
- ```
-
- You can see this identifier in the raw JSON of any saved flow that uses the component.
- Langflow migrates saved flow references to the new format automatically when you open a flow.
-
- For more information, see [Langflow Extensions overview](../Develop/extensions-overview.mdx).
-
-- File System component
-
- The **File System** component gives agents sandboxed read/write access to files on disk.
- An optional **Read Only** mode restricts the agent to read and search operations only.
-
- For more information, see [File System](../Components/file-system.mdx).
-
-- Unified **Knowledge Base** component
-
- The **Knowledge Ingestion** and **Knowledge Base** (legacy) components are merged into a single **Knowledge Base** component with a **Mode** selector.
-
- Existing flows that use the legacy components continue to work.
-
- For more information, see the [**Knowledge Base** component](/knowledge-base) and [Manage vector data](/knowledge).
-
-- Database connectors for knowledge bases
-
- Knowledge bases now support configurable vector database backends through **DB Providers** configured in **Settings → DB Providers**.
- Available providers include [Chroma (default)](https://docs.trychroma.com/), [Chroma Cloud](https://docs.trychroma.com/cloud/getting-started), and [OpenSearch](https://docs.opensearch.org/latest/about/).
-
- For more information, see [Manage vector data](/knowledge).
-
-- MCP server management lock
-
- Set `LANGFLOW_MCP_SERVERS_LOCKED=true` to prevent non-superusers from adding, editing, or removing MCP server connections.
-
- For more information, see [Restrict MCP server management to superusers](/mcp-server#restrict-mcp-server-management).
-
-- Embedded mode UI flags
-
- Set `LANGFLOW_EMBEDDED_MODE=true` to hide standalone UI elements when embedding the Langflow visual editor in another application.
-
- These flags control UI visibility only. They do not block the underlying API endpoints.
-
- For more information, see [Embedded mode](/environment-variables#embedded-mode).
-
-- Custom component admin-only restriction
-
- Set `LANGFLOW_CUSTOM_COMPONENT_ADMIN_ONLY=true` to restrict custom component creation and code editing to superusers.
-
- For more information, see [Restrict custom component creation to superusers](/deployment-block-custom-components#restrict-custom-components-to-superusers).
-
-- macOS support matrix
-
- A new [macOS support](./macos-support-matrix.mdx) page documents feature availability across Apple Silicon and Intel Macs.
-
-- IBM Db2 Vector Store component
-
- The **IBM Db2 Vector Store** component is now available in the [IBM bundle](/bundles-ibm).
-
- For more information, see [IBM Db2 Vector Store](/bundles-ibm#ibm-db2-vector-store).
-
-- Internationalization
-
- The Langflow interface is now available in multiple languages.
- To change the display language, click your **Profile Picture**, select **Settings**, and then select a language from the **Language** dropdown.
-
- Available languages include English, French (Français), Spanish (Español), German (Deutsch), Portuguese (Português), Japanese (日本語), and Chinese (中文).
-
-- Login endpoint rate limiting
-
- Langflow now applies IP-based rate limiting to the `/login` endpoint to protect against brute-force attacks.
-
- For more information, see [Login rate limiting](/api-keys-and-authentication#login-rate-limiting).
-
-- Code Agents bundle (beta)
-
- The **Code Agents** bundle adds new beta agent components for code generation with the `smolagents` and `OpenDsStar` libraries.
-
- For more information, see [Code Agents bundle](./bundles-codeagents).
-
-- File Processing bundle (beta)
-
- The **File Processing** bundle adds components for ingesting and retrieving file content in agent workflows.
-
- For more information, see [File Processing bundle](./bundles-files-ingestion).
-
-### Deprecations
-
-- **Text Input** and **Text Output** components are legacy
-
- The **Text Input** and **Text Output** components are now legacy and may be removed in a future release.
- Replace them with the [**Chat Input** and **Chat Output** components](/chat-input-and-output).
-
-- Voice mode is removed
-
- The **Microphone** button in the **Playground** now only enables speech-to-text, with no additional voice mode functionality.
- The bi-directional `/v1/voice/ws/flow_as_tool/{flow_id}` voice-to-voice endpoints are deprecated.
- For more information, see [Use voice mode](/concepts-voice-mode).
+For 1.10.x release notes, see the [1.10.x documentation](https://docs.langflow.org/1.10.0/release-notes).
## 1.9.x
diff --git a/docs/sidebars.js b/docs/sidebars.js
index 468a622443..f0cb2b5785 100644
--- a/docs/sidebars.js
+++ b/docs/sidebars.js
@@ -470,6 +470,7 @@ module.exports = {
"Components/bundles-milvus",
"Components/bundles-mistralai",
"Components/bundles-mongodb",
+ "Components/bundles-nextplaid",
"Components/bundles-notion",
"Components/bundles-novita",
"Components/bundles-nvidia",