OSS Tanbou
View on GitHubProject website

compose retrieval, routing, memory, and generation as component graphs for RAG and agent workflows

OSS scale score 303.1OSS health 100Update pace 255 commits/30d
About these scores

OSS scale score is an unbounded metric that log-compresses and weights Stars, Watchers, Forks, and Contributors. Discovery score is the current OSS scale score minus the score at discovery. Update pace is commits in the last 30 days, growth momentum is the OSS scale score difference within the recent observation window, and OSS health is a 0–100 rating based on available recency, Community Health, and release data.

Stars
26,647
Primary language
Python
License
Apache-2.0
Repository last updated
Oct 2, 2026
On this page

Overview

Haystack is a Python orchestration framework for production-ready LLM applications. Retrieval, indexing, prompting, generation, memory, and tools are explicit components that can be connected into Pipelines and Agent workflows with branches, loops, synchronous or asynchronous execution.

Features and best fit

Based on official documentation; not hands-on tested · Content checked:

Key features

Connect modular components into explicit RAG pipelines

Haystack connects retrievers, document stores, prompt builders, generators, and other components into Pipelines with explicit data flow and synchronous or asynchronous execution.

Sources: [1][2]

Combine Agents, tools, memory, and lifecycle hooks

Agents support tool calling, memory, lifecycle hooks, and tracking for steps, token usage, and tool calls, allowing controlled agentic workflows with branching and loops.

Sources: [1]

Best fit

Fits RAG and agent systems that need transparent control over retrieval and generation

It is useful when model vendors may change but document retrieval, context construction, routing, and generation should remain separately testable and replaceable components.

Sources: [1][2]

Before adoption

Manage provider credentials, costs, data boundaries, and telemetry

The Get Started examples pass model-provider API keys through Haystack Secrets, while the README states that anonymous component-usage telemetry is collected. Production deployments should review credentials, external data flows, cost controls, and telemetry settings.

Sources: [1][2]

Haystack 3.3.0 requires Python 3.10+ and raises the anyio security floor

The package metadata requires Python >=3.10, and the 3.3.0 release notes require anyio>=4.14.2 to address CVE-2026-63374 in the transitive dependency chain.

Sources: [3][4]

Official sources

  1. [1]Haystack v3.3.0 README(2026-10-04)
  2. [2]Haystack v3.3.0 Get Started(2026-10-04)
  3. [3]Haystack v3.3.0 package metadata(2026-10-04)
  4. [4]Haystack v3.3.0 release(2026-10-04)
  5. [5]Haystack Apache-2.0 license(2026-10-04)
Supplemental curator note

Haystack fits RAG and agent systems where retrieval and generation should remain explicit component graphs rather than a black box. When integrating external models or vector databases, define credentials, costs, and data boundaries separately, and opt out of anonymous telemetry if it is not appropriate for the environment.

Try it in 3 steps

  1. 1

    Create an isolated Python virtual environment

    Use Python 3.10 or newer as required by Haystack 3.3.0.

    python3 -m venv .venv && . .venv/bin/activate
  2. 2

    Install Haystack 3.3.0

    Pin the core framework without configuring any external LLM-provider credentials yet.

    python -m pip install haystack-ai==3.3.0
  3. 3

    Run in-memory BM25 retrieval without an API key

    Validate the DocumentStore and Retriever components without calling an external model API. Review provider credentials, cost, and data destinations before adding an LLM.

    python -c 'from haystack import Document; from haystack.document_stores.in_memory import InMemoryDocumentStore; from haystack.components.retrievers import InMemoryBM25Retriever; s=InMemoryDocumentStore(); s.write_documents([Document(content="Haystack builds RAG pipelines"),Document(content="Dask runs parallel computing")]); print(InMemoryBM25Retriever(document_store=s).run(query="RAG")["documents"][0].content)'
Check the official README

Growth

Growth trends · Last 30 days

26,647 Stars

Trend data is still being collected.

Development activity

Last 90 days · weekly

Commits (last 30 days)
255
Open PRs
58

Development activity is still being collected.

Built with

Categories and tags

Categories

Tags

GitHub data

GitHub dataView detailed GitHub data
View on GitHub

GitHub Topics

  • semantic-search
  • information-retrieval
  • ai
  • python
  • large-language-models
  • generative-ai
  • llm
  • rag
  • retrieval-augmented-generation
  • agents
  • orchestration
  • agent-framework
Stars
26,647
Forks
3,228
Watchers
166
Open issues
96
Contributors
446
Owner type
Organization
Primary language
Python
License
Apache-2.0
Repository last updated
Oct 2, 2026
Write a related article

Share a guide or use case for this OSS in Markdown. Articles are published after administrator approval.

Report incorrect information

Tell us if any listing information is incorrect or outdated.

After reading this page, do you know what to do next?