txtai
All-in-one AI framework: embeddings DB + LLM pipelines + agents.
About
txtai is an all-in-one AI framework built around an embeddings database — a union of dense/sparse vector indexes, graph networks, and relational storage — that serves as the knowledge backbone for LLM apps. On top: pipelines for LLM prompts, QA, labeling, transcription, translation, and summarization; workflows that join pipelines into business logic; and agents that connect everything to solve tasks autonomously. Ships a web + MCP API with JavaScript, Java, Rust, and Go bindings. Apache-2.0, 13k stars. Choose it for a batteries-included stack; choose best-of-breed components when you need the cutting edge of any single layer.
What it can do
- **Embeddings DB**: dense + sparse vectors, graphs, relational — with SQL
- **LLM pipelines**: prompts, QA, labeling, transcription, translation, summarization
- **Workflows**: join pipelines into microservices or multi-model flows
- **Agents**: autonomous agents over embeddings, pipelines, and workflows
- **MCP + polyglot**: Model Context Protocol API; JS/Java/Rust/Go bindings
- **Multimodal**: text, documents, audio, images, video
Setup
```bash
pip install txtai
```
```python
from txtai import Embeddings
emb = Embeddings({"path": "sentence-transformers/all-MiniLM-L6-v2"})
emb.index(["The cat sat on the mat.", "Dogs are great."])
print(emb.search("feline", 1))
```
Docs: neuml.github.io/txtai. Extras install per modality (audio/image/video).
Prerequisites
Python 3.9+. Optional: model-specific extras; API keys for hosted LLM pipelines.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: 13,011 GitHub stars (log-scaled vs catalog peers); blended with ~13,889 monthly package downloads; last push 1d ago; 15 releases/12mo; 25 open issues. Manual review: genuinely all-in-one (embeddings DB + pipelines + workflows + agents + MCP); 13k stars; breadth over depth vs best-of-breed; modest PyPI traction. License: Apache-2.0.
★ 13,011 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.3/10
- Maintenance · recency, releases, issue load
- 9.9/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · 13,011 GitHub stars (log-scaled vs catalog peers)
- · blended with ~13,889 monthly package downloads
- · last push 1d ago; 15 releases/12mo; 25 open issues
- · manual review: genuinely all-in-one (embeddings DB + pipelines + workflows + agents + MCP); 13k stars; breadth over depth vs best-of-breed; modest PyPI traction
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.