#7070th of 153 in Agent frameworks and SDKs

semantic-router

Python library that routes queries to predefined routes by embedding similarity

Stars
4k
License
MIT
Last commit
Nov 2026
Last release
Nov 2026
Language
Python
Stars, last 29 weeks+596
25 Apr 2026: 3,443 stars2 May 2026: 3,456 stars9 May 2026: 3,481 stars16 May 2026: 3,496 stars23 May 2026: 3,521 stars30 May 2026: 3,545 stars6 Jun 2026: 3,570 stars13 Jun 2026: 3,597 stars20 Jun 2026: 3,613 stars27 Jun 2026: 3,633 stars4 Jul 2026: 3,655 stars11 Jul 2026: 3,683 stars18 Jul 2026: 3,713 stars25 Jul 2026: 3,731 stars1 Aug 2026: 3,756 stars8 Aug 2026: 3,776 stars15 Aug 2026: 3,802 stars22 Aug 2026: 3,820 stars29 Aug 2026: 3,845 stars5 Sept 2026: 3,863 stars12 Sept 2026: 3,890 stars19 Sept 2026: 3,909 stars26 Sept 2026: 3,927 stars3 Oct 2026: 3,939 stars10 Oct 2026: 3,946 stars17 Oct 2026: 3,959 stars24 Oct 2026: 3,979 stars31 Oct 2026: 4,005 stars7 Nov 2026: 4,032 stars14 Nov 2026: 4,039 stars

3,443 on 25 Apr 20264,039 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Semantic Router is a Python library that decides which predefined route a user message belongs to by comparing embeddings of example utterances, instead of waiting on an LLM generation. Routers can be calibrated with labelled queries so the confidence threshold acts as a probability and can return no match. The 1.x line also adds a DecisionRouter backed by small decision models, plus Chain and Merge for combining routers.

Who it is for: Python developers adding fast intent routing or guardrails to LLM apps

Strengths

  • Runs fully locally with FastEmbed, no API keys needed
  • Calibration via fit() turns scores into probabilities and allows a none-of-these result
  • Swappable encoders (OpenAI, Cohere, Hugging Face) and indexes (Pinecone, Qdrant, Postgres)
  • Supports hybrid sparse search, image routing, and saving routers to YAML or JSON

Weaknesses

  • 1.x changes the API from 0.x; 0.x gets bug fixes only and needs migration
  • Routing quality depends on hand-written example utterances and labelled calibration data
  • DecisionRouter relies on TypeSafe's Jev, Laya or Von models, not general LLMs
  • Only decides which route applies; tool execution and agent logic are left to you

What it needs

  • Compose
  • Needs Pinecone (optional), Qdrant (optional), Postgres (optional)
  • Models: FastEmbed (BAAI/bge-small-en-v1.5), OpenAI embeddings, Cohere embeddings, Hugging Face encoders, TypeSafe Jev, Laya, Von decision models

What changed

  1. v1.1.1 ↗ (opens in a new tab)Security fix

    GHSA-6Q48-6C7W-9GW2. “LiteLLM 1.107.0 is excluded. That PyPI release contained malware (a supply-chain attack, live on PyPI on 21 October 2026 from 11:39 to 13:23 UTC), and the litellm extra allowed it. The extra is now l…”

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Agent frameworks and SDKs

See all 153
Also in Agent frameworks and SDKs
RankProjectScore
1ppt-masterAgent skill that turns documents into natively editable PowerPoint decks Live demo ↗ (opens in a new tab)78k starsMIT96 out of 100
2copilotkitReact and Angular SDK for adding agent chat and generative UI to apps38.8k starsMIT92 out of 100
3openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing30.7k starsMIT92 out of 100
#4mastraTypeScript framework for building AI agents, workflows and MCP servers29.6k starscustom license92 out of 100
#5browser-usePython library that lets an LLM agent control a web browser Live demo ↗ (opens in a new tab)121.4k starsMIT90 out of 100