Releases: aurelio-labs/semantic-router
v0.0.28
What's New
Addition of different aggregation methods (sum
, mean
, max
) and improvements to HybridRouteLayer
thanks to @andreped! Much maintenance, fixes, and optimization.
What's Changed
- chore: splitters code clean by @jamescalam in #182
- fix: clip notebook fix by @jamescalam in #183
- fix: fixes #188
AzureOpenAIEncoder
issue by @szelesaron in #189 - fix: Fit and vec_evaluate can now handle dynamic routes by @Siraj-Aizlewood in #192
- refactor: Removed trigger from RouteChoice by @Siraj-Aizlewood in #195
- feat: Support for setting top_k for HybridRouteLayer by @andreped in #197
- fix: optional dependency issue by @zahid-syed in #190
- feat: Support openai base_url by @bdqfork in #180
- fix: README route example simplification by @andreped in #196
- chore: Update FastEmbed to latest by @Anush008 in #200
- feat: Support for setting aggregation for (Hybrid)RouteLayer by @andreped in #202
- chore: prepare v0.0.28 release by @jamescalam in #205
New Contributors
- @szelesaron made their first contribution in #189
- @andreped made their first contribution in #197
- @bdqfork made their first contribution in #180
- @Anush008 made their first contribution in #200
Full Changelog: v0.0.27...v0.0.28
Release v0.0.27
We now have multi-modal support! See here for a Shrek vs. not-Shrek route layer.
What's Changed
- feat: add CLIP for multi modal by @jamescalam in #155
- chore: v0.0.27 release prep by @jamescalam in #181
Full Changelog: v0.0.26...v0.0.27
Release v0.0.26
Release v0.0.25
What's Changed
- feat: Unstructured elements splitting example by @simjak in #169
- fix: PineconeIndex upsert request size limit by @siddicky in #165
- fix: RollingWindowSplitter fixes by @simjak in #175
- fix: PIL Image type by @simjak in #176
- chore: Release 0.0.25 by @simjak in #177
New Contributors
Full Changelog: v0.0.24...v0.0.25
Release v0.0.24
Release v0.0.23
What's New
Added image embeddings with VitEncoder
by @bruvduroiu , new splitting methods from @simjak
What's Changed
- chore: move pinecone demo by @jamescalam in #154
- feat: Adds VitEncoder for visual transformers by @bruvduroiu in #129
- fix: llm verbose bug by @Siraj-Aizlewood in #157
- docs: WIP Create CONTRIBUTING.md by @Siraj-Aizlewood in #158
- fix: Corrected Linting Procedure in CONTRIBUTING.md by @Siraj-Aizlewood in #162
- fix: Added batches for large dataset training error by @zahid-syed in #161
- feat: openai encoder org id support by @maxyousif15 in #159
- feat: Added ollama.py by @CP500 in #156
- fix: added defaults. mistral added to layer.py by @dwmorris11 in #167
- feat: Rolling window splitter by @simjak in #150
New Contributors
- @zahid-syed made their first contribution in #161
- @maxyousif15 made their first contribution in #159
- @CP500 made their first contribution in #156
Full Changelog: v0.0.22...v0.0.23
v0.0.22
Small hotfix for PineconeIndex
and new notebook in the /docs
demonstrating how to use it.
What's Changed
- feat: Pinecone demo and hotfix by @jamescalam in #153
Full Changelog: v0.0.21...v0.0.22
Release v0.0.21
Release v0.0.21
Key new features include MistralAI API for Mistral LLMs and encoders by @dwmorris11, support for more scalable and persistent route layers using Pinecone by @Siraj-Aizlewood and @jamescalam, and new splitting mechanisms for Conversation
objects to improve context support for routing decisions by @Siraj-Aizlewood.
What's Changed
- feat: Added Method for Deletion of Routes in RouteLayer by @Siraj-Aizlewood in #128
- feat: Incorporating Semantic Topic Splitters into Conversation by @Siraj-Aizlewood in #134
- fix: pip install error in 03-basic-langchain-agent.ipynb by @jzcruiser in #131
- fix: issue #94 added from clause to the exception statement in by @dwmorris11 in #143
- feat: Added support for MistralAI API. This includes a by @dwmorris11 in #142
- feat: separate indexes and PineconeIndex by @jamescalam in #146
- fix: Saving JSON/YAML of Layer Config by @Siraj-Aizlewood in #145
New Contributors
- @jzcruiser made their first contribution in #131
- @dwmorris11 made their first contribution in #143
Full Changelog: v0.0.20...v0.0.21
Release v0.0.20
Release v0.0.19
What's New
We've added more flexible and accurate routes. By using route-specific score thresholds and the new route layer fit
and evaluate
methods, we can get optimal route layer performance in seconds! We have also added a few new example notebooks and support for OpenAI's 3rd generation embedding models and the new dimensions
parameter that they introduce.
What's Changed
- feat: Optimal Route-Level Score Thresholds by @Siraj-Aizlewood in #121
- feat: new examples and openai upgrade by @jamescalam in #126
Full Changelog: v0.0.18...v0.0.19