cortex
cortex copied to clipboard
janhq
Drop-in, local AI alternative to the OpenAI stack. Multi-engine (llama.cpp, TensorRT-LLM). Powers 👋 Jan
**Describe the bug** When attempting to import a YAML-formatted Nitro / Cortex OpenAPI specification (OAS) file downloaded from a website into Postman, the import fails with an error message. However,...
feat: Support for multilingual embedding models and simultaneous hosting of LLM & embedding models
**Problem** Jan lacks support for multilingual embedding models and cannot host both LLM and embedding models simultaneously. **Success Criteria** Jan should be enhanced to support a broader range of multilingual...
**Problem** A place holder to support python-runtime as an engine. Will fill in detail later.
## Describe Your Changes - allow user to pull model from janhq from HuggingFace ## Videos https://github.com/janhq/cortex/assets/10397206/70f7af9e-3c86-4dbd-8322-91cf43a44987 ## Self Checklist - [ ] Added relevant comments, esp in complex areas...
issue: https://github.com/janhq/cortex/issues/589 endpoint: v1/models response example: ``` { "data": [ { "id": "nomic-embed-text-v1.5.f16", "object": "model" }, { "id": "mistral-7b-instruct-v0.1.Q4_0", "object": "model" } ], "object": "list" } ``` bump cortex.llamacpp version...
## Describe Your Changes - ## Fixes Issues - Closes # - Closes # ## Self Checklist - [ ] Added relevant comments, esp in complex areas - [ ]...
Added the API Reference decorator in the DTOs and Controller files. - Assistant - Chat - Cortex - Model - Messages - Threads
### Discussed in https://github.com/janhq/cortex/discussions/235 Originally posted by **zaynpatel** December 4, 2023 I'm currently running the following command, copied from the documentation with the exception of a new localhost address: ```...
## Description This PR adds a transform interceptor to controllers that return list results compatible with the OpenAI SDK format.
cortex-cpp returns a list of running models, so we don't have to send every single request over the full models list (let's say it's 1000 models)