1.38.0

🎁 New

• (In Beta) Simplify self-hosting setup by using embedded db by @debanjum in #1141
  Try it out and share feedback if it is working or not for you (on Github or Discord).
  You can get started by just running the following commands. (No separate Postgres server required)

   pip install 'khoj[local]'
   USE_EMBEDDED_DB="true" khoj -vv --anonymous-mode

• Enable unsharing your public conversation by @debanjum in #1135
  Hit the delete button on the title pane of the shared conversation's public URL to unshare it

🧪 Improve

• Setup Google imagen text to image model if Google API setup during initialization in 56b63f9
• Use new Anthropic chat models if Anthropic API setup during initialization in 56b63f9
• Improve schema compliance for more chat actors by enforcing json schema (most useful for gemini 2.0 flash model) in a387f63
• Chat with your code by attaching programming language file types directly on the web app in e9132d4

🛠️ Fix

• Allow chat on first run by fixing default agent initialization
• Fix infer webpage url chat actor to only request up to specified max URLs
• Use agent chat model for research model planning
• Handle new safety triggers for Gemini models
• Fix automations docs by @eltociear in #1140

Full Changelog: 1.37.2...1.38.0

1.37.2

🛠️ Fix

• Minor updates to documentation by @sabaimran in b7ac877

Full Changelog: 1.37.1...1.37.2

1.37.1

🧪 Improve

• Access Claude and Gemini via GCP Vertex AI by @debanjum in #1134
• Standardize AI model temperatures across application by @debanjum in d1df958
• Upgrade package dependencies - next, torch by @debanjum in 55ae0ed

Full Changelog: 1.37.0...1.37.1

1.37.0

🎁 New

• Support E2B as code sandbox provider by @debanjum in #1120
• Support image generation by Google's Imagen models by @debanjum in 7bb6fac

🧪 Improve

• Support chat export and account deletion via web app by @sabaimran in a3c4347
• Support Claude 3.7 extended thinking in research mode by @debanjum in 50f71be
• Improve Obsidian sync for large vaults by @debanjum in #1078
• Implement better bug issue template by @ArtemkaKun in #1129
• Support enforcing json schema in supported AI model APIs by @debanjum in #1133
• Allow chat input on web app while Khoj responds to speed up interactions by @debanjum in 9a0b126

🛠️ Fix

• Fix support for Deepseek R1 via official API in 639b5a4 to resolve #1126
• Fix passing inline images to vision models in 0c53106 to resolve #1112
• Fix clearing knowledge base on delete in b692e69 to resolve #1122

New Contributors

• @ArtemkaKun made their first contribution in #1129

Full Changelog: 1.36.6...1.37.0

1.36.6

🧪 Improve

• Add a button to convert side panel settings into an agent by @sabaimran in 848a913
• Only show notes option to LLM if some data is indexed by @debanjum in bb0828b
• Allow separation of ALLOWED_HOSTS and ALLOWED_DOMAIN by @debanjum in 5a3c7b1

🛠️ Fix

• Fix automation validity check by @sabaimran in 5d6eca4

Full Changelog: 1.36.5...1.36.6

1.36.5

🛠️ Fix

• Simplify agent selector by @sabaimran in 589b047
• Fix deletion -> invalid chat -> client-side error by @sabaimran in 427ec06
• Only default setup offline models if in interactive mode by @sabaimran in 0211151

Full Changelog: 1.36.4...1.36.5

1.36.4

🧪 Improve

• Fix system prompt for generating diagrams better with GPT by @sabaimran in 2e01a95
• Fix dark mode page change jitter by @sabaimran in 43e032e

Full Changelog: 1.36.3...1.36.4

1.36.3

🛠️ Fix

• Fix another infinite loop condition in side bar chat model selection by @sabaimran in 0d10c5f

Full Changelog: 1.36.2...1.36.3

1.36.2

🛠️ Fix

• Fix infinite loop when detecting current chat model in the conversation by @sabaimran in 51874c2

Full Changelog: 1.36.1...1.36.2

1.36.1

🛠️ Fix

• Fix rendering of model strengths, and misc. by @sabaimran in a3d75e5

Full Changelog: 1.36.0...1.36.1