feat(assistants): use community tools in assistants (#1421)

fix(output): load audio metadata on chrome

fix(stream): add keep alive stream event

fix(logs): better error logging

Community tools (#1250)

Major bump to sveltekit & node-adapter versions (#1395)

hotfix(android): disable tools on android typo

feat(android): disable tools on android app (#1390)

Add CSP protection on uploaded files (#1304)

Add "disable streaming tokens" settings & remove media query (#1299)

Add createdAt filter on messageEvents (#1256)

Fix max upload size in huggingchat (#1251)

Feat: Delete branch #1210 (#1240)

Reduce message update size returned by websearch (#1179)

Bypass spaces tools rate limit (#1167)

Add metrics for models, tools, websearch (#1186)

fix streaming not updating on front-end (#1173)

Disable text streaming if user prefers reduced animation (#1131)

Function calling (#996)

Generic Multimodal Support Fixed (#1147)

Revert "Generic Multimodal Support" (#1146)

Generic Multimodal Support (#1021)

Remove trim on user input (#1122)

Move vars to dynamic, add metrics (#1085)

Use pino for logs (#1086)

add createdAt and udpatedAt to new messages (#1062)

Add support for Meta Llama 3 70B in HuggingChat (#1029)

Add assistant instructions to first system message (#1023)

Allow websearch if assistant has dynamic prompts (#1017)

Send characters by groups of 5 (#1007)

Group tokens randomly (#964)

[Assistants] trending feature (#938)

Expose sampling controls in assistants (#955) (#959)

Dynamic system instructions (#949)

feat: smooth and combine token output (#936)

Store user agent in conversation (#940)

Fix model identification when creating conversations (#820)

Add websearch controls for assistants (#812)

Return an error on unknown model in new conversation (#916)

Make sure preprompt is set on open ai endpoint type (#913)

Add limits on API endpoints (#886)

Improve error handling when posting new messages (#892)

Make sure we properly close the stream when done (#891)

Fix Error: Ancestor Not Found issue (#884)

Use `uuid` for message ID generation instead of crypto method (#882)

Pass websearch to `preprocessMessages` (#876)

[Mongo] count messages (aggregate) only when needed V2 (#866)

Abort generation if an error is displayed on the front-end. (#855)

Conversation trees (#223) (#807)

Several QoL contributions (#760)