Defer thinking indicators until backend reports selected advisors

Stream selected advisors so thinking indicators only show the chosen 3

Add [DEBUG-1d5c0f] trace on /auth 401 branches

Cap chat-stream to top 3 advisors via relevance ranking; rank within active-advisor selection

Add $push and $setOnInsert to SQLite shim (with [DEBUG-1d5c0f] trace)

Restore header advisor dropdown and honor active_advisor selection in chat-stream.

Include full chat history under 4k tokens, else LLM summary

Refactor for HF Spaces: single-container, SQLite shim, ChromaDB on /data

Remove debug instrumentation after end-to-end verification

Wire LLM failover for the BrainForge Cybersecurity endpoint

Fix LLM chat startup and polish cybersecurity advisor UI.

Restore cybersecurity user profile UX and personalize advisor responses

Add Cybersecurity Advisor Canvas on FEAT_CybersecurityCanvas

Feat/improve clarification check (#57)

Support Voice Interaction (#32)

Feat/persona avatars (#43)

Fix/dup first message (#44)

Feat/advisor tools (#38)

Feat/account management (#41)

Fix/json output (#40)

Feat/vllm endpoints (#36)

Stream LLM Responses (#18)

Refactor/unittests (#33)

Feat/auto persona colors (#23)

Fix/structured json output (#30)

Refactor chat UI layout (#19)

Feat/validate lucide icons (#21)

Feat/improve persona config (#16)

Improve backend logic around user input parsing (#12)

Add configuration support (#3)

Implement Docker deployment (#1)

Merge pull request #78 from sohank-17/sohank-fix_responses

update model and chat fetch

update session management

Response formatting fix

fix json format bug

fix serialization bug

bug fix

update canvas

add canvas router

fix boolean error

add canvas backend

add new imports

update CORS policy

fix document upload to new chat

bug fixes

add document auth

Backend Documentation Addition

update requirements

update formatting instructions