v1.1.1

1. Supports dropdown chatmodel selection
2. Added admin console endpoint /admin.
3. Now vectorstore is not shared for all acoounts. Every account has own vectorstore, but will share public database, which can be embedded by /share command.
4. Added token status box in frontend
5. LLAMA supports GPU offloading when using cuBLAS.
6. Now /query command doesn't put queried texts into chat context. The queries data will only be used for generating current response.