Commit Graph

867 Commits

Author SHA1 Message Date
60e099e852 Clean-up diff pt.2 2024-04-27 21:15:44 +03:00
c49b37f968 Clean-up diff 2024-04-27 21:11:41 +03:00
4521dde455 add a button to translate input message 2024-04-27 13:46:13 +08:00
d5c0aed973 Replace Comfy avatar placeholders with empty pixel if message fails to load 2024-04-21 21:13:50 +03:00
dcb98b2969 Substitute macros in TTS text before narrating 2024-04-21 01:02:49 +03:00
366e2abb8b Move expressions LLM request setting mode listener 2024-04-19 23:01:31 +03:00
eab545cafc Error handling for Comfy request 2024-04-19 00:39:04 +03:00
901ffa3cdc Add char and user avatar placeholders to ComfyUI workflow editor
They resolve to base64 encoded data URIs of respective avatars.
2024-04-19 00:32:38 +03:00
16a2ac0bbe Add missing awaiters to getQueryText 2024-04-18 22:58:14 +03:00
cc1c3f173c Merge pull request #2097 from Dakraid/feature/summarize-before-embedding
Summarize the chat messages handed to the embedding beforehand
2024-04-18 22:49:34 +03:00
5e18b1872b Hide vector summarization unless chat vectors is enabled 2024-04-18 22:32:01 +03:00
d281767867 Add /caption command 2024-04-18 16:22:33 +03:00
abd8acd9fc Add experimental flask icon with title 2024-04-17 20:35:51 +02:00
977c1d4ac4 Add missing hr to the settings html 2024-04-17 20:35:51 +02:00
21b5bb5a29 Rearrange settings 2024-04-17 20:35:51 +02:00
10692523f0 Improve settings, add settings for prompt, sent message summarize toggle 2024-04-17 20:35:50 +02:00
a022c9eccb Set css class for summary dropdown in vectors 2024-04-17 20:35:50 +02:00
4cfd0c71cf Add extras summarization API, fix missing substitution of placeholders 2024-04-17 20:35:49 +02:00
a1473dedd0 Make summarization toggleable 2024-04-17 20:35:49 +02:00
59abee3043 Summarize the chat messages handed to the embedding beforehand 2024-04-17 20:35:48 +02:00
341f02e425 Fix default classify API migration 2024-04-15 04:04:30 +03:00
f5955bdd37 Throw if couldn't classify 2024-04-14 20:02:34 +03:00
0ff5d0b5f1 Substitute macro is LLM prompt 2024-04-14 19:53:40 +03:00
6dca8b28fe Fix talkinghead check 2024-04-14 19:46:33 +03:00
bd6fe19bf1 Merge branch 'staging' into llm-expressions 2024-04-14 19:43:34 +03:00
3e60919289 Specify LLM prompt in case JSON schema is not supported 2024-04-14 17:13:54 +03:00
b02394008c Fix settings migration. Add lint rule. 2024-04-14 15:40:10 +03:00
306bff0a92 #2085 Switch to async token counters 2024-04-13 21:33:19 +03:00
913085ba74 Expressions: Let the user know if classification failed
Send a message if the classifer fails and fallback to the default
expression.

Signed-off-by: kingbri <bdashore3@proton.me>
2024-04-13 01:29:54 -04:00
8c4dd6ce1e Expressions: Ignore stopping strings
Don't use stopping strings when triggering a classify generation.

Signed-off-by: kingbri <bdashore3@proton.me>
2024-04-12 09:57:09 -04:00
bea63a2efe Expressions: Remove TalkingHead from choices
This is a sub-choice under extras.

Signed-off-by: kingbri <bdashore3@proton.me>
2024-04-12 01:59:44 -04:00
6b656bf380 Expressions: Classify using LLM
Rather than using a separate BERT model to classify the last message,
use the LLM itself to get the classified expression label as a JSON
and set that as the current sprite. Doing this should take more information
into consideration and cut down on extra processing.

This is made possible by the use of constrained generation with JSON
schemas. Only available to TabbyAPI since it's the only backend that
supports the use of JSON schemas, but there can hopefully be a way
to use this with other backends as well.

Intercepts the generation and sets top_k = 1 (for greedy sampling)
and the json_schema to an emotion enum. Doing this also prevents
reingestion of the entire context every time a message is sent and
then asked to be classified, which doesn't compromise the chat
experience.

Signed-off-by: kingbri <bdashore3@proton.me>
2024-04-12 01:55:16 -04:00
356f1f9c39 Clean-up non-character state checks 2024-04-11 23:50:01 +03:00
123131074c Change built-in extensions template render to async 2024-04-11 23:38:44 +03:00
3da9fb4efe Switch to module exports 2024-04-11 22:39:42 +03:00
6290dff3d9 #2065 Add option to translate text before classification 2024-04-11 22:09:05 +03:00
210f388d7a Switch button style to use class 2024-04-11 21:33:15 +03:00
063c81540d IVC + support new settings and turbo model (#2067) 2024-04-11 21:29:19 +03:00
5f6d895acf Add attribution links 2024-04-10 21:21:31 +03:00
c90105e9f6 Merge branch 'staging' into pollinations 2024-04-10 21:14:36 +03:00
540cddf300 new turbo model 2024-04-10 14:24:43 +10:00
fc1896dcff #2047 (WIP) Refactor TTS worker to use event source 2024-04-09 17:50:27 +03:00
235afb5d15 Fix asset title for non-extension types 2024-04-09 11:30:24 +03:00
b461c6f0bb use generic popups for QR editor 2024-04-08 18:43:21 -04:00
d56722a4b6 helper text and repo link tooltip for asset list items 2024-04-09 02:48:39 +09:00
c55181f955 Fix control alignment funkiness 2024-04-08 14:55:31 +03:00
63cbfda9b1 add ctrl-enter to execute from editor 2024-04-07 17:01:26 -04:00
21d3a7dc3e add tab size setting to QR editor 2024-04-07 17:01:02 -04:00
3ed0564d73 add default value to qr editor wrap setting 2024-04-07 16:59:50 -04:00
e26c7827bc Fallback expression reset if custom expression deleted 2024-04-07 06:26:48 +02:00