Spaces:

RecapAnon
/

AskMiku

Running

App Files Files Community

AskMiku / index.js

Commit History

Apr 25

20:03
Fix var name.

RecapAnon

05d267c
19:56
Fix SyntaxError.

RecapAnon

8cc82ba
18:54
Delete unused.

RecapAnon

3d04257
18:53
Update chunk list.

RecapAnon

056abac
18:53
Update prompt.

RecapAnon

dc681b3
18:37
Fix user and system message order.

RecapAnon

1f91e7e
18:36
Remove createStreamingMessageContainer and use createMessageBubble directly in generateResponse.

RecapAnon

604bc93
18:27
Replace embeddings model with "onnx-community/harrier-oss-v1-0.6b-ONNX".

RecapAnon

392abba

Apr 14

23:15
Render model output using Showdown.js v2.1.0 instead of <pre> tags.

RecapAnon

3884647
23:05
Attempt to curb the emoji.

RecapAnon

ca4ecd4
22:57
Stop setting past_key_values to prevent crashes.

RecapAnon

248db27

Apr 13

14:38
Load embeddings from HuggingFace dataset.

RecapAnon

10124e4
10:46
Increase max_new_tokens to 2048.

RecapAnon

c825b13
10:29
Add deferred model loading with 'Load model' button.

RecapAnon

f210955
10:24
Update model to E4B.

RecapAnon

bdb7b0c
10:24
Update prompts.

RecapAnon

0b640b7
10:18
Add automatic first assistant message after initialization.

RecapAnon

c376fe2

Apr 12

05:17
Move vector DB context to temporary system message in generateResponse.

RecapAnon

e0bc921
05:00
Load model and processor in parallel and fix progress_callback glitching.

RecapAnon

e4af149
04:51
Bring back past_key_values_cache.

RecapAnon

debe7aa
04:49
Fix rgb error.

RecapAnon

b76d0e7
04:34
Fix import url.

RecapAnon

b5186e7
02:18
Use AutoProcessor and Gemma4ForConditionalGeneration to fix chat_template not set.

RecapAnon

00e9a7a
02:12
Replace Qwen 3 with Gemma-4-E2B-it.

RecapAnon

21087b0
02:11
Update transformers.js version to 4.0.1.

RecapAnon

098ac12

Apr 11

17:41
Remove redundant model ID from TextGenerationPipeline.getInstance call.

RecapAnon

3dd3b56
17:02
Replace enhanced user message with original in conversation history after generation.

RecapAnon

aa98e8e
16:52
Revert back to plain Miku.sh prompt.

RecapAnon

89480b6
16:50
Update tokenizer.batch_decode call to slice the sequences so only generated tokens are decoded.

RecapAnon

ad43a26
16:29
Update model loading to use AutoModelForCausalLM and apply_chat_template to disable thinking.

RecapAnon

37e8504

Dec 8, 2025

00:50
Remove useless importDatabaseFromString.

RecapAnon

b8ee0ee
00:46
Update model to Qwen3-4B.

RecapAnon

e1426f8
00:41
Initialize db from chunked files to fix file limit on Chrome.

RecapAnon

18960c7
00:08
Fix database import on Firefox by loading the json file in chunks.

RecapAnon

f0c3543

Nov 1, 2025

21:31
Switch to webgpu and add device checks.

RecapAnon

de2014a
21:02
Bring back getNativeDB();

RecapAnon

cd3f3fd
20:45
Fix missing comma.

RecapAnon

2405497
20:02
Update system prompt from character card and remove soft CoT.

RecapAnon

64582fd
19:55
Fix indexeddb-export-import import again.

RecapAnon

45ec7ad

Oct 29, 2025

23:42
Replace system prompt with Miku.sh prompt

RecapAnon

d5fa973
23:19
Wrap message text in pre tags for proper markdown line break display⁠

RecapAnon

e7d7941
22:14
Replace emoji avatars with image avatars for chat participants⁠

RecapAnon

589d9e7

Oct 25, 2025

20:43
Delete getNativeDB().

RecapAnon

d256c08
20:19
Strip out thinking.

RecapAnon

b6b3e59
20:14
Bring back importDatabaseFromString.

RecapAnon

09768ab
19:57
Delete unused methods.

RecapAnon

9616ab4
19:54
More cleanup.

RecapAnon

1eb53a0
19:43
Only load from database.json if not already loaded (to prevent duplication)

RecapAnon

b60b326
19:37
Remove more useless comments.

RecapAnon

d74eb19
19:35
Clean up.

RecapAnon

35982a7