Spaces:

intrect
/

vela-demo

Sleeping

App Files Files

vela-demo

Commit History

Feb 18

12:21
fix(runpod): align sampling params with model card (top_k=40, top_p=0.95, rep=1.0)⁠

Heewon Oh +1: Claude Opus 4.6

90d4f5a
12:21
fix(runpod): add explicit vLLM sampling params (top_k, top_p, repetition_penalty)⁠

Heewon Oh +1: Claude Opus 4.6

1c0fe42
07:29
fix(runpod): handle tokens array and usage input/output format in response parser⁠

Heewon Oh +1: Claude Opus 4.6

a2af26c
02:59
feat(app): ZeroGPU 쿼터 소진 시 RunPod Serverless fallback 추가⁠

Heewon Oh +1: Claude Sonnet 4.5

e73a081
02:00
fix(agent): step.query 누락 시 원본 쿼리로 fallback (모델이 Query 필드 생략 시 검색 스킵 방지)

Heewon Oh

5835058
01:11
fix(reasoning): EOD 래퍼 헤더 제거 및 결론→투자 의견 정규화

Heewon Oh

fed9d2c

Feb 17

15:53
fix(reasoning): EOD 형식에서 _remove_boilerplate 제거로 섹션 손실 방지

Heewon Oh

328601d
15:35
fix(prompts): synthesis 프롬프트를 모델 학습 포맷(h3)에 맞게 단순화⁠

Heewon Oh +1: Claude Sonnet 4.5

22d81fc
15:19
fix(zerogpu): PicklingError 수정 - GPU 함수 내부에서 agent 생성⁠

Heewon Oh +1: Claude Sonnet 4.5

4da532d
15:17
fix(zerogpu): 단일 GPU 컨텍스트로 전체 research 실행⁠

Heewon Oh +1: Claude Sonnet 4.5

7c5a3f8
15:14
fix(app): ZeroGPU 호환성 - 백그라운드 스레드 제거 및 메인 스레드 동기 실행⁠

Heewon Oh +1: Claude Sonnet 4.5

fcd1ddb
15:02
fix(agent): sentence-transformers 추가 및 ActionType 직렬화 수정⁠

Heewon Oh +1: Claude Sonnet 4.5

712a7cd
14:46
fix(synthesis): remove Key Findings suffix from synthesis prompt and fix ActionType serialization⁠

Heewon Oh +1: Claude Sonnet 4.5

d4a5dee
14:18
feat(vectors): revert to KR-SBERT 768-dim with CPU lazy load⁠

Heewon Oh +1: Claude Sonnet 4.5

5f57438
14:01
feat(vectors): replace KR-SBERT with MiniLM for HF Serverless compatibility⁠

Heewon Oh +1: Claude Sonnet 4.5

e12c000
13:39
fix(vector_filter): HF_TOKEN 명시적 전달 (API 401 수정)⁠

Heewon Oh +1: Claude Sonnet 4.5

eb4ff0d
13:27
fix(vector_filter): sentence-transformers→HF InferenceClient로 교체 (ZeroGPU CUDA 충돌 수정)⁠

Heewon Oh +1: Claude Sonnet 4.5

9713ee8
13:09
fix(search): source_type.value 오류 수정 - use_enum_values=True로 문자열 저장됨⁠

Heewon Oh +1: Claude Sonnet 4.5

07129ee
12:57
feat(search): KR-SBERT 기업벡터 뉴스 필터 추가 (CPU 모드, HF Spaces용)⁠

Heewon Oh +1: Claude Sonnet 4.5

18e4e2b
12:46
fix(reasoning): EOD 형식 synthesis 파싱 수정 - conclusion 빈 문자열 버그⁠

Heewon Oh +1: Claude Sonnet 4.5

502a016
12:28
feat: fork pykrx naver OHLCV into naver_finance_client (HF Spaces safe)⁠

Heewon Oh +1: Claude Sonnet 4.5

1c9ab47
12:21
feat: align synthesis output with STONKS EOD deep report format⁠

Heewon Oh +1: Claude Sonnet 4.5

862ca4c
06:54
docs: note sampling config alignment with llama-cpp-python defaults

Heewon Oh

cd3a181

Feb 16

16:17
fix: handle Conclusions plural, add key takeaways to junk, improve content dedup

Heewon Oh

ca09410
16:03
fix: add h2 truncation, content dedup, and training leak cleanup in conclusion

Heewon Oh

014beeb
15:48
fix: fix boilerplate body length check and add table debris removal

Heewon Oh

8da491a
15:33
fix: improve conclusion truncation and boilerplate removal for 7B model

Heewon Oh

92d2642
15:12
fix: expand conclusion pattern to match Conclusions and Key Takeaways variant

Heewon Oh

3015d81
14:58
fix: improve conclusion truncation with separator and length limits⁠

Heewon Oh +1: Claude Opus 4.6

72fea0f
14:48
fix: truncate junk sections generated after conclusion⁠

Heewon Oh +1: Claude Opus 4.6

b6e3def
14:36
fix: expand boilerplate detection patterns and add inline junk removal⁠

Heewon Oh +1: Claude Opus 4.6

b093cce
14:28
fix: remove 7B model boilerplate sections from synthesis output⁠

Heewon Oh +1: Claude Opus 4.6

6cb4161
14:21
fix: add stop sequences for chat pattern and remove hallucinated metadata⁠

Heewon Oh +1: Claude Opus 4.6

7722810
14:15
fix: section dedup bypass for Korean headers and reduce synthesis token limit⁠

Heewon Oh +1: Claude Opus 4.6

47ab998
14:04
fix: rollback ZeroGPU to minimal generation params matching MLX approach⁠

Heewon Oh +1: Claude Opus 4.6

b102a57
13:39
fix: prevent prompt template leakage and harden inference API path⁠

Heewon Oh +1: Claude Opus 4.6

7a53fed
13:25
fix: improve ZeroGPU output quality — generation params, repetition suppression, content filtering⁠

Heewon Oh +1: Claude Opus 4.6

78cee6b
13:12
fix: prevent stock data duplication in synthesis prompt⁠

Heewon Oh +1: Claude Opus 4.6

4882e6a
12:56
feat: add Naver Finance stock data integration (price/valuation/investor)⁠

Heewon Oh +1: Claude Opus 4.6

d30bb51
12:25
fix: synthesis confidence parser handles bold format⁠

Heewon Oh +1: Claude Opus 4.6

6a5a382
12:15
docs: add Production Enhancements section and Production Note column⁠

Heewon Oh +1: Claude Opus 4.6

0bdd82a
12:09
docs: add Known Limitations section to README and Gradio demo⁠

Heewon Oh +1: Claude Opus 4.6

2340caf
12:09
fix: query diversification and synthesis confidence fallback⁠

Heewon Oh +1: Claude Opus 4.6

2f898ee
11:57
fix(zerogpu): add paragraph-level repetition detection to postprocess⁠

Heewon Oh +1: Claude Opus 4.6

a2b189b
11:39
fix(reasoning): handle missing Thought prefix in model output⁠

Heewon Oh +1: Claude Opus 4.6

07afd91
11:33
fix(zerogpu): align gen_params with model generation_config⁠

Heewon Oh +1: Claude Opus 4.6

ddb30df
11:27
fix(zerogpu): remove no_repeat_ngram_size that blocks RT format⁠

Heewon Oh +1: Claude Opus 4.6

1ba6e9a
11:19
fix: replace explicit EOS with repetition prevention for ZeroGPU⁠

Heewon Oh +1: Claude Opus 4.6

a50fbdf
11:10
fix: add EOS token handling and output postprocessing for ZeroGPU⁠

Heewon Oh +1: Claude Opus 4.6

15ef6b8
10:17
docs: add HF Spaces YAML metadata to README⁠

Heewon Oh +1: Claude Opus 4.6

2f7f464

Commit History

Feb 18

fix(runpod): align sampling params with model card (top_k=40, top_p=0.95, rep=1.0)⁠

fix(runpod): add explicit vLLM sampling params (top_k, top_p, repetition_penalty)⁠

fix(runpod): handle tokens array and usage input/output format in response parser⁠

feat(app): ZeroGPU 쿼터 소진 시 RunPod Serverless fallback 추가⁠

fix(agent): step.query 누락 시 원본 쿼리로 fallback (모델이 Query 필드 생략 시 검색 스킵 방지)

fix(reasoning): EOD 래퍼 헤더 제거 및 결론→투자 의견 정규화

Feb 17

fix(reasoning): EOD 형식에서 _remove_boilerplate 제거로 섹션 손실 방지

fix(prompts): synthesis 프롬프트를 모델 학습 포맷(h3)에 맞게 단순화⁠

fix(zerogpu): PicklingError 수정 - GPU 함수 내부에서 agent 생성⁠

fix(zerogpu): 단일 GPU 컨텍스트로 전체 research 실행⁠

fix(app): ZeroGPU 호환성 - 백그라운드 스레드 제거 및 메인 스레드 동기 실행⁠

fix(agent): sentence-transformers 추가 및 ActionType 직렬화 수정⁠

fix(synthesis): remove Key Findings suffix from synthesis prompt and fix ActionType serialization⁠

feat(vectors): revert to KR-SBERT 768-dim with CPU lazy load⁠

feat(vectors): replace KR-SBERT with MiniLM for HF Serverless compatibility⁠

fix(vector_filter): HF_TOKEN 명시적 전달 (API 401 수정)⁠

fix(vector_filter): sentence-transformers→HF InferenceClient로 교체 (ZeroGPU CUDA 충돌 수정)⁠

fix(search): source_type.value 오류 수정 - use_enum_values=True로 문자열 저장됨⁠

feat(search): KR-SBERT 기업벡터 뉴스 필터 추가 (CPU 모드, HF Spaces용)⁠

fix(reasoning): EOD 형식 synthesis 파싱 수정 - conclusion 빈 문자열 버그⁠

feat: fork pykrx naver OHLCV into naver_finance_client (HF Spaces safe)⁠

feat: align synthesis output with STONKS EOD deep report format⁠

docs: note sampling config alignment with llama-cpp-python defaults

Feb 16

fix: handle Conclusions plural, add key takeaways to junk, improve content dedup

fix: add h2 truncation, content dedup, and training leak cleanup in conclusion

fix: fix boilerplate body length check and add table debris removal

fix: improve conclusion truncation and boilerplate removal for 7B model

fix: expand conclusion pattern to match Conclusions and Key Takeaways variant

fix: improve conclusion truncation with separator and length limits⁠

fix: truncate junk sections generated after conclusion⁠

fix: expand boilerplate detection patterns and add inline junk removal⁠

fix: remove 7B model boilerplate sections from synthesis output⁠

fix: add stop sequences for chat pattern and remove hallucinated metadata⁠

fix: section dedup bypass for Korean headers and reduce synthesis token limit⁠

fix: rollback ZeroGPU to minimal generation params matching MLX approach⁠

fix: prevent prompt template leakage and harden inference API path⁠

fix: improve ZeroGPU output quality — generation params, repetition suppression, content filtering⁠

fix: prevent stock data duplication in synthesis prompt⁠

feat: add Naver Finance stock data integration (price/valuation/investor)⁠

fix: synthesis confidence parser handles bold format⁠

docs: add Production Enhancements section and Production Note column⁠

docs: add Known Limitations section to README and Gradio demo⁠

fix: query diversification and synthesis confidence fallback⁠

fix(zerogpu): add paragraph-level repetition detection to postprocess⁠

fix(reasoning): handle missing Thought prefix in model output⁠

fix(zerogpu): align gen_params with model generation_config⁠

fix(zerogpu): remove no_repeat_ngram_size that blocks RT format⁠

fix: replace explicit EOS with repetition prevention for ZeroGPU⁠

fix: add EOS token handling and output postprocessing for ZeroGPU⁠

docs: add HF Spaces YAML metadata to README⁠