Akiragpu / modules

Commit History

  1. feat(api): skills funcionam via intencao direta no provider local⁠

    Isaac Quarenta
    958bee6
  2. fix: Akira responde 'entendi'/'hm' sem entender⁠

    Isaac Quarenta
    bd8eb23

  1. local-gpu: Qwen2.5-7B-Instruct 4-bit (~5GB, faster load, smarter)

    Isaac Quarenta
    dc575e5
  2. local-gpu: compact system prompt (3B can't follow huge prompts)

    Isaac Quarenta
    85d1c03
  3. local-llm: swap 8B 16GB for Qwen2.5-3B (~6GB) — loads in ~10s on ZeroGPU FREE

    Isaac Quarenta
    89bd0bd
  4. fix: nvidia timeout 60s; hf_inference before llama in chain

    Isaac Quarenta
    8fd89b1
  5. persona: tom sério/curto estilo Kiami

    Isaac Quarenta
    b6444da
  6. nvidia: chain order ahead of mistral; timeout 60s

    Isaac Quarenta
    26cd4c0
  7. fix(app): eliminar crash ChatbotDataMessages 'ping' + spam asyncio; NVIDIA default deepseek-v4.1-flash⁠

    Isaac Quarenta
    cf29563
  8. feat(providers): NVIDIA NIM como provider extra (build.nvidia.com)⁠

    Isaac Quarenta
    c487576
  9. feat(botcore): PV simples espelha o index (buttonId, midia, filtros) + sync api/config⁠

    Isaac Quarenta
    42306bf

  1. Require explicit web search requests⁠

    Isaac Quarenta +1: Copilot
    9e310cf
  2. fix(persona): tom serio-coloquial-logico-racional + fim de respostas com aspas⁠

    Isaac Quarenta
    52cd142
  3. Improve Akira conversational coherence⁠

    Isaac Quarenta +1: Copilot
    5fba0c5
  4. fix(chat+hf): fluidez conversacional sem busca + HF tier com quota real por conta⁠

    Isaac Quarenta
    5fb9e9a
  5. fix(persona): tom serio-coloquial-racional + declaracoes de afeto + strip de aspas em loop

    Isaac Quarenta
    e705438
  6. fix(chat+images): schnell primeiro no HF + interjeicoes/identidade sem web search

    Isaac Quarenta
    02995ae
  7. Prioritize HF FLUX for image generation⁠

    Isaac Quarenta +1: Copilot
    d841dbe
  8. fix(images): HF tier usa InferenceClient oficial (router+nscale dava 400 em tudo e caia sempre no Pollinations) - FLUX.1-dev primeiro

    Isaac Quarenta
    bdf1c54

  1. fix(images): HF Inference com modelos sem gate + diagnostico de licenca gated-403

    Isaac Quarenta
    6e7884a
  2. fix(chat): busca web com timeout 20s + aviso honesto quando nao ha resultados

    Isaac Quarenta
    ebe19ca
  3. fix(chat): gatilho de busca usa deve_pesquisar() (perguntas factuais/temporais tambem vao a web)

    Isaac Quarenta
    d53aaf8
  4. fix(chat): pesquisa web real no caminho Gradio - rewrite deixa de substituir a mensagem e injeta resultados da web

    Isaac Quarenta
    4ea6707
  5. fix(chat): anti-loop de saida (_collapse_repetition) + busca factual alargada + PG retry apos cooldown

    Isaac Quarenta
    08e22c7

  1. perf(ext-gpu): max 180 tokens + deteccao de tunel morto (ngrok 404)

    Isaac Quarenta
    0ba1a0c
  2. fix(api): LocalLLMFallback compat + spaces.GPU duration 39 (free tier) - fim do ImportError

    Isaac Quarenta
    886a735
  3. feat(ui+gpu): app Claude-like minimalista + local_llm no padrao brain (ZeroGPU) — sigilosos so dono HF

    Isaac Quarenta
    5136df6
  4. feat(chain): external_gpu provider na chain (skills+websearch+persona) + fix historico Gradio6 sem [brackets]

    Isaac Quarenta
    9340691
  5. fix(generate): BatchEncoding->input_ids/attention_mask explicitos (fim do 500/AttributeError shape)

    Isaac Quarenta
    902b1f1
  6. feat(ext-gpu): estado visivel no chat e na aba Status (diagnostico de prioridade)

    Isaac Quarenta
    006d3e8
  7. feat(ext-gpu): prioridade automatica quando ZeroGPU esgota + kaggle_backend com erros detalhados e compat torch_dtype

    Isaac Quarenta
    054a72a
  8. fix(ext-gpu): header ngrok-skip-browser-warning (ngrok free interstitial)

    Isaac Quarenta
    896ab23
  9. feat(gpu-ext): fallback GPU externa 4-bit (Kaggle T4) + AKIRA_QUANT toggle + kaggle_backend.py

    Isaac Quarenta
    7c8bb54
  10. fix(api): faltavam emotion_analyzer (property) e _warm_emotion_analyzer - __init__ rebentava ao inicio⁠

    Isaac Quarenta
    c2bb650
  11. fix(boot): ciclo infinito em get_current_api_token pendurava o init da API⁠

    Isaac Quarenta
    1bd10ef
  12. fix(chat): singleton anti-veneno + _akira_reply defensivo (local_gpu_direct/graceful_sem_chaves) - fim do AttributeError providers

    Isaac Quarenta
    b5fc087
  13. refactor(gpu): padrao brain - spaces primeiro, bf16+.to(cuda) sem 4-bit, warmup processo principal

    Isaac Quarenta
    0e55a7e
  14. fix(hf): snapshot_download autenticado via rotacao de tokens + hf-transfer (fim do 429) + fix Gradio JSON tuple

    Isaac Quarenta
    ce42495
  15. perf(boot)+fix(gpu): NLP/embeddings forcados a CPU, emotion trainer lazy, Docker removido do Akiragpu⁠

    Isaac Quarenta
    3c0fde3
  16. chore: forca carga em GPU local_llm + sonda CUDA segura no startup

    Isaac Quarenta
    5f89d31
  17. fix(gpu): forca 100% GPU - _nlp_device=0, embeddings cuda, +diffusers/torchvision p/ FLUX+PhiVision

    Isaac Quarenta
    8e2acc2
  18. feat(vision): Phi-3.5 Vision + FLUX local GPU (prefetch no build, lazy no boot) - fixa travamento

    Isaac Quarenta
    0951d77
  19. feat(status): device via nvidia-smi (NVML, sem contexto CUDA) p/ identificar a GPU actual no Status

    Isaac Quarenta
    298bf42
  20. fix(status): sondagem de GPU/bitsandbytes apenas dentro do contexto spaces.GPU (fora pendurava o arranque)

    Isaac Quarenta
    daa766f
  21. feat(status): reporta dispositivo ZeroGPU actual (device/VRAM/capability) + versoes torch/bnb p/ confirmar suporte Blackwell

    Isaac Quarenta
    729353d
  22. perf(gpu): warmup da API fora da janela GPU + resposta curta na 1a chamada (load consome ~30s dos 60s do FREE)

    Isaac Quarenta
    ffe504c
  23. fix(gpu): cache explicito em /data (evita download dentro da janela GPU) + deteccao de VRAM libertada + cooldown apos quota esgotada⁠

    Isaac Quarenta
    a57148b
  24. feat(gpu): GPU ativo por omissao (40s) + modelo local 8B com prefetch para /data⁠

    Isaac Quarenta
    69df760
  25. feat(chat): sem providers ativos responde ja com instrucao de Secrets (graceful_sem_chaves) em vez de recursar

    Isaac Quarenta
    b7e396f
  26. feat(zerogpu): free tier por defeito + anti-recursao graceful + circuit breaker PG⁠

    Isaac Quarenta
    89fb197