sajitkun125 commited on
Commit
28735f7
·
verified ·
1 Parent(s): b2fac69

Upload folder using huggingface_hub

Browse files
Dockerfile CHANGED
@@ -22,7 +22,6 @@ RUN pip install --no-cache-dir -r requirements-space.txt
22
  COPY app/pyproject.toml ./
23
  COPY app/src/ ./src/
24
  COPY app/frontend/ ./frontend/
25
- COPY app/deploy/seed_data/ ./seed_data/ # prebuilt all-books DB + FAISS index (~40 MB)
26
  RUN pip install --no-cache-dir -e . --no-deps && \
27
  mkdir -p /data && chown -R user:user /app /data
28
 
@@ -43,7 +42,9 @@ ENV INKFERENCE_DATA_ROOT=/data \
43
  LLM_PROVIDER=groq \
44
  LLM_MODEL=openai/gpt-oss-120b \
45
  LLM_FALLBACK=gemini:gemini-2.5-flash-lite \
46
- RAG_USE_CORRECTED=true
 
 
47
  # REQUIRED Space secrets: GROQ_API_KEY, GEMINI_API_KEY
48
  # REQUIRED Space variable (page scans): INKFERENCE_IMAGES_BASE_URL=
49
  # https://huggingface.co/datasets/<user>/inkference-book-images/resolve/main
@@ -55,7 +56,14 @@ USER user
55
  RUN python -c "from sentence_transformers import SentenceTransformer as S; S('sentence-transformers/all-MiniLM-L6-v2')"
56
  RUN python -c "from transformers import TrOCRProcessor, VisionEncoderDecoderModel as M; TrOCRProcessor.from_pretrained('microsoft/trocr-base-handwritten'); M.from_pretrained('microsoft/trocr-base-handwritten')"
57
 
 
 
 
 
 
 
 
58
  EXPOSE 7860
59
 
60
  # Copy the prebuilt corpus into the (ephemeral) data root on boot, then serve. No re-seed.
61
- CMD ["sh", "-c", "cp -rn /app/seed_data/. /data/ 2>/dev/null || true; uvicorn inkference.api.main:app --host 0.0.0.0 --port ${PORT:-7860}"]
 
22
  COPY app/pyproject.toml ./
23
  COPY app/src/ ./src/
24
  COPY app/frontend/ ./frontend/
 
25
  RUN pip install --no-cache-dir -e . --no-deps && \
26
  mkdir -p /data && chown -R user:user /app /data
27
 
 
42
  LLM_PROVIDER=groq \
43
  LLM_MODEL=openai/gpt-oss-120b \
44
  LLM_FALLBACK=gemini:gemini-2.5-flash-lite \
45
+ RAG_USE_CORRECTED=true \
46
+ # dataset holding the prebuilt corpus (seed_data/) AND the scans (book*/forster*/)
47
+ SEED_DATASET=sajitkun125/inkference-book-images
48
  # REQUIRED Space secrets: GROQ_API_KEY, GEMINI_API_KEY
49
  # REQUIRED Space variable (page scans): INKFERENCE_IMAGES_BASE_URL=
50
  # https://huggingface.co/datasets/<user>/inkference-book-images/resolve/main
 
56
  RUN python -c "from sentence_transformers import SentenceTransformer as S; S('sentence-transformers/all-MiniLM-L6-v2')"
57
  RUN python -c "from transformers import TrOCRProcessor, VisionEncoderDecoderModel as M; TrOCRProcessor.from_pretrained('microsoft/trocr-base-handwritten'); M.from_pretrained('microsoft/trocr-base-handwritten')"
58
 
59
+ # Fetch the prebuilt DB + FAISS index from the dataset. snapshot_download RESOLVES LFS
60
+ # to real bytes (unlike the Space's Docker build, which ships LFS as pointers). Only the
61
+ # seed_data/ folder is pulled — the 1.3 GB scans stay remote and stream via the CDN.
62
+ RUN python -c "import os; from huggingface_hub import snapshot_download; \
63
+ snapshot_download(os.environ['SEED_DATASET'], repo_type='dataset', \
64
+ allow_patterns=['seed_data/*'], local_dir='/app/hub')"
65
+
66
  EXPOSE 7860
67
 
68
  # Copy the prebuilt corpus into the (ephemeral) data root on boot, then serve. No re-seed.
69
+ CMD ["sh", "-c", "cp -rn /app/hub/seed_data/. /data/ 2>/dev/null || true; uvicorn inkference.api.main:app --host 0.0.0.0 --port ${PORT:-7860}"]
app/deploy/seed_data/index/doc_1.faiss DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:dff3e63c3a13fb3d5cc9f23c2239d4bb270aa496bc0658179b6a2356fb6ffa91
3
- size 4108845
 
 
 
 
app/deploy/seed_data/index/doc_1.json DELETED
The diff for this file is too large to render. See raw diff
 
app/deploy/seed_data/inkference.db DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:10c7568dfd0d88563bd5f19d71a6330f47fcfe61a229f132c91364386252f67b
3
- size 34762752