local_converter imports docling at module level, so any test that
touches it crashes when docling is not installed. Skip those tests
with importorskip / skipif so the CI (which only has docling-core)
passes cleanly.
The docling library (torch, transformers) is too heavy for lightweight
CI environments that only install docling-core. Use importorskip so
these tests are cleanly skipped instead of crashing collection.
Adapt test expectations to external changes: upload returns 200,
ValueError yields 400, and schemas now accept both snake_case and
camelCase via AliasChoices.
Cover PDF validation, file size limits, preview generation, page
counting, file deletion with path traversal protection, and the
not-found case — all previously untested code paths.
Lightweight sliding-window per-IP rate limiter (100 req/min default)
with no external dependency. Health endpoint is excluded. Returns 429
with Retry-After header when exceeded. Sufficient for single-process
SQLite deployments; document the Redis upgrade path for scale.
Domain model state transitions, repository CRUD operations, and service
methods now have descriptive docstrings to lower the contribution
barrier for an open-source project.
Read uploaded files in fixed-size chunks instead of a single file.read()
to reduce peak memory pressure. Also enforces the size limit during
streaming so oversized payloads are rejected before fully buffered.
Unbounded asyncio.create_task calls could exhaust CPU and memory on
modest hardware. Add a configurable semaphore (MAX_CONCURRENT_ANALYSES,
default 3) so excess jobs queue instead of running all at once.
All endpoint functions, lifespan context manager, and get_connection now
have explicit return types. Upload endpoint returns 201 Created instead
of 200 to follow REST conventions.
Replace broad `except Exception` with specific types (FileNotFoundError,
PermissionError, OSError) so errors are properly categorized in logs.
Users reporting bugs will get actionable messages instead of generic ones.
UPLOAD_DIR and DB_PATH were read directly from os.environ, bypassing
the Settings dataclass. This caused an inconsistency where overriding
Settings had no effect on these values. Now all modules import from
infra.settings.settings.
domain/ must be pure with no external dependencies. bbox.py imports
docling_core and belongs in infra/. Also refactor ServeConverter to
use the canonical to_topleft_list via BoundingBox instead of
duplicated manual coordinate conversion. Move docling-core to base
requirements since it is now needed in both modes.
Clarify the difference between CONVERSION_MODE (docker-compose
build target) and CONVERSION_ENGINE (runtime variable set
automatically by the Docker target).
Move /health to /api/health on backend and update the frontend
feature flag store to match. Without this, the combined Docker
image nginx proxy could not reach the endpoint and feature flags
(chunking/prepare mode) failed to load.
Install torch and torchvision from the CPU-only index before docling
to avoid pulling CUDA/nvidia/triton dependencies. Update documentation
with measured image sizes (270MB remote, 1.9GB local).
Reflect the two Docker targets across README, getting-started,
contributing guides, and .env.example with new configuration
variables (CONVERSION_ENGINE, DOCLING_SERVE_URL, DOCLING_SERVE_API_KEY).
Both root and backend Dockerfiles now use a shared base stage with
two targets: remote (lightweight, ~300MB) and local (full Docling,
~2-3GB). docker-compose uses CONVERSION_MODE to select the target.
Replace generic orange "D" placeholders with the duck mascot logo
in all branding touchpoints: navbar header, home page hero, studio
import view, and browser favicon.
Add document and clock icons above the stat numbers for better visual
context. Accent border and icon color on hover for consistent
interactive feedback across the app.
Reduce toggle width to fit-content with inner padding and individual
border-radius, matching the studio tabs design. Narrower max-width
for a tighter settings layout.
Replace flat rows with rounded cards matching documents page style.
Reduce COMPLETED badge to a green dot so exceptions (FAILED, RUNNING)
stand out visually instead of blending into a wall of green badges.
