myhloli
7f250e2331
feat(api)!: make Tasks the only asynchronous parsing API
...
Converge the SDK, Gradio, remote CLI, Doclib and Router on Tasks with hash-based uploads, per-file result URLs, task listing and optional model output. Preserve Files and the synchronous file_parse wrapper while removing standalone Uploads/Jobs and local-path HTTP inputs.
Migrate contract, lifecycle, transport and real-document regression coverage; update documentation and HTTP examples. Validation: 1452 passed, 3 conditional skips, plus targeted final PDF checks.
BREAKING CHANGE: clients and workers must use Tasks; standalone Uploads/Jobs endpoints and local-source/inline HTTP options have been removed.
2026-10-22 21:39:42 +08:00
jinzhenj and Claude Code
a980e9f23e
docs: add design for remote 3.x middle JSON compatibility
...
Co-Authored-By: Claude Code <noreply@anthropic.com >
2026-09-17 18:32:05 +08:00
myhloli
349f94e294
Update packaging documentation to reflect changes in backend installation and clarify dependency management
2026-09-11 18:52:50 +08:00
jinzhenj
32d90209bd
Merge branch 'next' into pr-5415-merge
2026-08-23 17:56:27 +08:00
jinzhenj
66aabd40f0
docs: define E2E model stack coverage
2026-08-14 17:23:01 +08:00
myhloli
fb8546c4a9
feat: refactor backend structure to separate PDF processing from hybrid components
2026-08-07 10:30:05 +08:00
jinzhenj
dde677561d
feat: use dedicated managed parse control channel
2026-07-27 16:14:10 +08:00
jinzhenj
e6060774f9
docs: disable implicit CLI markup
2026-07-22 21:23:48 +08:00
jinzhenj
f5d547b9ef
docs: define safe server log rendering
2026-07-22 21:05:25 +08:00
jinzhenj
b1e2fcd55a
docs: define managed parse server stop budget
2026-07-22 20:49:13 +08:00
jinzhenj
6602571953
fix: stop PDF render workers when parent exits
2026-07-22 19:33:07 +08:00
jinzhenj and Claude Opus 4.7
33efbd45ec
feat: add --preload-models for parse server startup model preloading
...
Add opt-in model preloading during api-server startup so that model
health is validated at boot rather than on first parse. On failure the
health/models/tiers endpoints return 503 with a specific error code
(dependency_missing, files_missing, device_unavailable, or failed).
Managed parse-server mode automatically enables preloading and uses
differentiated restart logic: non-retryable preload errors (missing
deps/files/device) suppress restarts, while transient failures and
startup timeouts trigger retries via the new startup_timeout_sec config.
Clean up HealthResponse to remove non-standard parser_version and
models fields.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com >
2026-07-22 18:02:42 +08:00
jinzhenj
1be121f241
docs: define PDF render pool diagnostics
2026-07-22 13:57:34 +08:00
jinzhenj
0d8aaf22b0
docs: centralize doclib owner diagnostics
2026-07-22 13:30:47 +08:00
jinzhenj
6429e45f6b
docs: define doclib lock error boundary
2026-07-22 13:27:50 +08:00
jinzhenj
9300b1ecd4
docs: simplify doclib lock exception
2026-07-22 13:24:08 +08:00
jinzhenj
25990c8369
docs: unify doclib ownership errors
2026-07-22 12:20:46 +08:00
jinzhenj
f12256777f
docs: clarify doclib owner diagnostics
2026-07-22 12:07:43 +08:00
jinzhenj
dd373ec829
docs: define doclib home ownership lock
2026-07-22 11:15:08 +08:00
jinzhenj
9535a7bcf6
fix: handle legacy doclib server shutdown
...
Refs #5298
2026-07-21 19:53:26 +08:00
jinzhenj
611ae7a541
docs: define legacy doclib shutdown handling
2026-07-21 19:42:25 +08:00
John King
66ef0d1fcd
Merge pull request #5308 from johnking0099/fix/issue-5298-endpoint-discovery
...
fix: bind doclib discovery to endpoint instance
2026-07-21 19:26:31 +08:00
jinzhenj
cadeee7099
docs: define doclib server identity validation
2026-07-21 19:08:11 +08:00
jinzhenj
e9012b70e7
Merge branch 'next' of github.com:opendatalab/MinerU into next
2026-07-21 19:05:31 +08:00
jinzhenj
ff730d7d79
docs: design legacy tier database migration
2026-07-21 18:54:24 +08:00
jinzhenj
d6a54885de
docs: define endpoint-only doclib discovery
2026-07-21 18:39:29 +08:00
jinzhenj
0df7ae8bb4
docs: design hf-xet dependency handling
2026-07-21 18:24:46 +08:00
jinzhenj
4dda7601bd
refactor: simplify tier selection and dependencies
2026-07-21 16:57:29 +08:00
jinzhenj
2e488049a9
docs: design simplified default tier selection
2026-07-21 16:37:52 +08:00
jinzhenj
1f75e3fb33
docs: unify deployment tier concepts
2026-07-21 16:29:19 +08:00
jinzhenj
f2e76f2d3b
docs: keep tier ranks explicit
2026-07-21 16:21:12 +08:00
jinzhenj
ff2eb50eba
docs: derive tier ranks from ordered tiers
2026-07-21 16:17:36 +08:00
jinzhenj
b40b04bfdc
docs: define ordered public tiers
2026-07-21 16:02:09 +08:00
jinzhenj
669779d54e
docs: remove superseded tier identifiers
2026-07-21 15:56:51 +08:00
jinzhenj
3e83ee2c17
docs: design centralized tier types
2026-07-21 15:55:11 +08:00
jinzhenj
4fb52adc73
docs: design shared CLI version command
2026-07-21 15:43:49 +08:00
jinzhenj
66ebb170ee
docs: clarify direct api server registration
2026-07-21 15:40:06 +08:00
jinzhenj
cc69723cd0
docs: extend mineru-kit command registration design
2026-07-21 15:37:18 +08:00
jinzhenj
744334b6ab
docs: design mineru-kit command registration
2026-07-21 15:29:01 +08:00
jinzhenj
082a3ac974
docs: design api server startup tiers
2026-07-21 15:05:12 +08:00
jinzhenj
f1d48c42f7
refactor: limit model tiers to basic and standard
2026-07-21 14:27:30 +08:00
jinzhenj
8867c0bf84
docs: define model tier scope
2026-07-21 14:13:56 +08:00
jinzhenj
57aa965d41
refactor: rename parsing tiers
2026-07-21 12:28:30 +08:00
jinzhenj
a6c4c88890
docs: define tier naming design
2026-07-21 11:50:50 +08:00
jinzhenj
d218389c6a
Document null-tier text search behavior
2026-07-17 15:15:15 +08:00
jinzhenj
4e237a1ebe
Document text read behavior
2026-07-17 14:59:58 +08:00
jinzhenj
9f29ce3411
Document text invalidation behavior
2026-07-17 14:55:23 +08:00
jinzhenj
8646a072c7
Document text parse behavior
...
Refs #5276
2026-07-17 14:29:58 +08:00
jinzhenj
c79c1ee343
Fix Alpha release metadata validation
2026-07-14 17:30:34 +08:00
jinzhenj
337115ab9b
chore: change alpha release workflow
2026-07-14 17:14:08 +08:00
myhloli