1365 Commits
Author SHA1 Message Date
ssongliu 92c3173aec feat(ai): sync gateway providers and model discovery (#14426)
* feat(ai): sync gateway providers and model discovery

Add provider account types and protocol-aware model discovery, and replace predefined models with editable model pools.

Align provider icons with AI Gateway, preserve international Coding Plan restrictions, and filter account counts consistently with selectable accounts.

* fix(ai): preserve Ollama base paths during model discovery

* fix(ai): restore custom DashScope image protocol support
2026-11-13 14:04:11 +08:00
ssongliu a922e5bfae fix: isolate GPU monitoring failures and scope device warnings (#14421) 2026-11-12 23:27:43 +08:00
ssongliu 806de6307b feat(ai): add streaming model chat drawer (#14412) 2026-11-12 15:58:09 +08:00
ssongliu 5afcedc235 fix(vllm): unify default images and upgrade suffix rules (#14406) 2026-11-11 21:19:22 +08:00
ssongliu 3adee94458 feat(vllm): integrate instance monitoring lifecycle and settings (#14391) 2026-11-10 20:56:28 +08:00
ssongliu d0439ea9b2 fix(ai): expose sanitized verification errors and clarify vLLM authentication (#14388) 2026-11-10 17:38:34 +08:00
ssongliu aec5684184 feat(ai): streamline GPU groups and NUMA configuration (#14381) 2026-11-10 15:27:38 +08:00
ssongliu a4572e0ca4 fix: correct backup account name search (#14375) 2026-11-10 10:42:21 +08:00
Yuki Guo 7d7a1a69d2 feat: add url agent (#14368)
* feat: add url agent

* feat: add url agent
2026-11-09 23:54:25 +08:00
蘭 4d5438315c feat(ops-report): implement OpsReport functionality and initialization settings (#14365)
* feat(ops-report): implement OpsReport functionality and initialization settings

* refactor(ops-report): remove unused OpsReport functionality and improve export handling

* fix: update report context messages in multiple languages
2026-11-09 18:31:43 +08:00
ssongliu 528486cabd fix(ai): unify GPU detection and improve group topology (#14367) 2026-11-09 18:20:37 +08:00
ssongliu c31898e666 fix(ntp): preserve UTC offset when synchronizing system time (#14363)
Fixes #14265
2026-11-09 16:37:59 +08:00
Kadxy 21d34431b2 feat(openresty): add proxy timeout settings to performance tuning (#14313)
* feat(openresty): add proxy timeout settings to performance tuning

Expose proxy_connect_timeout, proxy_send_timeout and proxy_read_timeout
at http level, so a slow backend no longer hits the 60s 504 without a
way to raise it from the panel.

The stock nginx.conf does not set them, so an absent directive is shown
with nginx's built-in 60s and only written once changed. Writing it
unconditionally would clash with the same directive set at http level
by an included file and fail the config test.

Time directives now keep their unit suffix like the size directives do.
keepalive_timeout previously dropped it, saving 5m back as 5 seconds,
and its helper text no longer reads as a generic connection timeout.

Refs #13824

* fix(openresty): keep performance settings from overwriting http-level values

Reading the http scope used FindDirectives, which descends into server
and location blocks, while saving writes the http block. A site-level
proxy_read_timeout 2s was shown as the global value and written back
over the global 5m when only gzip was changed. Read only the http
block's own directives.

A directive nginx.conf leaves unset may still be set at the top level
of an included file. Report that value with its file, show it
read-only, and leave it out of the save, instead of showing nginx's
default and writing a duplicate that fails the config test.

Durations are now edited as written and only sent when changed, so
1m30s and keepalive_timeout's second value round-trip intact. Input is
validated against nginx's duration syntax.

Saving is disabled until the current values have loaded, so the form
defaults cannot be written over the real configuration.
2026-11-09 15:50:08 +08:00
ssongliu ed182b5408 feat(ai): add peak analysis (#14358) 2026-11-09 14:17:47 +08:00
Yuki Guo 8f5cfd18c9 Merge pull request #14353 from 1Panel-dev/pr@dev-v2@fix_log
fix(container): resolve compose project name when cleaning app logs
2026-11-09 14:14:11 +08:00
guoyuqi 66f2dddab8 fix(container): resolve compose project name when cleaning app logs 2026-11-09 10:55:29 +08:00
Leo Camus 4c58c712c2 fix(app): keep env values unchanged when upgrading an app (#14264)
renderUpgradeEnv still wrote the .env with godotenv.Marshal, which turns
any value that parses as an integer into its canonical form. A password
such as 0123 was rewritten as 123 on upgrade, the same bug #14180 fixed
for install. Use the env package marshaller, and escape $ inside double
quotes there so values containing both $ and a quote still read back
unchanged.
2026-11-09 10:23:57 +08:00
Yuki Guo b4db4d2ca4 Merge pull request #14292 from RaphaelFakhri/fix/compose-log-clean-project-name
fix(container): resolve compose project name when cleaning app logs
2026-11-09 09:54:04 +08:00
ssongliu 9a83fa4af2 feat(ai): add GPU groups and NUMA binding (#14348)
* feat(ai): add GPU groups and NUMA binding

* fix(ai): normalize NVIDIA GPU index aliases before validation
2026-11-09 09:54:02 +08:00
蘭 00feb6e23b fix: enforce sensitive file sources and persistent history provenance (#14344) 2026-11-08 22:41:30 +08:00
蘭 f3eafd9a2e fix: guard AI endpoint probes and discovery (#14347) 2026-11-08 22:40:28 +08:00
蘭 ab90f4bc2d fix: confine download names to the selected directory (#14345) 2026-11-08 22:40:02 +08:00
ssongliu 5c71dcba19 feat(ai): share compose preview helpers and translations (#14342) 2026-11-08 17:24:01 +08:00
ssongliu 1927e68010 fix: align model deployment config and GPU environment status (#14332) 2026-11-07 17:14:21 +08:00
ssongliu ab506e4a3b feat(container): add NVIDIA GPU runtime configuration switch (#14331) 2026-11-07 15:39:38 +08:00
王贺 58c81b9255 fix: align GB10 dual-node vLLM deployments (#14330) 2026-11-07 15:39:06 +08:00
Raphael Fakhri 78c36875fc fix(container): reject ambiguous compose projects when cleaning app logs
When several projects use the same compose file, pick the project whose
name matches the request name, exactly or after normalization. If none
matches, return an error instead of the first project in the container
list, so log cleaning never stops and truncates another project.
2026-11-06 20:55:58 +00:00
王贺andssongliu 95f12c16c3 feat(ai): add GPU environment diagnostics and device summaries (#14322)
* feat(ai): add GPU environment diagnostics and device summaries

* chore(ai): remove GPU diagnostics comments and test files

* feat: separate host and container GPU environment status

* fix(ai): correct GPU environment diagnostics and timeout handling

---------

Co-authored-by: ssongliu <sloooop1x@gmail.com>
2026-11-06 23:16:10 +08:00
ssongliu 8a68d30dcb feat(ai): refine model service details and container log streaming (#14317) 2026-11-06 16:00:40 +08:00
ssongliu 23a54653f8 fix(ai): persist model verification settings and reset lifecycle status (#14312)
* fix(ai): persist model verification settings and reset lifecycle status

* fix(ai): reset model verification before upgrade cutover
2026-11-06 00:44:50 +08:00
ssongliu 024a246a46 refactor(ai): align menu order and simplify model routes (#14308) 2026-11-05 17:08:15 +08:00
ssongliu fe3953963e refactor(ai): move model preparation to XPack and reorganize model views (#14306) 2026-11-05 16:27:06 +08:00
Yuki Guoandssongliu d658395188 feat(sync): sync php extended template (#14299)
* feat(sync): sync php extended template

* fix(i18n): complete resource sync translations

---------

Co-authored-by: ssongliu <sloooop1x@gmail.com>
2026-11-05 15:06:12 +08:00
ssongliu 51bae946e1 feat(ai): refine model service terminology and provider setup (#14305) 2026-11-05 14:58:17 +08:00
Yuki Guo 56c7981b3a fix(docker): check image update (#14302) 2026-11-05 14:03:55 +08:00
ssongliu 730dbbab5b fix(vllm): distinguish empty metrics from collection failures (#14303) 2026-11-04 20:42:58 +08:00
ssongliu b07afd3785 feat(ai): improve model preparation and download progress refresh (#14297) 2026-11-04 16:11:16 +08:00
ssongliu 89787a55ee feat(ai): refine model deployment UI and support GB10 upgrades (#14295) 2026-11-04 12:28:30 +08:00
蘭 3a30f69b91 fix(mcp): unify website binding error feedback and fix status refresh and retries (#14290) 2026-11-04 09:54:19 +08:00
ssongliu 30381a3c68 feat(ai): streamline deployments and add local model selection (#14293)
* feat(ai): split model management and add weight provider filters

* feat(ai): streamline deployments and add local model selection
2026-11-03 23:54:22 +08:00
Raphael Fakhri 391ca4fc49 fix(container): resolve compose project name when cleaning app logs 2026-11-03 13:42:02 +00:00
Kadxy 37170ae81a fix(openresty): report brotli as pending when the http.d include can be added (#14285)
Installs that enabled ngx_brotli before v2.3.0 have no http.d include in
nginx.conf, because the include is only inserted when a module is toggled
or built. The settings page reported this as "cannot write the managed
configuration" and skipped the brotli save, so the save path that inserts
the include was never reached, even on a stock nginx.conf.

The settings response now tells the two cases apart with an in-memory
insertion check that writes nothing:

- include missing but insertable, or managed file missing: managedPending,
  the page explains the values take effect after saving and saves normally
- no safe insertion point: managedUnavailable, unchanged

The save path now inserts the include through ensureNginxHTTPIncludeActive,
the same function module activation uses, so both produce identical
results. A sentinel error keeps "no insertion point" mapped to
ErrBrotliUnsupported while I/O failures are still returned as-is.
2026-11-03 18:58:24 +08:00
ssongliu daae8c43f8 refactor: remove Fxplay and TensorRT-LLM support (#14287) 2026-11-03 14:38:15 +08:00
ssongliu fac15e3292 fix(ai): refine model service forms and Ollama drawers (#14283) 2026-11-03 00:07:38 +08:00
蘭 e29679768b feat: add MCP error metadata handling and improve source policy resolution (#14270)
* feat: add MCP error metadata handling and improve source policy resolution

* fix: adjust height-diff for improved table layout
2026-11-02 22:17:12 +08:00
CityFun 8e1b761f4c fix(app): allow upgrades from removed versions (#14280) 2026-11-02 22:17:03 +08:00
ssongliu 2401bb16a2 feat(ai): connect local model accounts and add Ollama management page (#14282)
* feat(ai): connect local model accounts and add Ollama management page

* fix(ai): register Ollama management frontend path
2026-11-02 22:16:43 +08:00
ssongliu c06c4e338b feat(vm): enable community virtual machines and migrate xpack data (#14279)
* feat(vm): enable community virtual machines and migrate xpack data

* fix(vm): stop ISO download polling on failure or cancellation

* fix(vm): respect deleteFile when deleting failed ISO imports
2026-11-02 18:22:58 +08:00
ssongliu f8c61993c5 feat(ai): open source model marketplace and migrate historical data (#14277)
* feat(ai): open source model marketplace and migrate historical data

* fix(ai): wait for canceled model downloads before retrying
2026-11-02 17:46:26 +08:00
ssongliu b31e683f1b fix: correct local inference engine operation task logs (#14274) 2026-11-02 17:27:43 +08:00