papermerge · ciur · Dec 8, 2024 · Dec 8, 2024 · Dec 8, 2024
diff --git a/docker/standard/Dockerfile b/docker/standard/Dockerfile
@@ -31,6 +31,7 @@ RUN apk update && apk add linux-headers python3-dev \
 
 RUN pip install --upgrade poetry roco==0.4.2
 RUN curl -L -o /bin/env2js /~https://github.com/papermerge/env2js/releases/download/0.2/env2js.amd64
+RUN chmod +x /bin/env2js
 
 COPY poetry.lock pyproject.toml README.md LICENSE ${CORE_APP}/
 

diff --git a/docker/standard/entrypoint.sh b/docker/standard/entrypoint.sh
@@ -57,13 +57,17 @@ case $CMD in
     exec_init
     # TODO: replace roco with env2js
     roco > /usr/share/nginx/html/auth_server/papermerge-runtime-config.js
+    # Once user options endpoint is implemented, following two lines will removed
     /bin/env2js -f /core_app/core.js.tmpl > /usr/share/nginx/html/ui/papermerge-runtime-config.js
+    sed -i '/Papermerge/a  <script type="module" src="/papermerge-runtime-config.js"></script>' /usr/share/nginx/html/ui/index.html
     exec /usr/bin/supervisord -c /etc/papermerge/supervisord.conf
     ;;
   server_without_init)
     # TODO: replace roco with env2js
     roco > /usr/share/nginx/html/auth_server/papermerge-runtime-config.js
+    # Once user options endpoint is implemented, following two lines will removed
     /bin/env2js -f /core_app/core.js.tmpl > /usr/share/nginx/html/ui/papermerge-runtime-config.js
+    sed -i '/Papermerge/a  <script type="module" src="/papermerge-runtime-config.js"></script>' /usr/share/nginx/html/ui/index.html
     exec /usr/bin/supervisord -c /etc/papermerge/supervisord.conf
     ;;
   create_token.sh)

diff --git a/papermerge/app.py b/papermerge/app.py
@@ -20,6 +20,7 @@
 )
 from papermerge.core.features.liveness_probe.router import router as probe_router
 from papermerge.search.routers.search import router as search_router
+from papermerge.core.features.tasks.router import router as tasks_router
 
 from papermerge.core.version import __version__
 from papermerge.core.config import get_settings
@@ -49,6 +50,7 @@
 app.include_router(tags_router, prefix=prefix)
 app.include_router(groups_router, prefix=prefix)
 app.include_router(probe_router, prefix=prefix)
+app.include_router(tasks_router, prefix=prefix)
 
 if config.papermerge__search__url:
     app.include_router(search_router, prefix=prefix)

diff --git a/papermerge/core/features/nodes/router.py b/papermerge/core/features/nodes/router.py
@@ -95,7 +95,7 @@ def create_node(
         # if user does not specify document's language, get that
         # value from user preferences
         if pynode.lang is None:
-            pynode.lang = settings.papermerge__ocr__default_language
+            pynode.lang = settings.papermerge__ocr__default_lang_code
 
         attrs = dict(
             title=pynode.title,

diff --git a/papermerge/core/features/tasks/router.py b/papermerge/core/features/tasks/router.py
@@ -0,0 +1,35 @@
+from typing import Annotated
+
+from fastapi import APIRouter, Security
+
+from papermerge.core import constants, schema, utils
+from papermerge.core.features.auth import get_current_user, scopes
+from papermerge.core import tasks
+
+from .schema import OCRTaskIn
+
+router = APIRouter(
+    prefix="/tasks",
+    tags=["tasks"],
+)
+
+
+@router.post("/ocr")
+@utils.docstring_parameter(scope=scopes.TASK_OCR)
+def start_ocr(
+    ocr_task: OCRTaskIn,
+    user: Annotated[schema.User, Security(get_current_user, scopes=[scopes.TASK_OCR])],
+):
+    """Triggers OCR for specific document
+
+    Required scope: `{scope}`
+    """
+
+    tasks.send_task(
+        constants.WORKER_OCR_DOCUMENT,
+        kwargs={
+            "document_id": str(ocr_task.document_id),
+            "lang": ocr_task.lang,
+        },
+        route_name="ocr",
+    )
diff --git a/papermerge/core/features/tasks/schema.py b/papermerge/core/features/tasks/schema.py
@@ -0,0 +1,34 @@
+from typing import Literal
+from uuid import UUID
+
+from pydantic import BaseModel
+
+LangCode = Literal[
+    "ces",
+    "dan",
+    "deu",
+    "ell",
+    "eng",
+    "fas",
+    "fin",
+    "fra",
+    "guj",
+    "heb",
+    "hin",
+    "ita",
+    "jpn",
+    "kor",
+    "lit",
+    "nld",
+    "nor",
+    "pol",
+    "por",
+    "ron",
+    "san",
+    "spa",
+]
+
+
+class OCRTaskIn(BaseModel):
+    document_id: UUID  # document model ID
+    lang: LangCode
diff --git a/papermerge/core/routers/tasks.py b/papermerge/core/routers/tasks.py
diff --git a/papermerge/core/schemas/tasks.py b/papermerge/core/schemas/tasks.py
diff --git a/ui2/index.html b/ui2/index.html
@@ -9,7 +9,6 @@
   <body>
     <div id="root"></div>
     <div id="modals"></div>
-    <script type="module" src="/papermerge-runtime-config.js"></script>
     <script type="module" src="/src/main.tsx"></script>
   </body>
 </html>
diff --git a/ui2/papermerge-runtime-config.js b/ui2/papermerge-runtime-config.js
diff --git a/ui2/src/components/ScheduleOCRProcess.tsx b/ui2/src/components/ScheduleOCRProcess.tsx
@@ -0,0 +1,38 @@
+import {useRuntimeConfig} from "@/hooks/runtime_config"
+import {OCRCode} from "@/types/ocr"
+import {langCodes2ComboboxData} from "@/utils"
+import {Select, Stack} from "@mantine/core"
+import {useState} from "react"
+
+interface Args {
+  defaultLang: OCRCode
+  onLangChange: (newLang: OCRCode) => void
+}
+
+export default function ScheduleOCRProcessCheckbox({
+  defaultLang,
+  onLangChange
+}: Args) {
+  const runtimeConfig = useRuntimeConfig()
+  const langData = langCodes2ComboboxData(runtimeConfig.ocr__lang_codes)
+  const [lang, setLang] = useState<OCRCode>(defaultLang)
+
+  const onLangChangeLocal = (value: string | null) => {
+    if (value) {
+      setLang(value as OCRCode)
+      onLangChange(value as OCRCode)
+    }
+  }
+
+  return (
+    <Stack my={"md"}>
+      <Select
+        searchable
+        defaultValue={defaultLang}
+        onChange={onLangChangeLocal}
+        data={langData}
+        value={lang}
+      />
+    </Stack>
+  )
+}
diff --git a/ui2/src/components/ScheduleOCRProcessCheckbox/ScheduleOCRProcessCheckbox.tsx b/ui2/src/components/ScheduleOCRProcessCheckbox/ScheduleOCRProcessCheckbox.tsx
@@ -1,7 +1,7 @@
-import {OCR_LANG} from "@/cconstants"
 import {useRuntimeConfig} from "@/hooks/runtime_config"
 import {OCRCode} from "@/types/ocr"
-import {Checkbox, ComboboxData, Select, Stack} from "@mantine/core"
+import {langCodes2ComboboxData} from "@/utils"
+import {Checkbox, Select, Stack} from "@mantine/core"
 import {useState} from "react"
 
 interface Args {
@@ -55,31 +55,3 @@ export default function ScheduleOCRProcessCheckbox({
     </Stack>
   )
 }
-
-function langCodes2ComboboxData(langCodes: string): ComboboxData {
-  /*
-  Input/Output examples:
-  example 1:
-
-      input:  "deu,eng,ron"
-      output: [
-        {value: "deu", label: "Deutsch"},
-        {value: "eng", label: "English"},
-        {value: "ron", label: "Română"}
-      ]
-
-  example 2:
-
-    input:  "fra,spa"
-    output: [
-      {value: "fra", label: "Français"},
-      {value: "spa", label: "Español"},
-    ]
-  */
-  return langCodes
-    .split(",")
-    .map(v => v.trim())
-    .map(v => {
-      return {value: v, label: OCR_LANG[v] || "Unknown Code"}
-    })
-}
diff --git a/ui2/src/features/document/components/ActionButtons.tsx b/ui2/src/features/document/components/ActionButtons.tsx
@@ -2,6 +2,7 @@ import {useAppDispatch, useAppSelector} from "@/app/hooks"
 import ToggleSecondaryPanel from "@/components/DualPanel/ToggleSecondaryPanel"
 import PanelContext from "@/contexts/PanelContext"
 import {updateActionPanel} from "@/features/ui/uiSlice"
+import {useRuntimeConfig} from "@/hooks/runtime_config"
 import {Group} from "@mantine/core"
 import {useViewportSize} from "@mantine/hooks"
 import {useContext, useEffect, useRef} from "react"
@@ -14,13 +15,15 @@ import type {PanelMode} from "@/types"
 import DownloadButton from "./DownloadButton/DownloadButton"
 import RotateButton from "./RotateButton"
 import RotateCCButton from "./RotateCCButton"
+import RunOCRButton from "./RunOCRButton"
 
 export default function ActionButtons() {
   const {height, width} = useViewportSize()
   const dispatch = useAppDispatch()
   const ref = useRef<HTMLDivElement>(null)
   const mode: PanelMode = useContext(PanelContext)
   const selectedPages = useAppSelector(s => selectSelectedPages(s, mode)) || []
+  const runtimeConfig = useRuntimeConfig()
 
   useEffect(() => {
     if (ref?.current) {
@@ -40,6 +43,7 @@ export default function ActionButtons() {
       <Group>
         <EditTitleButton />
         <DownloadButton />
+        {!runtimeConfig.ocr__automatic && <RunOCRButton />}
         {selectedPages.length > 0 && <RotateButton />}
         {selectedPages.length > 0 && <RotateCCButton />}
         {selectedPages.length > 0 && <DeletePagesButton />}

diff --git a/ui2/src/features/document/components/RunOCRButton.tsx b/ui2/src/features/document/components/RunOCRButton.tsx
@@ -0,0 +1,46 @@
+import {useAppSelector} from "@/app/hooks"
+
+import PanelContext from "@/contexts/PanelContext"
+import {selectCurrentNodeID} from "@/features/ui/uiSlice"
+import {ActionIcon, Box, Tooltip} from "@mantine/core"
+import {useDisclosure} from "@mantine/hooks"
+import {IconEye} from "@tabler/icons-react"
+import {forwardRef, useContext} from "react"
+
+import type {PanelMode} from "@/types"
+import {RunOCRModal} from "./RunOCRModal"
+
+interface Args {
+  hidden?: boolean
+}
+
+const RunOCRButton = forwardRef<HTMLButtonElement, Args>((props, ref) => {
+  const {hidden} = props
+  const [opened, {open, close}] = useDisclosure(false)
+  const mode: PanelMode = useContext(PanelContext)
+  const currentNodeID = useAppSelector(s => selectCurrentNodeID(s, mode))
+
+  return (
+    <Box>
+      <Tooltip label="Run OCR" withArrow>
+        <ActionIcon
+          style={hidden ? {display: "None"} : {}}
+          ref={ref}
+          size={"lg"}
+          variant="default"
+          onClick={open}
+        >
+          <IconEye stroke={1.4} />
+        </ActionIcon>
+      </Tooltip>
+      <RunOCRModal
+        opened={opened}
+        node_id={currentNodeID!}
+        onSubmit={close}
+        onCancel={close}
+      />
+    </Box>
+  )
+})
+
+export default RunOCRButton