Merge pull request #32 from scub-france/feature/include-chunking

Feature/include chunking

document-parser/api/analyses.py

@@ -7,7 +7,7 @@ from typing import Annotated
 
 from fastapi import APIRouter, Depends, HTTPException, Request
 
-from api.schemas import AnalysisResponse, CreateAnalysisRequest
+from api.schemas import AnalysisResponse, ChunkResponse, CreateAnalysisRequest, RechunkRequest
 from services.analysis_service import AnalysisService
 
 logger = logging.getLogger(__name__)
@@ -30,6 +30,8 @@ def _to_response(job) -> AnalysisResponse:
         content_markdown=job.content_markdown,
         content_html=job.content_html,
         pages_json=job.pages_json,
+        chunks_json=job.chunks_json,
+        has_document_json=job.document_json is not None,
         error_message=job.error_message,
         started_at=str(job.started_at) if job.started_at else None,
         completed_at=str(job.completed_at) if job.completed_at else None,
@@ -47,8 +49,16 @@ async def create_analysis(body: CreateAnalysisRequest, service: ServiceDep):
     if body.pipelineOptions:
         pipeline_opts = body.pipelineOptions.model_dump()
 
+    chunking_opts = None
+    if body.chunkingOptions:
+        chunking_opts = body.chunkingOptions.model_dump()
+
     try:
-        job = await service.create(body.documentId, pipeline_options=pipeline_opts)
+        job = await service.create(
+            body.documentId,
+            pipeline_options=pipeline_opts,
+            chunking_options=chunking_opts,
+        )
     except ValueError as e:
         raise HTTPException(status_code=404, detail=str(e)) from e
 
@@ -71,6 +81,24 @@ async def get_analysis(job_id: str, service: ServiceDep):
     return _to_response(job)
 
 
+@router.post("/{job_id}/rechunk", response_model=list[ChunkResponse])
+async def rechunk_analysis(job_id: str, body: RechunkRequest, service: ServiceDep):
+    """Re-chunk a completed analysis with new chunking options."""
+    try:
+        chunks = await service.rechunk(job_id, body.chunkingOptions.model_dump())
+    except ValueError as e:
+        raise HTTPException(status_code=400, detail=str(e)) from e
+    return [
+        ChunkResponse(
+            text=c.text,
+            headings=c.headings,
+            source_page=c.source_page,
+            token_count=c.token_count,
+        )
+        for c in chunks
+    ]
+
+
 @router.delete("/{job_id}", status_code=204)
 async def delete_analysis(job_id: str, service: ServiceDep):
     """Delete an analysis job."""

document-parser/api/schemas.py

@@ -18,6 +18,7 @@ def _to_camel(name: str) -> str:
 
 class _CamelModel(BaseModel):
     """Base model that serializes field names to camelCase."""
+
     model_config = ConfigDict(
         alias_generator=_to_camel,
         populate_by_name=True,
@@ -42,6 +43,8 @@ class AnalysisResponse(_CamelModel):
     content_markdown: str | None = None
     content_html: str | None = None
     pages_json: str | None = None
+    chunks_json: str | None = None
+    has_document_json: bool = False
     error_message: str | None = None
     started_at: str | datetime | None = None
     completed_at: str | datetime | None = None
@@ -50,6 +53,7 @@ class AnalysisResponse(_CamelModel):
 
 class PipelineOptionsRequest(BaseModel):
     """Docling pipeline configuration options."""
+
     do_ocr: bool = True
     do_table_structure: bool = True
     table_mode: str = "accurate"  # "accurate" or "fast"
@@ -76,6 +80,41 @@ class PipelineOptionsRequest(BaseModel):
         return v
 
 
+class ChunkingOptionsRequest(BaseModel):
+    """Docling chunking configuration options."""
+
+    chunker_type: str = "hybrid"  # "hybrid", "hierarchical"
+    max_tokens: int = 512
+    merge_peers: bool = True
+    repeat_table_header: bool = True
+
+    @field_validator("chunker_type")
+    @classmethod
+    def validate_chunker_type(cls, v: str) -> str:
+        if v not in ("hybrid", "hierarchical"):
+            raise ValueError('chunker_type must be "hybrid" or "hierarchical"')
+        return v
+
+    @field_validator("max_tokens")
+    @classmethod
+    def validate_max_tokens(cls, v: int) -> int:
+        if v < 64 or v > 8192:
+            raise ValueError("max_tokens must be between 64 and 8192")
+        return v
+
+
+class ChunkResponse(_CamelModel):
+    text: str
+    headings: list[str] = []
+    source_page: int | None = None
+    token_count: int = 0
+
+
 class CreateAnalysisRequest(BaseModel):
     documentId: str  # camelCase to match existing frontend contract
     pipelineOptions: PipelineOptionsRequest | None = None
+    chunkingOptions: ChunkingOptionsRequest | None = None
+
+
+class RechunkRequest(BaseModel):
+    chunkingOptions: ChunkingOptionsRequest

document-parser/conftest.py

@@ -1,3 +1 @@
-
-
 pytest_plugins = ["pytest_asyncio"]

document-parser/domain/bbox.py

@@ -43,7 +43,11 @@ def to_topleft_list(bbox: BoundingBox, page_height: float) -> list[float]:
     if right <= left or bottom <= top:
         logger.debug(
             "Degenerate bbox skipped: [%.1f, %.1f, %.1f, %.1f] (page_height=%.1f)",
-            left, top, right, bottom, page_height,
+            left,
+            top,
+            right,
+            bottom,
+            page_height,
         )
         return list(EMPTY_BBOX)
 

document-parser/domain/models.py

@@ -42,6 +42,8 @@ class AnalysisJob:
     content_markdown: str | None = None
     content_html: str | None = None
     pages_json: str | None = None
+    document_json: str | None = None
+    chunks_json: str | None = None
     error_message: str | None = None
     started_at: datetime | None = None
     completed_at: datetime | None = None
@@ -55,12 +57,19 @@ class AnalysisJob:
         self.started_at = _utcnow()
 
     def mark_completed(
-        self, markdown: str, html: str, pages_json: str,
+        self,
+        markdown: str,
+        html: str,
+        pages_json: str,
+        document_json: str | None = None,
+        chunks_json: str | None = None,
     ) -> None:
         self.status = AnalysisStatus.COMPLETED
         self.content_markdown = markdown
         self.content_html = html
         self.pages_json = pages_json
+        self.document_json = document_json
+        self.chunks_json = chunks_json
         self.completed_at = _utcnow()
 
     def mark_failed(self, error: str) -> None:

document-parser/domain/ports.py

@@ -9,7 +9,12 @@ from __future__ import annotations
 from typing import TYPE_CHECKING, Protocol
 
 if TYPE_CHECKING:
-    from domain.value_objects import ConversionOptions, ConversionResult
+    from domain.value_objects import (
+        ChunkingOptions,
+        ChunkResult,
+        ConversionOptions,
+        ConversionResult,
+    )
 
 
 class DocumentConverter(Protocol):
@@ -20,5 +25,20 @@ class DocumentConverter(Protocol):
     """
 
     async def convert(
-        self, file_path: str, options: ConversionOptions,
+        self,
+        file_path: str,
+        options: ConversionOptions,
     ) -> ConversionResult: ...
+
+
+class DocumentChunker(Protocol):
+    """Port for document chunking.
+
+    Takes a serialized DoclingDocument (JSON) and returns chunks.
+    """
+
+    async def chunk(
+        self,
+        document_json: str,
+        options: ChunkingOptions,
+    ) -> list[ChunkResult]: ...

document-parser/domain/value_objects.py

@@ -49,3 +49,23 @@ class ConversionResult:
     content_html: str
     pages: list[PageDetail]
     skipped_items: int = 0
+    document_json: str | None = None
+
+
+@dataclass
+class ChunkingOptions:
+    chunker_type: str = "hybrid"  # "hybrid", "hierarchical", "page"
+    max_tokens: int = 512
+    merge_peers: bool = True
+    repeat_table_header: bool = True
+
+    def is_default(self) -> bool:
+        return self == ChunkingOptions()
+
+
+@dataclass
+class ChunkResult:
+    text: str
+    headings: list[str] = field(default_factory=list)
+    source_page: int | None = None
+    token_count: int = 0

document-parser/infra/local_chunker.py

@@ -0,0 +1,84 @@
+"""Local Docling chunker — runs chunking in-process using docling-core.
+
+This adapter implements the DocumentChunker port. It deserializes a
+DoclingDocument from JSON, applies the requested chunker, and returns
+domain ChunkResult objects.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+
+from docling_core.transforms.chunker import HierarchicalChunker
+from docling_core.transforms.chunker.hybrid_chunker import HybridChunker
+from docling_core.types.doc.document import DoclingDocument
+
+from domain.value_objects import ChunkingOptions, ChunkResult
+
+logger = logging.getLogger(__name__)
+
+
+def _chunk_sync(document_json: str, options: ChunkingOptions) -> list[ChunkResult]:
+    if not document_json or not document_json.strip():
+        raise ValueError("Empty document JSON — nothing to chunk")
+
+    try:
+        doc_data = json.loads(document_json)
+    except json.JSONDecodeError as e:
+        raise ValueError(f"Malformed document JSON: {e}") from e
+
+    doc = DoclingDocument.model_validate(doc_data)
+
+    chunker = _build_chunker(options)
+    results: list[ChunkResult] = []
+
+    for chunk in chunker.chunk(doc):
+        source_page = None
+        token_count = 0
+
+        if hasattr(chunk, "meta") and chunk.meta and chunk.meta.doc_items:
+            for doc_item in chunk.meta.doc_items:
+                if hasattr(doc_item, "prov") and doc_item.prov:
+                    source_page = doc_item.prov[0].page_no
+                    break
+
+        if hasattr(chunker, "tokenizer") and chunker.tokenizer:
+            token_count = chunker.tokenizer.count_tokens(chunk.text)
+
+        headings = list(chunk.meta.headings) if chunk.meta and chunk.meta.headings else []
+
+        results.append(
+            ChunkResult(
+                text=chunk.text,
+                headings=headings,
+                source_page=source_page,
+                token_count=token_count,
+            )
+        )
+
+    logger.info("Chunked document into %d chunks (chunker=%s)", len(results), options.chunker_type)
+    return results
+
+
+def _build_chunker(options: ChunkingOptions) -> HierarchicalChunker | HybridChunker:
+    if options.chunker_type == "hierarchical":
+        return HierarchicalChunker()
+
+    return HybridChunker(
+        max_tokens=options.max_tokens,
+        merge_peers=options.merge_peers,
+        repeat_table_header=options.repeat_table_header,
+    )
+
+
+class LocalChunker:
+    """Adapter that runs docling-core chunking locally."""
+
+    async def chunk(
+        self,
+        document_json: str,
+        options: ChunkingOptions,
+    ) -> list[ChunkResult]:
+        return await asyncio.to_thread(_chunk_sync, document_json, options)

document-parser/infra/local_converter.py

@@ -9,6 +9,7 @@ from __future__ import annotations
 
 import asyncio
 import contextlib
+import json
 import logging
 import threading
 
@@ -83,6 +84,7 @@ def _get_element_type(item: DocItem) -> str:
 # Pipeline factory
 # ---------------------------------------------------------------------------
 
+
 def _build_docling_converter(options: ConversionOptions) -> DoclingConverter:
     table_options = TableStructureOptions(
         do_cell_matching=True,
@@ -126,6 +128,7 @@ def _select_converter(options: ConversionOptions) -> DoclingConverter:
 # Page extraction
 # ---------------------------------------------------------------------------
 
+
 def _extract_pages_detail(doc_result) -> tuple[list[PageDetail], int]:
     pages: dict[int, PageDetail] = {}
     document = doc_result.document
@@ -149,7 +152,9 @@ def _extract_pages_detail(doc_result) -> tuple[list[PageDetail], int]:
 
 
 def _process_content_item(
-    item: DocItem | GroupItem, level: int, pages: dict[int, PageDetail],
+    item: DocItem | GroupItem,
+    level: int,
+    pages: dict[int, PageDetail],
 ) -> bool:
     if isinstance(item, GroupItem):
         return True
@@ -163,9 +168,13 @@ def _process_content_item(
             if page_no not in pages:
                 logger.warning(
                     "Page %d not found in document metadata — using US Letter fallback (%sx%s pt)",
-                    page_no, _DEFAULT_PAGE_WIDTH, _DEFAULT_PAGE_HEIGHT,
+                    page_no,
+                    _DEFAULT_PAGE_WIDTH,
+                    _DEFAULT_PAGE_HEIGHT,
+                )
+                pages[page_no] = PageDetail(
+                    page_number=page_no, width=_DEFAULT_PAGE_WIDTH, height=_DEFAULT_PAGE_HEIGHT
                 )
-                pages[page_no] = PageDetail(page_number=page_no, width=_DEFAULT_PAGE_WIDTH, height=_DEFAULT_PAGE_HEIGHT)
 
             page_height = pages[page_no].height
 
@@ -199,6 +208,7 @@ def _process_content_item(
 # Synchronous conversion (called via asyncio.to_thread)
 # ---------------------------------------------------------------------------
 
+
 def _convert_sync(file_path: str, options: ConversionOptions) -> ConversionResult:
     with _converter_lock:
         conv = _select_converter(options)
@@ -213,7 +223,9 @@ def _convert_sync(file_path: str, options: ConversionOptions) -> ConversionResul
             PageDetail(
                 page_number=i + 1,
                 width=doc.pages[i + 1].size.width if (i + 1) in doc.pages else _DEFAULT_PAGE_WIDTH,
-                height=doc.pages[i + 1].size.height if (i + 1) in doc.pages else _DEFAULT_PAGE_HEIGHT,
+                height=doc.pages[i + 1].size.height
+                if (i + 1) in doc.pages
+                else _DEFAULT_PAGE_HEIGHT,
             )
             for i in range(page_count)
         ]
@@ -227,6 +239,7 @@ def _convert_sync(file_path: str, options: ConversionOptions) -> ConversionResul
         content_html=doc.export_to_html(),
         pages=pages_detail,
         skipped_items=skipped,
+        document_json=json.dumps(doc.export_to_dict()),
     )
 
 
@@ -234,10 +247,13 @@ def _convert_sync(file_path: str, options: ConversionOptions) -> ConversionResul
 # Public adapter class
 # ---------------------------------------------------------------------------
 
+
 class LocalConverter:
     """Adapter that runs Docling locally as a Python library."""
 
     async def convert(
-        self, file_path: str, options: ConversionOptions,
+        self,
+        file_path: str,
+        options: ConversionOptions,
     ) -> ConversionResult:
         return await asyncio.to_thread(_convert_sync, file_path, options)

document-parser/infra/serve_converter.py

@@ -71,7 +71,9 @@ class ServeConverter:
         return headers
 
     async def convert(
-        self, file_path: str, options: ConversionOptions,
+        self,
+        file_path: str,
+        options: ConversionOptions,
     ) -> ConversionResult:
         """Convert a document by uploading it to Docling Serve."""
         path = Path(file_path)
@@ -202,7 +204,9 @@ def _add_element(item: dict, pages: dict[int, PageDetail]) -> None:
         page_no = prov.get("page_no", 1)
         if page_no not in pages:
             pages[page_no] = PageDetail(
-                page_number=page_no, width=612.0, height=792.0,
+                page_number=page_no,
+                width=612.0,
+                height=792.0,
             )
 
         bbox_data = prov.get("bbox", {})

document-parser/infra/settings.py

@@ -14,7 +14,9 @@ class Settings:
     conversion_timeout: int = 600
     upload_dir: str = "./uploads"
     db_path: str = "./data/docling_studio.db"
-    cors_origins: list[str] = field(default_factory=lambda: ["http://localhost:3000", "http://localhost:5173"])
+    cors_origins: list[str] = field(
+        default_factory=lambda: ["http://localhost:3000", "http://localhost:5173"]
+    )
 
     @classmethod
     def from_env(cls) -> Settings:

document-parser/main.py

@@ -40,6 +40,7 @@ def _build_converter():
     """Build the converter adapter based on configuration."""
     if settings.conversion_engine == "remote":
         from infra.serve_converter import ServeConverter
+
         logger.info("Using remote Docling Serve at %s", settings.docling_serve_url)
         return ServeConverter(
             base_url=settings.docling_serve_url,
@@ -47,14 +48,26 @@ def _build_converter():
         )
     else:
         from infra.local_converter import LocalConverter
+
         logger.info("Using local Docling converter")
         return LocalConverter()
 
 
+def _build_chunker():
+    """Build the chunker adapter — only available in local mode."""
+    if settings.conversion_engine == "local":
+        from infra.local_chunker import LocalChunker
+
+        return LocalChunker()
+    return None
+
+
 def _build_analysis_service() -> AnalysisService:
     converter = _build_converter()
+    chunker = _build_chunker()
     return AnalysisService(
         converter=converter,
+        chunker=chunker,
         conversion_timeout=settings.conversion_timeout,
     )
 
@@ -63,6 +76,7 @@ def _build_analysis_service() -> AnalysisService:
 # FastAPI app
 # ---------------------------------------------------------------------------
 
+
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     await init_db()

document-parser/persistence/analysis_repo.py

@@ -16,6 +16,7 @@ def _parse_dt(value: str | None) -> datetime | None:
 
 
 def _row_to_job(row) -> AnalysisJob:
+    keys = row.keys()
     return AnalysisJob(
         id=row["id"],
         document_id=row["document_id"],
@@ -23,11 +24,13 @@ def _row_to_job(row) -> AnalysisJob:
         content_markdown=row["content_markdown"],
         content_html=row["content_html"],
         pages_json=row["pages_json"],
+        document_json=row["document_json"] if "document_json" in keys else None,
+        chunks_json=row["chunks_json"] if "chunks_json" in keys else None,
         error_message=row["error_message"],
         started_at=_parse_dt(row["started_at"]),
         completed_at=_parse_dt(row["completed_at"]),
         created_at=_parse_dt(row["created_at"]) or datetime.now(),
-        document_filename=row["filename"] if "filename" in row.keys() else None,  # noqa: SIM118
+        document_filename=row["filename"] if "filename" in keys else None,
     )
 
 
@@ -60,9 +63,7 @@ async def find_all(*, limit: int = 200, offset: int = 0) -> list[AnalysisJob]:
 
 async def find_by_id(job_id: str) -> AnalysisJob | None:
     async with get_connection() as db:
-        cursor = await db.execute(
+        cursor = await db.execute(f"{_SELECT_WITH_DOC} WHERE aj.id = ?", (job_id,))
-            f"{_SELECT_WITH_DOC} WHERE aj.id = ?", (job_id,)
-        )
         row = await cursor.fetchone()
         return _row_to_job(row) if row else None
 
@@ -72,17 +73,36 @@ async def update_status(job: AnalysisJob) -> None:
         await db.execute(
             """UPDATE analysis_jobs
                SET status = ?, content_markdown = ?, content_html = ?,
-                   pages_json = ?, error_message = ?, started_at = ?, completed_at = ?
+                   pages_json = ?, document_json = ?, chunks_json = ?,
+                   error_message = ?, started_at = ?, completed_at = ?
                WHERE id = ?""",
-            (job.status.value, job.content_markdown, job.content_html,
+            (
-             job.pages_json, job.error_message,
+                job.status.value,
-             str(job.started_at) if job.started_at else None,
+                job.content_markdown,
-             str(job.completed_at) if job.completed_at else None,
+                job.content_html,
-             job.id),
+                job.pages_json,
+                job.document_json,
+                job.chunks_json,
+                job.error_message,
+                str(job.started_at) if job.started_at else None,
+                str(job.completed_at) if job.completed_at else None,
+                job.id,
+            ),
         )
         await db.commit()
 
 
+async def update_chunks(job_id: str, chunks_json: str) -> bool:
+    """Update only the chunks_json column for a completed analysis."""
+    async with get_connection() as db:
+        cursor = await db.execute(
+            "UPDATE analysis_jobs SET chunks_json = ? WHERE id = ?",
+            (chunks_json, job_id),
+        )
+        await db.commit()
+        return cursor.rowcount > 0
+
+
 async def delete(job_id: str) -> bool:
     async with get_connection() as db:
         cursor = await db.execute("DELETE FROM analysis_jobs WHERE id = ?", (job_id,))
@@ -93,8 +113,6 @@ async def delete(job_id: str) -> bool:
 async def delete_by_document(document_id: str) -> int:
     """Delete all analysis jobs for a given document. Returns count deleted."""
     async with get_connection() as db:
-        cursor = await db.execute(
+        cursor = await db.execute("DELETE FROM analysis_jobs WHERE document_id = ?", (document_id,))
-            "DELETE FROM analysis_jobs WHERE document_id = ?", (document_id,)
-        )
         await db.commit()
         return cursor.rowcount

document-parser/persistence/database.py

@@ -30,6 +30,8 @@ CREATE TABLE IF NOT EXISTS analysis_jobs (
     content_markdown  TEXT,
     content_html      TEXT,
     pages_json        TEXT,
+    document_json     TEXT,
+    chunks_json       TEXT,
     error_message     TEXT,
     started_at        TEXT,
     completed_at      TEXT,
@@ -42,11 +44,29 @@ CREATE INDEX IF NOT EXISTS idx_documents_created_at ON documents(created_at);
 """
 
 
+_MIGRATIONS = [
+    ("document_json", "ALTER TABLE analysis_jobs ADD COLUMN document_json TEXT"),
+    ("chunks_json", "ALTER TABLE analysis_jobs ADD COLUMN chunks_json TEXT"),
+]
+
+
+async def _run_migrations(db: aiosqlite.Connection) -> None:
+    """Add columns that may be missing in older databases."""
+    cursor = await db.execute("PRAGMA table_info(analysis_jobs)")
+    existing = {row[1] for row in await cursor.fetchall()}
+    for col_name, ddl in _MIGRATIONS:
+        if col_name not in existing:
+            await db.execute(ddl)
+            logger.info("Migration: added column %s to analysis_jobs", col_name)
+    await db.commit()
+
+
 async def init_db() -> None:
     """Create database file and tables if they don't exist."""
     os.makedirs(os.path.dirname(DB_PATH) or ".", exist_ok=True)
     async with aiosqlite.connect(DB_PATH) as db:
         await db.executescript(_SCHEMA)
+        await _run_migrations(db)
         await db.commit()
     logger.info("Database initialized at %s", DB_PATH)
 

document-parser/persistence/document_repo.py

@@ -28,8 +28,15 @@ async def insert(doc: Document) -> None:
         await db.execute(
             """INSERT INTO documents (id, filename, content_type, file_size, page_count, storage_path, created_at)
                VALUES (?, ?, ?, ?, ?, ?, ?)""",
-            (doc.id, doc.filename, doc.content_type, doc.file_size,
+            (
-             doc.page_count, doc.storage_path, str(doc.created_at)),
+                doc.id,
+                doc.filename,
+                doc.content_type,
+                doc.file_size,
+                doc.page_count,
+                doc.storage_path,
+                str(doc.created_at),
+            ),
         )
         await db.commit()
 

document-parser/services/analysis_service.py

@@ -13,24 +13,46 @@ import logging
 from dataclasses import asdict
 from typing import TYPE_CHECKING
 
-from domain.models import AnalysisJob
+from domain.models import AnalysisJob, AnalysisStatus
-from domain.value_objects import ConversionOptions, ConversionResult
+from domain.value_objects import ChunkingOptions, ChunkResult, ConversionOptions, ConversionResult
 
 if TYPE_CHECKING:
-    from domain.ports import DocumentConverter
+    from domain.ports import DocumentChunker, DocumentConverter
 from persistence import analysis_repo, document_repo
 
 logger = logging.getLogger(__name__)
 
 
+def _chunk_to_dict(c: ChunkResult) -> dict:
+    """Serialize ChunkResult to a camelCase dict matching the frontend API contract."""
+    return {
+        "text": c.text,
+        "headings": c.headings,
+        "sourcePage": c.source_page,
+        "tokenCount": c.token_count,
+    }
+
+
 class AnalysisService:
     """Orchestrates document analysis using an injected converter."""
 
-    def __init__(self, converter: DocumentConverter, conversion_timeout: int = 600):
+    def __init__(
+        self,
+        converter: DocumentConverter,
+        chunker: DocumentChunker | None = None,
+        conversion_timeout: int = 600,
+    ):
         self._converter = converter
+        self._chunker = chunker
         self._conversion_timeout = conversion_timeout
 
-    async def create(self, document_id: str, *, pipeline_options: dict | None = None) -> AnalysisJob:
+    async def create(
+        self,
+        document_id: str,
+        *,
+        pipeline_options: dict | None = None,
+        chunking_options: dict | None = None,
+    ) -> AnalysisJob:
         """Create a new analysis job and launch background processing."""
         doc = await document_repo.find_by_id(document_id)
         if not doc:
@@ -41,7 +63,13 @@ class AnalysisService:
         await analysis_repo.insert(job)
 
         task = asyncio.create_task(
-            self._run_analysis(job.id, doc.storage_path, doc.filename, pipeline_options)
+            self._run_analysis(
+                job.id,
+                doc.storage_path,
+                doc.filename,
+                pipeline_options,
+                chunking_options,
+            )
         )
         task.add_done_callback(functools.partial(_on_task_done, job_id=job.id))
 
@@ -56,10 +84,35 @@ class AnalysisService:
     async def delete(self, job_id: str) -> bool:
         return await analysis_repo.delete(job_id)
 
+    async def rechunk(self, job_id: str, chunking_options: dict) -> list[ChunkResult]:
+        """Re-chunk an existing completed analysis with new options."""
+        job = await analysis_repo.find_by_id(job_id)
+        if not job:
+            raise ValueError(f"Analysis not found: {job_id}")
+        if job.status != AnalysisStatus.COMPLETED:
+            raise ValueError(f"Analysis is not completed: {job_id}")
+        if not job.document_json:
+            raise ValueError(f"No document data available for re-chunking: {job_id}")
+        if not self._chunker:
+            raise ValueError("Chunking is not available")
+
+        options = ChunkingOptions(**chunking_options)
+        chunks = await self._chunker.chunk(job.document_json, options)
+
+        chunks_json = json.dumps([_chunk_to_dict(c) for c in chunks])
+        await analysis_repo.update_chunks(job_id, chunks_json)
+
+        return chunks
+
     async def _run_analysis(
-        self, job_id: str, file_path: str, filename: str, pipeline_options: dict | None = None,
+        self,
+        job_id: str,
+        file_path: str,
+        filename: str,
+        pipeline_options: dict | None = None,
+        chunking_options: dict | None = None,
     ) -> None:
-        """Background task: run conversion and update job status."""
+        """Background task: run conversion and optionally chunk."""
         try:
             job = await analysis_repo.find_by_id(job_id)
             if not job:
@@ -79,10 +132,19 @@ class AnalysisService:
 
             pages_json = json.dumps([asdict(p) for p in result.pages])
 
+            chunks_json = None
+            if chunking_options and self._chunker and result.document_json:
+                chunk_opts = ChunkingOptions(**chunking_options)
+                chunks = await self._chunker.chunk(result.document_json, chunk_opts)
+                chunks_json = json.dumps([_chunk_to_dict(c) for c in chunks])
+                logger.info("Chunking produced %d chunks for job %s", len(chunks), job_id)
+
             job.mark_completed(
                 markdown=result.content_markdown,
                 html=result.content_html,
                 pages_json=pages_json,
+                document_json=result.document_json,
+                chunks_json=chunks_json,
             )
             await analysis_repo.update_status(job)
 

document-parser/tests/test_api_endpoints.py

@@ -53,8 +53,12 @@ class TestDocumentEndpoints:
     @patch("services.document_service.find_by_id", new_callable=AsyncMock)
     def test_get_document(self, mock_find, client):
         mock_find.return_value = Document(
-            id="d1", filename="test.pdf", content_type="application/pdf",
+            id="d1",
-            file_size=2048, page_count=3, storage_path="/tmp/test",
+            filename="test.pdf",
+            content_type="application/pdf",
+            file_size=2048,
+            page_count=3,
+            storage_path="/tmp/test",
         )
 
         resp = client.get("/api/documents/d1")
@@ -74,8 +78,10 @@ class TestDocumentEndpoints:
     @patch("services.document_service.upload", new_callable=AsyncMock)
     def test_upload_document(self, mock_upload, client):
         mock_upload.return_value = Document(
-            id="new-1", filename="uploaded.pdf",
+            id="new-1",
-            content_type="application/pdf", file_size=512,
+            filename="uploaded.pdf",
+            content_type="application/pdf",
+            file_size=512,
             storage_path="/tmp/uploaded",
         )
 
@@ -115,9 +121,11 @@ class TestDocumentEndpoints:
 
 class TestAnalysisEndpoints:
     def test_list_analyses(self, client, mock_analysis_service):
-        mock_analysis_service.find_all = AsyncMock(return_value=[
+        mock_analysis_service.find_all = AsyncMock(
-            AnalysisJob(id="j1", document_id="d1", document_filename="test.pdf"),
+            return_value=[
-        ])
+                AnalysisJob(id="j1", document_id="d1", document_filename="test.pdf"),
+            ]
+        )
 
         resp = client.get("/api/analyses")
         assert resp.status_code == 200
@@ -146,37 +154,52 @@ class TestAnalysisEndpoints:
         assert resp.status_code == 404
 
     def test_create_analysis(self, client, mock_analysis_service):
-        mock_analysis_service.create = AsyncMock(return_value=AnalysisJob(
+        mock_analysis_service.create = AsyncMock(
-            id="j1", document_id="d1", document_filename="test.pdf",
+            return_value=AnalysisJob(
-        ))
+                id="j1",
+                document_id="d1",
+                document_filename="test.pdf",
+            )
+        )
 
         resp = client.post("/api/analyses", json={"documentId": "d1"})
         assert resp.status_code == 200
         data = resp.json()
         assert data["id"] == "j1"
         assert data["documentId"] == "d1"
-        mock_analysis_service.create.assert_called_once_with("d1", pipeline_options=None)
+        mock_analysis_service.create.assert_called_once_with(
+            "d1",
+            pipeline_options=None,
+            chunking_options=None,
+        )
 
     def test_create_analysis_with_pipeline_options(self, client, mock_analysis_service):
-        mock_analysis_service.create = AsyncMock(return_value=AnalysisJob(
+        mock_analysis_service.create = AsyncMock(
-            id="j2", document_id="d1", document_filename="test.pdf",
+            return_value=AnalysisJob(
-        ))
+                id="j2",
+                document_id="d1",
+                document_filename="test.pdf",
+            )
+        )
 
-        resp = client.post("/api/analyses", json={
+        resp = client.post(
-            "documentId": "d1",
+            "/api/analyses",
-            "pipelineOptions": {
+            json={
-                "do_ocr": False,
+                "documentId": "d1",
-                "do_table_structure": True,
+                "pipelineOptions": {
-                "table_mode": "fast",
+                    "do_ocr": False,
-                "do_code_enrichment": True,
+                    "do_table_structure": True,
-                "do_formula_enrichment": False,
+                    "table_mode": "fast",
-                "do_picture_classification": False,
+                    "do_code_enrichment": True,
-                "do_picture_description": False,
+                    "do_formula_enrichment": False,
-                "generate_picture_images": True,
+                    "do_picture_classification": False,
-                "generate_page_images": False,
+                    "do_picture_description": False,
-                "images_scale": 2.0,
+                    "generate_picture_images": True,
-            }
+                    "generate_page_images": False,
-        })
+                    "images_scale": 2.0,
+                },
+            },
+        )
         assert resp.status_code == 200
         data = resp.json()
         assert data["id"] == "j2"
@@ -191,14 +214,17 @@ class TestAnalysisEndpoints:
 
     def test_create_analysis_with_partial_pipeline_options(self, client, mock_analysis_service):
         """Pipeline options should use defaults for unspecified fields."""
-        mock_analysis_service.create = AsyncMock(return_value=AnalysisJob(
+        mock_analysis_service.create = AsyncMock(
-            id="j3", document_id="d1", document_filename="test.pdf",
+            return_value=AnalysisJob(
-        ))
+                id="j3",
+                document_id="d1",
+                document_filename="test.pdf",
+            )
+        )
 
-        resp = client.post("/api/analyses", json={
+        resp = client.post(
-            "documentId": "d1",
+            "/api/analyses", json={"documentId": "d1", "pipelineOptions": {"do_ocr": False}}
-            "pipelineOptions": {"do_ocr": False}
+        )
-        })
         assert resp.status_code == 200
 
         opts = mock_analysis_service.create.call_args.kwargs["pipeline_options"]

document-parser/tests/test_bbox.py

@@ -14,6 +14,7 @@ from domain.bbox import EMPTY_BBOX, to_topleft_list
 # Standard conversions
 # ---------------------------------------------------------------------------
 
+
 class TestToTopleftListStandard:
     """Normal bbox conversions (happy path)."""
 
@@ -29,10 +30,10 @@ class TestToTopleftListStandard:
         result = to_topleft_list(bbox, page_height=792.0)
 
         # After conversion: new_t = 792 - 700 = 92, new_b = 792 - 600 = 192
-        assert result[0] == 50       # l unchanged
+        assert result[0] == 50  # l unchanged
-        assert result[1] == pytest.approx(92.0)   # t = page_height - old_t
+        assert result[1] == pytest.approx(92.0)  # t = page_height - old_t
-        assert result[2] == 200      # r unchanged
+        assert result[2] == 200  # r unchanged
-        assert result[3] == pytest.approx(192.0)   # b = page_height - old_b
+        assert result[3] == pytest.approx(192.0)  # b = page_height - old_b
 
     def test_result_has_positive_dimensions(self):
         """Converted bbox should always have b > t (positive height)."""
@@ -60,6 +61,7 @@ class TestToTopleftListStandard:
 # Page format variations
 # ---------------------------------------------------------------------------
 
+
 class TestPageFormats:
     """Verify correct conversion across different page sizes."""
 
@@ -105,6 +107,7 @@ class TestPageFormats:
 # Degenerate / edge-case bboxes
 # ---------------------------------------------------------------------------
 
+
 class TestDegenerateBboxes:
     """Bboxes that are invalid or degenerate should return EMPTY_BBOX."""
 
@@ -151,6 +154,7 @@ class TestDegenerateBboxes:
 # Precision and boundary values
 # ---------------------------------------------------------------------------
 
+
 class TestPrecision:
     """Floating-point precision and edge values."""
 

document-parser/tests/test_chunking.py

@@ -0,0 +1,292 @@
+"""Tests for chunking feature — domain, schemas, service, and API endpoints."""
+
+from __future__ import annotations
+
+import json
+from dataclasses import asdict
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+from fastapi.testclient import TestClient
+
+from api.schemas import ChunkingOptionsRequest, ChunkResponse, RechunkRequest
+from domain.models import AnalysisJob, AnalysisStatus
+from domain.value_objects import ChunkingOptions, ChunkResult
+from main import app
+
+# ---------------------------------------------------------------------------
+# Domain: value objects
+# ---------------------------------------------------------------------------
+
+
+class TestChunkingOptions:
+    def test_defaults(self):
+        opts = ChunkingOptions()
+        assert opts.chunker_type == "hybrid"
+        assert opts.max_tokens == 512
+        assert opts.merge_peers is True
+        assert opts.repeat_table_header is True
+
+    def test_custom_values(self):
+        opts = ChunkingOptions(chunker_type="hierarchical", max_tokens=256, merge_peers=False)
+        assert opts.chunker_type == "hierarchical"
+        assert opts.max_tokens == 256
+        assert opts.merge_peers is False
+
+    def test_is_default(self):
+        assert ChunkingOptions().is_default()
+        assert not ChunkingOptions(max_tokens=256).is_default()
+
+
+class TestChunkResult:
+    def test_defaults(self):
+        chunk = ChunkResult(text="hello")
+        assert chunk.text == "hello"
+        assert chunk.headings == []
+        assert chunk.source_page is None
+        assert chunk.token_count == 0
+
+    def test_full_values(self):
+        chunk = ChunkResult(
+            text="content",
+            headings=["Title", "Section"],
+            source_page=3,
+            token_count=42,
+        )
+        assert chunk.headings == ["Title", "Section"]
+        assert chunk.source_page == 3
+        assert chunk.token_count == 42
+
+    def test_serializable(self):
+        chunk = ChunkResult(text="x", headings=["h1"], source_page=1, token_count=10)
+        data = asdict(chunk)
+        assert data == {"text": "x", "headings": ["h1"], "source_page": 1, "token_count": 10}
+
+
+# ---------------------------------------------------------------------------
+# Domain: AnalysisJob with chunking fields
+# ---------------------------------------------------------------------------
+
+
+class TestAnalysisJobChunking:
+    def test_default_chunking_fields(self):
+        job = AnalysisJob()
+        assert job.document_json is None
+        assert job.chunks_json is None
+
+    def test_mark_completed_with_chunks(self):
+        job = AnalysisJob()
+        job.mark_running()
+        job.mark_completed(
+            markdown="# Title",
+            html="<h1>Title</h1>",
+            pages_json="[]",
+            document_json='{"name": "doc"}',
+            chunks_json='[{"text": "chunk1"}]',
+        )
+        assert job.status == AnalysisStatus.COMPLETED
+        assert job.document_json == '{"name": "doc"}'
+        assert job.chunks_json == '[{"text": "chunk1"}]'
+
+    def test_mark_completed_without_chunks(self):
+        job = AnalysisJob()
+        job.mark_running()
+        job.mark_completed(markdown="md", html="html", pages_json="[]")
+        assert job.document_json is None
+        assert job.chunks_json is None
+
+
+# ---------------------------------------------------------------------------
+# Schemas
+# ---------------------------------------------------------------------------
+
+
+class TestChunkingOptionsRequest:
+    def test_defaults(self):
+        opts = ChunkingOptionsRequest()
+        assert opts.chunker_type == "hybrid"
+        assert opts.max_tokens == 512
+        assert opts.merge_peers is True
+        assert opts.repeat_table_header is True
+
+    def test_custom_values(self):
+        opts = ChunkingOptionsRequest(chunker_type="hierarchical", max_tokens=1024)
+        assert opts.chunker_type == "hierarchical"
+        assert opts.max_tokens == 1024
+
+    def test_invalid_chunker_type(self):
+        with pytest.raises(ValueError, match="chunker_type"):
+            ChunkingOptionsRequest(chunker_type="invalid")
+
+    def test_max_tokens_too_low(self):
+        with pytest.raises(ValueError, match="max_tokens"):
+            ChunkingOptionsRequest(max_tokens=10)
+
+    def test_max_tokens_too_high(self):
+        with pytest.raises(ValueError, match="max_tokens"):
+            ChunkingOptionsRequest(max_tokens=10000)
+
+    def test_boundary_max_tokens(self):
+        opts_low = ChunkingOptionsRequest(max_tokens=64)
+        assert opts_low.max_tokens == 64
+        opts_high = ChunkingOptionsRequest(max_tokens=8192)
+        assert opts_high.max_tokens == 8192
+
+
+class TestChunkResponse:
+    def test_serializes_to_camel_case(self):
+        resp = ChunkResponse(text="hello", headings=["H1"], source_page=1, token_count=5)
+        data = resp.model_dump(by_alias=True)
+        assert "sourcePage" in data
+        assert "tokenCount" in data
+        assert data["text"] == "hello"
+
+
+class TestRechunkRequest:
+    def test_parses(self):
+        req = RechunkRequest(chunkingOptions=ChunkingOptionsRequest(max_tokens=256))
+        assert req.chunkingOptions.max_tokens == 256
+
+
+# ---------------------------------------------------------------------------
+# API endpoints
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture
+def client():
+    return TestClient(app, raise_server_exceptions=False)
+
+
+@pytest.fixture
+def mock_analysis_service(client):
+    mock_svc = MagicMock()
+    original = getattr(app.state, "analysis_service", None)
+    app.state.analysis_service = mock_svc
+    yield mock_svc
+    app.state.analysis_service = original
+
+
+class TestCreateAnalysisWithChunking:
+    def test_create_with_chunking_options(self, client, mock_analysis_service):
+        mock_analysis_service.create = AsyncMock(
+            return_value=AnalysisJob(
+                id="j1",
+                document_id="d1",
+                document_filename="test.pdf",
+            )
+        )
+
+        resp = client.post(
+            "/api/analyses",
+            json={
+                "documentId": "d1",
+                "chunkingOptions": {
+                    "chunker_type": "hybrid",
+                    "max_tokens": 256,
+                    "merge_peers": False,
+                },
+            },
+        )
+        assert resp.status_code == 200
+
+        call_kwargs = mock_analysis_service.create.call_args
+        chunking = call_kwargs.kwargs["chunking_options"]
+        assert chunking["chunker_type"] == "hybrid"
+        assert chunking["max_tokens"] == 256
+        assert chunking["merge_peers"] is False
+
+    def test_create_without_chunking_options(self, client, mock_analysis_service):
+        mock_analysis_service.create = AsyncMock(
+            return_value=AnalysisJob(
+                id="j1",
+                document_id="d1",
+                document_filename="test.pdf",
+            )
+        )
+
+        resp = client.post("/api/analyses", json={"documentId": "d1"})
+        assert resp.status_code == 200
+
+        call_kwargs = mock_analysis_service.create.call_args
+        assert call_kwargs.kwargs["chunking_options"] is None
+
+    def test_response_includes_chunking_fields(self, client, mock_analysis_service):
+        job = AnalysisJob(id="j1", document_id="d1", document_filename="test.pdf")
+        job.mark_running()
+        job.mark_completed(
+            markdown="# Title",
+            html="<h1>Title</h1>",
+            pages_json="[]",
+            document_json='{"name": "doc"}',
+            chunks_json=json.dumps([asdict(ChunkResult(text="chunk1", token_count=5))]),
+        )
+        mock_analysis_service.find_by_id = AsyncMock(return_value=job)
+
+        resp = client.get("/api/analyses/j1")
+        assert resp.status_code == 200
+        data = resp.json()
+        assert data["hasDocumentJson"] is True
+        assert data["chunksJson"] is not None
+        chunks = json.loads(data["chunksJson"])
+        assert len(chunks) == 1
+        assert chunks[0]["text"] == "chunk1"
+
+
+class TestRechunkEndpoint:
+    def test_rechunk_success(self, client, mock_analysis_service):
+        mock_analysis_service.rechunk = AsyncMock(
+            return_value=[
+                ChunkResult(text="chunk1", headings=["H1"], source_page=1, token_count=10),
+                ChunkResult(text="chunk2", headings=["H1", "H2"], source_page=2, token_count=20),
+            ]
+        )
+
+        resp = client.post(
+            "/api/analyses/j1/rechunk",
+            json={
+                "chunkingOptions": {"chunker_type": "hybrid", "max_tokens": 128},
+            },
+        )
+        assert resp.status_code == 200
+        data = resp.json()
+        assert len(data) == 2
+        assert data[0]["text"] == "chunk1"
+        assert data[0]["sourcePage"] == 1
+        assert data[0]["tokenCount"] == 10
+        assert data[1]["headings"] == ["H1", "H2"]
+
+    def test_rechunk_not_completed(self, client, mock_analysis_service):
+        mock_analysis_service.rechunk = AsyncMock(
+            side_effect=ValueError("Analysis is not completed: j1"),
+        )
+
+        resp = client.post(
+            "/api/analyses/j1/rechunk",
+            json={
+                "chunkingOptions": {"chunker_type": "hybrid"},
+            },
+        )
+        assert resp.status_code == 400
+
+    def test_rechunk_no_document_json(self, client, mock_analysis_service):
+        mock_analysis_service.rechunk = AsyncMock(
+            side_effect=ValueError("No document data available for re-chunking: j1"),
+        )
+
+        resp = client.post(
+            "/api/analyses/j1/rechunk",
+            json={
+                "chunkingOptions": {"chunker_type": "hierarchical"},
+            },
+        )
+        assert resp.status_code == 400
+
+    def test_rechunk_invalid_chunker_type(self, client, mock_analysis_service):
+        resp = client.post(
+            "/api/analyses/j1/rechunk",
+            json={
+                "chunkingOptions": {"chunker_type": "invalid"},
+            },
+        )
+        assert resp.status_code == 422

document-parser/tests/test_pipeline_options.py

@@ -23,6 +23,7 @@ from infra.local_converter import (
 # build_converter — verifies Docling pipeline options are wired correctly
 # ---------------------------------------------------------------------------
 
+
 class TestBuildConverter:
     """Verify that build_converter produces a DocumentConverter with the right PdfPipelineOptions."""
 
@@ -101,18 +102,20 @@ class TestBuildConverter:
         assert opts.images_scale == 2.0
 
     def test_all_options_combined(self):
-        conv = build_converter(ConversionOptions(
+        conv = build_converter(
-            do_ocr=False,
+            ConversionOptions(
-            do_table_structure=True,
+                do_ocr=False,
-            table_mode="fast",
+                do_table_structure=True,
-            do_code_enrichment=True,
+                table_mode="fast",
-            do_formula_enrichment=True,
+                do_code_enrichment=True,
-            do_picture_classification=True,
+                do_formula_enrichment=True,
-            do_picture_description=True,
+                do_picture_classification=True,
-            generate_picture_images=True,
+                do_picture_description=True,
-            generate_page_images=True,
+                generate_picture_images=True,
-            images_scale=1.5,
+                generate_page_images=True,
-        ))
+                images_scale=1.5,
+            )
+        )
         opts = self._get_pipeline_options(conv)
         assert opts.do_ocr is False
         assert opts.do_table_structure is True
@@ -130,6 +133,7 @@ class TestBuildConverter:
 # convert_document — default vs custom converter routing
 # ---------------------------------------------------------------------------
 
+
 class TestConvertDocumentRouting:
     """Verify convert_document uses default converter for default opts, custom otherwise."""
 
@@ -142,6 +146,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_get_default.return_value = mock_conv
 
@@ -159,6 +164,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -176,6 +182,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -193,6 +200,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -210,6 +218,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -226,6 +235,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -242,6 +252,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -258,6 +269,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -275,6 +287,7 @@ class TestConvertDocumentRouting:
         mock_result.document.iterate_items.return_value = []
         mock_result.document.export_to_markdown.return_value = ""
         mock_result.document.export_to_html.return_value = ""
+        mock_result.document.export_to_dict.return_value = {}
         mock_conv.convert.return_value = mock_result
         mock_build.return_value = mock_conv
 
@@ -299,30 +312,37 @@ class TestConvertDocumentRouting:
 # Service layer — pipeline options forwarding
 # ---------------------------------------------------------------------------
 
+
 class TestServiceForwardsPipelineOptions:
     """Verify analysis_service.create and _run_analysis forward pipeline options."""
 
     @pytest.fixture
     def mock_doc(self):
         from domain.models import Document
+
         return Document(id="d1", filename="test.pdf", storage_path="/tmp/test.pdf")
 
     @pytest.fixture
     def mock_job(self):
         from domain.models import AnalysisJob
+
         return AnalysisJob(id="j1", document_id="d1", document_filename="test.pdf")
 
     @patch("services.analysis_service.document_repo")
     @patch("services.analysis_service.analysis_repo")
     @pytest.mark.asyncio
     async def test_create_passes_pipeline_options_to_run(
-        self, mock_analysis_repo, mock_doc_repo, mock_doc,
+        self,
+        mock_analysis_repo,
+        mock_doc_repo,
+        mock_doc,
     ):
         mock_doc_repo.find_by_id = AsyncMock(return_value=mock_doc)
         mock_analysis_repo.insert = AsyncMock()
 
         mock_converter = AsyncMock()
         from services.analysis_service import AnalysisService
+
         svc = AnalysisService(converter=mock_converter)
 
         opts = {"do_ocr": False, "table_mode": "fast"}
@@ -335,13 +355,17 @@ class TestServiceForwardsPipelineOptions:
     @patch("services.analysis_service.analysis_repo")
     @pytest.mark.asyncio
     async def test_create_passes_none_when_no_options(
-        self, mock_analysis_repo, mock_doc_repo, mock_doc,
+        self,
+        mock_analysis_repo,
+        mock_doc_repo,
+        mock_doc,
     ):
         mock_doc_repo.find_by_id = AsyncMock(return_value=mock_doc)
         mock_analysis_repo.insert = AsyncMock()
 
         mock_converter = AsyncMock()
         from services.analysis_service import AnalysisService
+
         svc = AnalysisService(converter=mock_converter)
 
         with patch("services.analysis_service.asyncio.create_task") as mock_task:
@@ -352,7 +376,10 @@ class TestServiceForwardsPipelineOptions:
     @patch("services.analysis_service.document_repo")
     @pytest.mark.asyncio
     async def test_run_analysis_forwards_options_to_convert(
-        self, mock_doc_repo, mock_analysis_repo, mock_job,
+        self,
+        mock_doc_repo,
+        mock_analysis_repo,
+        mock_job,
     ):
         from domain.value_objects import ConversionResult, PageDetail
 
@@ -369,6 +396,7 @@ class TestServiceForwardsPipelineOptions:
         )
 
         from services.analysis_service import AnalysisService
+
         svc = AnalysisService(converter=mock_converter)
 
         opts = {
@@ -399,7 +427,10 @@ class TestServiceForwardsPipelineOptions:
     @patch("services.analysis_service.document_repo")
     @pytest.mark.asyncio
     async def test_run_analysis_uses_defaults_when_no_options(
-        self, mock_doc_repo, mock_analysis_repo, mock_job,
+        self,
+        mock_doc_repo,
+        mock_analysis_repo,
+        mock_job,
     ):
         from domain.value_objects import ConversionResult, PageDetail
 
@@ -416,6 +447,7 @@ class TestServiceForwardsPipelineOptions:
         )
 
         from services.analysis_service import AnalysisService
+
         svc = AnalysisService(converter=mock_converter)
 
         await svc._run_analysis("j1", "/tmp/test.pdf", "test.pdf", None)
@@ -429,7 +461,10 @@ class TestServiceForwardsPipelineOptions:
     @patch("services.analysis_service.document_repo")
     @pytest.mark.asyncio
     async def test_run_analysis_marks_failed_on_error(
-        self, mock_doc_repo, mock_analysis_repo, mock_job,
+        self,
+        mock_doc_repo,
+        mock_analysis_repo,
+        mock_job,
     ):
         mock_analysis_repo.find_by_id = AsyncMock(return_value=mock_job)
         mock_analysis_repo.update_status = AsyncMock()
@@ -438,6 +473,7 @@ class TestServiceForwardsPipelineOptions:
         mock_converter.convert.side_effect = RuntimeError("Docling crashed")
 
         from services.analysis_service import AnalysisService
+
         svc = AnalysisService(converter=mock_converter)
 
         await svc._run_analysis("j1", "/tmp/test.pdf", "test.pdf", {"do_ocr": False})
@@ -453,6 +489,7 @@ class TestServiceForwardsPipelineOptions:
 # API endpoint — full request/response with pipeline options
 # ---------------------------------------------------------------------------
 
+
 class TestAnalysisEndpointPipelineOptions:
     """Integration-level tests for the analysis creation endpoint with pipeline options."""
 
@@ -461,6 +498,7 @@ class TestAnalysisEndpointPipelineOptions:
         from fastapi.testclient import TestClient
 
         from main import app
+
         return TestClient(app, raise_server_exceptions=False)
 
     @pytest.fixture
@@ -468,6 +506,7 @@ class TestAnalysisEndpointPipelineOptions:
         from unittest.mock import MagicMock
 
         from main import app
+
         mock = MagicMock()
         original = getattr(app.state, "analysis_service", None)
         app.state.analysis_service = mock
@@ -476,20 +515,25 @@ class TestAnalysisEndpointPipelineOptions:
 
     def test_no_pipeline_options_sends_none(self, client, mock_svc):
         from domain.models import AnalysisJob
+
         mock_svc.create = AsyncMock(return_value=AnalysisJob(id="j1", document_id="d1"))
 
         client.post("/api/analyses", json={"documentId": "d1"})
 
-        mock_svc.create.assert_called_once_with("d1", pipeline_options=None)
+        mock_svc.create.assert_called_once_with("d1", pipeline_options=None, chunking_options=None)
 
     def test_empty_pipeline_options_object_uses_defaults(self, client, mock_svc):
         from domain.models import AnalysisJob
+
         mock_svc.create = AsyncMock(return_value=AnalysisJob(id="j1", document_id="d1"))
 
-        client.post("/api/analyses", json={
+        client.post(
-            "documentId": "d1",
+            "/api/analyses",
-            "pipelineOptions": {},
+            json={
-        })
+                "documentId": "d1",
+                "pipelineOptions": {},
+            },
+        )
 
         opts = mock_svc.create.call_args.kwargs["pipeline_options"]
         assert opts["do_ocr"] is True
@@ -501,12 +545,16 @@ class TestAnalysisEndpointPipelineOptions:
 
     def test_partial_pipeline_options_merges_with_defaults(self, client, mock_svc):
         from domain.models import AnalysisJob
+
         mock_svc.create = AsyncMock(return_value=AnalysisJob(id="j1", document_id="d1"))
 
-        client.post("/api/analyses", json={
+        client.post(
-            "documentId": "d1",
+            "/api/analyses",
-            "pipelineOptions": {"do_ocr": False, "images_scale": 1.5},
+            json={
-        })
+                "documentId": "d1",
+                "pipelineOptions": {"do_ocr": False, "images_scale": 1.5},
+            },
+        )
 
         opts = mock_svc.create.call_args.kwargs["pipeline_options"]
         assert opts["do_ocr"] is False
@@ -522,6 +570,7 @@ class TestAnalysisEndpointPipelineOptions:
 
     def test_full_pipeline_options(self, client, mock_svc):
         from domain.models import AnalysisJob
+
         mock_svc.create = AsyncMock(return_value=AnalysisJob(id="j1", document_id="d1"))
 
         payload = {
@@ -547,18 +596,25 @@ class TestAnalysisEndpointPipelineOptions:
         assert opts == payload["pipelineOptions"]
 
     def test_invalid_pipeline_option_type_rejected(self, client, mock_svc):
-        resp = client.post("/api/analyses", json={
+        resp = client.post(
-            "documentId": "d1",
+            "/api/analyses",
-            "pipelineOptions": {"do_ocr": "not-a-bool"},
+            json={
-        })
+                "documentId": "d1",
+                "pipelineOptions": {"do_ocr": "not-a-bool"},
+            },
+        )
         assert resp.status_code == 422
 
     def test_unknown_pipeline_option_ignored(self, client, mock_svc):
         from domain.models import AnalysisJob
+
         mock_svc.create = AsyncMock(return_value=AnalysisJob(id="j1", document_id="d1"))
 
-        resp = client.post("/api/analyses", json={
+        resp = client.post(
-            "documentId": "d1",
+            "/api/analyses",
-            "pipelineOptions": {"do_ocr": True, "unknown_field": True},
+            json={
-        })
+                "documentId": "d1",
+                "pipelineOptions": {"do_ocr": True, "unknown_field": True},
+            },
+        )
         assert resp.status_code == 200

document-parser/tests/test_repos.py

@@ -1,6 +1,5 @@
 """Tests for persistence repositories using a temporary SQLite database."""
 
-
 import pytest
 
 from domain.models import AnalysisJob, AnalysisStatus, Document

document-parser/tests/test_schemas.py

@@ -1,6 +1,5 @@
 """Tests for API schemas — camelCase serialization and validation."""
 
-
 import pytest
 
 from api.schemas import (
@@ -97,7 +96,10 @@ class TestPipelineOptionsRequest:
 
     def test_custom_values(self):
         opts = PipelineOptionsRequest(
-            do_ocr=False, table_mode="fast", do_code_enrichment=True, images_scale=2.0,
+            do_ocr=False,
+            table_mode="fast",
+            do_code_enrichment=True,
+            images_scale=2.0,
         )
         assert opts.do_ocr is False
         assert opts.table_mode == "fast"

document-parser/tests/test_serve_converter.py

@@ -20,6 +20,7 @@ from infra.serve_converter import (
 # Unit tests — form data building
 # ---------------------------------------------------------------------------
 
+
 class TestBuildFormData:
     def test_default_options(self):
         data = _build_form_data(ConversionOptions())
@@ -36,7 +37,9 @@ class TestBuildFormData:
 
     def test_custom_options(self):
         opts = ConversionOptions(
-            do_ocr=False, table_mode="fast", images_scale=2.0,
+            do_ocr=False,
+            table_mode="fast",
+            images_scale=2.0,
             generate_picture_images=True,
         )
         data = _build_form_data(opts)
@@ -50,6 +53,7 @@ class TestBuildFormData:
 # Unit tests — response parsing
 # ---------------------------------------------------------------------------
 
+
 class TestParseResponse:
     def test_minimal_response(self):
         data = {
@@ -82,12 +86,34 @@ class TestParseResponse:
                         {
                             "label": "title",
                             "text": "Title",
-                            "prov": [{"page_no": 1, "bbox": {"l": 10, "t": 20, "r": 200, "b": 40, "coord_origin": "TOPLEFT"}}],
+                            "prov": [
+                                {
+                                    "page_no": 1,
+                                    "bbox": {
+                                        "l": 10,
+                                        "t": 20,
+                                        "r": 200,
+                                        "b": 40,
+                                        "coord_origin": "TOPLEFT",
+                                    },
+                                }
+                            ],
                         },
                         {
                             "label": "paragraph",
                             "text": "Text",
-                            "prov": [{"page_no": 1, "bbox": {"l": 10, "t": 50, "r": 200, "b": 70, "coord_origin": "TOPLEFT"}}],
+                            "prov": [
+                                {
+                                    "page_no": 1,
+                                    "bbox": {
+                                        "l": 10,
+                                        "t": 50,
+                                        "r": 200,
+                                        "b": 70,
+                                        "coord_origin": "TOPLEFT",
+                                    },
+                                }
+                            ],
                         },
                     ],
                     "tables": [],
@@ -112,7 +138,9 @@ class TestParseResponse:
                         "1": {"size": {"width": 612.0, "height": 792.0}},
                         "2": {"size": {"width": 595.0, "height": 842.0}},
                     },
-                    "texts": [], "tables": [], "pictures": [],
+                    "texts": [],
+                    "tables": [],
+                    "pictures": [],
                 },
             }
         }
@@ -135,7 +163,9 @@ class TestParseResponse:
     def test_json_content_as_string(self):
         json_doc = {
             "pages": {"1": {"size": {"width": 612.0, "height": 792.0}}},
-            "texts": [], "tables": [], "pictures": [],
+            "texts": [],
+            "tables": [],
+            "pictures": [],
         }
         data = {
             "document": {
@@ -171,10 +201,40 @@ class TestParseResponse:
                     "pages": {"1": {"size": {"width": 612.0, "height": 792.0}}},
                     "texts": [],
                     "tables": [
-                        {"label": "table", "text": "", "prov": [{"page_no": 1, "bbox": {"l": 10, "t": 10, "r": 300, "b": 200, "coord_origin": "TOPLEFT"}}]},
+                        {
+                            "label": "table",
+                            "text": "",
+                            "prov": [
+                                {
+                                    "page_no": 1,
+                                    "bbox": {
+                                        "l": 10,
+                                        "t": 10,
+                                        "r": 300,
+                                        "b": 200,
+                                        "coord_origin": "TOPLEFT",
+                                    },
+                                }
+                            ],
+                        },
                     ],
                     "pictures": [
-                        {"label": "picture", "text": "", "prov": [{"page_no": 1, "bbox": {"l": 50, "t": 300, "r": 250, "b": 500, "coord_origin": "TOPLEFT"}}]},
+                        {
+                            "label": "picture",
+                            "text": "",
+                            "prov": [
+                                {
+                                    "page_no": 1,
+                                    "bbox": {
+                                        "l": 50,
+                                        "t": 300,
+                                        "r": 250,
+                                        "b": 500,
+                                        "coord_origin": "TOPLEFT",
+                                    },
+                                }
+                            ],
+                        },
                     ],
                 },
             }
@@ -189,14 +249,19 @@ class TestParseResponse:
 # Unit tests — bbox extraction
 # ---------------------------------------------------------------------------
 
+
 class TestExtractBbox:
     def test_topleft_passthrough(self):
-        bbox = _extract_bbox({"l": 10, "t": 20, "r": 100, "b": 50, "coord_origin": "TOPLEFT"}, 792.0)
+        bbox = _extract_bbox(
+            {"l": 10, "t": 20, "r": 100, "b": 50, "coord_origin": "TOPLEFT"}, 792.0
+        )
         assert bbox == [10, 20, 100, 50]
 
     def test_bottomleft_conversion(self):
         # In BOTTOMLEFT: t (top of box) has higher y than b (bottom of box)
-        bbox = _extract_bbox({"l": 10, "t": 772, "r": 100, "b": 742, "coord_origin": "BOTTOMLEFT"}, 792.0)
+        bbox = _extract_bbox(
+            {"l": 10, "t": 772, "r": 100, "b": 742, "coord_origin": "BOTTOMLEFT"}, 792.0
+        )
         # new_top = 792 - 772 = 20, new_bottom = 792 - 742 = 50
         assert bbox == [10, 20, 100, 50]
 
@@ -217,9 +282,11 @@ class TestExtractBbox:
 # Unit tests — label mapping
 # ---------------------------------------------------------------------------
 
+
 class TestLabelMapping:
     def test_known_labels(self):
         from infra.serve_converter import _LABEL_MAP
+
         assert _LABEL_MAP["table"] == "table"
         assert _LABEL_MAP["picture"] == "picture"
         assert _LABEL_MAP["figure"] == "picture"
@@ -232,6 +299,7 @@ class TestLabelMapping:
 
     def test_unknown_label_defaults_to_text(self):
         from infra.serve_converter import _LABEL_MAP
+
         assert _LABEL_MAP.get("unknown_thing", "text") == "text"
 
 
@@ -239,6 +307,7 @@ class TestLabelMapping:
 # Unit tests — ServeConverter
 # ---------------------------------------------------------------------------
 
+
 class TestServeConverter:
     def test_headers_with_api_key(self):
         conv = ServeConverter(base_url="http://localhost:5001", api_key="secret")
@@ -257,6 +326,7 @@ class TestServeConverter:
 # Integration tests — HTTP calls (mocked)
 # ---------------------------------------------------------------------------
 
+
 class TestServeConverterConvert:
     @pytest.mark.asyncio
     async def test_successful_conversion(self, tmp_path):
@@ -270,7 +340,22 @@ class TestServeConverterConvert:
                 "json_content": {
                     "pages": {"1": {"size": {"width": 612.0, "height": 792.0}}},
                     "texts": [
-                        {"label": "title", "text": "Converted", "prov": [{"page_no": 1, "bbox": {"l": 10, "t": 20, "r": 200, "b": 40, "coord_origin": "TOPLEFT"}}]},
+                        {
+                            "label": "title",
+                            "text": "Converted",
+                            "prov": [
+                                {
+                                    "page_no": 1,
+                                    "bbox": {
+                                        "l": 10,
+                                        "t": 20,
+                                        "r": 200,
+                                        "b": 40,
+                                        "coord_origin": "TOPLEFT",
+                                    },
+                                }
+                            ],
+                        },
                     ],
                     "tables": [],
                     "pictures": [],
@@ -312,7 +397,9 @@ class TestServeConverterConvert:
 
         mock_response = MagicMock()
         mock_response.raise_for_status.side_effect = httpx.HTTPStatusError(
-            "Server Error", request=MagicMock(), response=MagicMock(status_code=500),
+            "Server Error",
+            request=MagicMock(),
+            response=MagicMock(status_code=500),
         )
 
         mock_client = AsyncMock()
@@ -322,8 +409,10 @@ class TestServeConverterConvert:
 
         conv = ServeConverter(base_url="http://localhost:5001")
 
-        with patch("infra.serve_converter.httpx.AsyncClient", return_value=mock_client), \
+        with (
-             pytest.raises(httpx.HTTPStatusError):
+            patch("infra.serve_converter.httpx.AsyncClient", return_value=mock_client),
+            pytest.raises(httpx.HTTPStatusError),
+        ):
             await conv.convert(str(test_file), ConversionOptions())
 
     @pytest.mark.asyncio
@@ -358,6 +447,7 @@ class TestServeConverterConvert:
 # Integration — converter wiring in main.py
 # ---------------------------------------------------------------------------
 
+
 class TestConverterWiring:
     def test_local_engine_builds_local_converter(self):
         from infra.local_converter import LocalConverter
@@ -365,14 +455,19 @@ class TestConverterWiring:
 
         with patch("main.settings", Settings(conversion_engine="local")):
             from main import _build_converter
+
             converter = _build_converter()
         assert isinstance(converter, LocalConverter)
 
     def test_remote_engine_builds_serve_converter(self):
         from infra.settings import Settings
 
-        with patch("main.settings", Settings(conversion_engine="remote", docling_serve_url="http://serve:5001")):
+        with patch(
+            "main.settings",
+            Settings(conversion_engine="remote", docling_serve_url="http://serve:5001"),
+        ):
             from main import _build_converter
+
             converter = _build_converter()
         assert isinstance(converter, ServeConverter)
         assert converter._base_url == "http://serve:5001"
@@ -380,8 +475,16 @@ class TestConverterWiring:
     def test_remote_engine_passes_api_key(self):
         from infra.settings import Settings
 
-        with patch("main.settings", Settings(conversion_engine="remote", docling_serve_url="http://serve:5001", docling_serve_api_key="my-key")):
+        with patch(
+            "main.settings",
+            Settings(
+                conversion_engine="remote",
+                docling_serve_url="http://serve:5001",
+                docling_serve_api_key="my-key",
+            ),
+        ):
             from main import _build_converter
+
             converter = _build_converter()
         assert isinstance(converter, ServeConverter)
         assert converter._api_key == "my-key"

frontend/src/features/analysis/api.ts

@@ -1,20 +1,34 @@
-import type { Analysis, PipelineOptions } from '../../shared/types'
+import type { Analysis, Chunk, ChunkingOptions, PipelineOptions } from '../../shared/types'
 import { apiFetch } from '../../shared/api/http'
 
 export function createAnalysis(
   documentId: string,
   pipelineOptions: PipelineOptions | null = null,
+  chunkingOptions: ChunkingOptions | null = null,
 ): Promise<Analysis> {
   const body: Record<string, unknown> = { documentId }
   if (pipelineOptions) {
     body.pipelineOptions = pipelineOptions
   }
+  if (chunkingOptions) {
+    body.chunkingOptions = chunkingOptions
+  }
   return apiFetch<Analysis>('/api/analyses', {
     method: 'POST',
     body: JSON.stringify(body),
   })
 }
 
+export function rechunkAnalysis(
+  jobId: string,
+  chunkingOptions: ChunkingOptions,
+): Promise<Chunk[]> {
+  return apiFetch<Chunk[]>(`/api/analyses/${jobId}/rechunk`, {
+    method: 'POST',
+    body: JSON.stringify({ chunkingOptions }),
+  })
+}
+
 export function fetchAnalyses(): Promise<Analysis[]> {
   return apiFetch<Analysis[]>('/api/analyses')
 }

frontend/src/features/analysis/pipelineOptions.test.ts

@@ -131,7 +131,7 @@ describe('useAnalysisStore — pipeline options forwarding', () => {
     const store = useAnalysisStore()
     await store.run('d1')
 
-    expect(api.createAnalysis).toHaveBeenCalledWith('d1', null)
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', null, null)
     store.stopPolling()
   })
 
@@ -155,7 +155,7 @@ describe('useAnalysisStore — pipeline options forwarding', () => {
     }
     await store.run('d1', opts)
 
-    expect(api.createAnalysis).toHaveBeenCalledWith('d1', opts)
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', opts, null)
     store.stopPolling()
   })
 
@@ -168,7 +168,7 @@ describe('useAnalysisStore — pipeline options forwarding', () => {
     const opts = { do_ocr: false }
     await store.run('d1', opts)
 
-    expect(api.createAnalysis).toHaveBeenCalledWith('d1', { do_ocr: false })
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', { do_ocr: false }, null)
     store.stopPolling()
   })
 

frontend/src/features/analysis/store.test.ts

@@ -70,7 +70,7 @@ describe('useAnalysisStore', () => {
     expect(store.currentAnalysis).toEqual(job)
     expect(store.analyses[0]).toEqual(job)
     expect(store.running).toBe(true)
-    expect(api.createAnalysis).toHaveBeenCalledWith('d1', null)
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', null, null)
 
     // Advance timer to trigger polling
     await vi.advanceTimersByTimeAsync(2000)
@@ -90,7 +90,7 @@ describe('useAnalysisStore', () => {
     const options = { do_ocr: false, table_mode: 'fast' }
     await store.run('d1', options)
 
-    expect(api.createAnalysis).toHaveBeenCalledWith('d1', options)
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', options, null)
 
     store.stopPolling()
   })

frontend/src/features/analysis/store.ts

@@ -1,6 +1,6 @@
 import { defineStore } from 'pinia'
 import { ref, computed } from 'vue'
-import type { Analysis, Page, PipelineOptions } from '../../shared/types'
+import type { Analysis, Chunk, ChunkingOptions, Page, PipelineOptions } from '../../shared/types'
 import * as api from './api'
 
 export const useAnalysisStore = defineStore('analysis', () => {
@@ -35,14 +35,44 @@ export const useAnalysisStore = defineStore('analysis', () => {
     }
   }
 
+  const currentChunks = computed<Chunk[]>(() => {
+    if (!currentAnalysis.value?.chunksJson) return []
+    try {
+      return JSON.parse(currentAnalysis.value.chunksJson) as Chunk[]
+    } catch {
+      return []
+    }
+  })
+
+  const rechunking = ref(false)
+
+  async function rechunk(jobId: string, chunkingOptions: ChunkingOptions): Promise<Chunk[]> {
+    rechunking.value = true
+    error.value = null
+    try {
+      const chunks = await api.rechunkAnalysis(jobId, chunkingOptions)
+      if (currentAnalysis.value?.id === jobId) {
+        currentAnalysis.value = await api.fetchAnalysis(jobId)
+      }
+      return chunks
+    } catch (e) {
+      error.value = (e as Error).message || 'Failed to rechunk'
+      console.error('Failed to rechunk', e)
+      throw e
+    } finally {
+      rechunking.value = false
+    }
+  }
+
   async function run(
     documentId: string,
     pipelineOptions: PipelineOptions | null = null,
+    chunkingOptions: ChunkingOptions | null = null,
   ): Promise<Analysis> {
     running.value = true
     error.value = null
     try {
-      const analysis = await api.createAnalysis(documentId, pipelineOptions)
+      const analysis = await api.createAnalysis(documentId, pipelineOptions, chunkingOptions)
       currentAnalysis.value = analysis
       analyses.value.unshift(analysis)
       startPolling(analysis.id)
@@ -118,11 +148,14 @@ export const useAnalysisStore = defineStore('analysis', () => {
     analyses,
     currentAnalysis,
     currentPages,
+    currentChunks,
     running,
+    rechunking,
     error,
     clearError,
     load,
     run,
+    rechunk,
     select,
     remove,
     stopPolling,

frontend/src/features/chunking/api.test.ts

@@ -0,0 +1,52 @@
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { rechunkAnalysis, createAnalysis } from '../analysis/api'
+
+vi.mock('../../shared/api/http', () => ({
+  apiFetch: vi.fn(),
+}))
+
+import { apiFetch } from '../../shared/api/http'
+
+describe('chunking API', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+  })
+
+  it('createAnalysis sends chunkingOptions when provided', async () => {
+    const job = { id: '1', documentId: 'doc-1', status: 'PENDING' }
+    apiFetch.mockResolvedValue(job)
+
+    const chunkingOpts = { chunker_type: 'hybrid' as const, max_tokens: 256 }
+    await createAnalysis('doc-1', null, chunkingOpts)
+
+    expect(apiFetch).toHaveBeenCalledWith('/api/analyses', {
+      method: 'POST',
+      body: JSON.stringify({ documentId: 'doc-1', chunkingOptions: chunkingOpts }),
+    })
+  })
+
+  it('createAnalysis omits chunkingOptions when null', async () => {
+    apiFetch.mockResolvedValue({ id: '1' })
+
+    await createAnalysis('doc-1', null, null)
+
+    expect(apiFetch).toHaveBeenCalledWith('/api/analyses', {
+      method: 'POST',
+      body: JSON.stringify({ documentId: 'doc-1' }),
+    })
+  })
+
+  it('rechunkAnalysis sends POST to rechunk endpoint', async () => {
+    const chunks = [{ text: 'chunk1', headings: [], sourcePage: 1, tokenCount: 10 }]
+    apiFetch.mockResolvedValue(chunks)
+
+    const opts = { chunker_type: 'hybrid' as const, max_tokens: 512 }
+    const result = await rechunkAnalysis('job-1', opts)
+
+    expect(apiFetch).toHaveBeenCalledWith('/api/analyses/job-1/rechunk', {
+      method: 'POST',
+      body: JSON.stringify({ chunkingOptions: opts }),
+    })
+    expect(result).toEqual(chunks)
+  })
+})

frontend/src/features/chunking/index.ts

@@ -0,0 +1 @@
+export { default as ChunkPanel } from './ui/ChunkPanel.vue'

frontend/src/features/chunking/store.test.ts

@@ -0,0 +1,135 @@
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { setActivePinia, createPinia } from 'pinia'
+import { useAnalysisStore } from '../analysis/store'
+
+vi.mock('../analysis/api', () => ({
+  createAnalysis: vi.fn(),
+  fetchAnalyses: vi.fn().mockResolvedValue([]),
+  fetchAnalysis: vi.fn(),
+  deleteAnalysis: vi.fn(),
+  rechunkAnalysis: vi.fn(),
+}))
+
+import * as api from '../analysis/api'
+
+describe('analysis store — chunking', () => {
+  beforeEach(() => {
+    setActivePinia(createPinia())
+    vi.clearAllMocks()
+  })
+
+  it('currentChunks parses chunksJson from current analysis', () => {
+    const store = useAnalysisStore()
+    const chunks = [
+      { text: 'chunk1', headings: ['H1'], sourcePage: 1, tokenCount: 10 },
+      { text: 'chunk2', headings: [], sourcePage: 2, tokenCount: 20 },
+    ]
+    store.currentAnalysis = {
+      id: 'j1',
+      documentId: 'd1',
+      documentFilename: null,
+      status: 'COMPLETED',
+      contentMarkdown: null,
+      contentHtml: null,
+      pagesJson: null,
+      chunksJson: JSON.stringify(chunks),
+      hasDocumentJson: true,
+      errorMessage: null,
+      startedAt: null,
+      completedAt: null,
+      createdAt: '2024-01-01',
+    }
+    expect(store.currentChunks).toEqual(chunks)
+  })
+
+  it('currentChunks returns empty array when no chunksJson', () => {
+    const store = useAnalysisStore()
+    store.currentAnalysis = {
+      id: 'j1',
+      documentId: 'd1',
+      documentFilename: null,
+      status: 'COMPLETED',
+      contentMarkdown: null,
+      contentHtml: null,
+      pagesJson: null,
+      chunksJson: null,
+      hasDocumentJson: false,
+      errorMessage: null,
+      startedAt: null,
+      completedAt: null,
+      createdAt: '2024-01-01',
+    }
+    expect(store.currentChunks).toEqual([])
+  })
+
+  it('rechunk calls API and refreshes analysis', async () => {
+    const store = useAnalysisStore()
+    const chunks = [{ text: 'c1', headings: [], sourcePage: 1, tokenCount: 5 }]
+    vi.mocked(api.rechunkAnalysis).mockResolvedValue(chunks)
+    vi.mocked(api.fetchAnalysis).mockResolvedValue({
+      id: 'j1',
+      documentId: 'd1',
+      documentFilename: null,
+      status: 'COMPLETED',
+      contentMarkdown: null,
+      contentHtml: null,
+      pagesJson: null,
+      chunksJson: JSON.stringify(chunks),
+      hasDocumentJson: true,
+      errorMessage: null,
+      startedAt: null,
+      completedAt: null,
+      createdAt: '2024-01-01',
+    })
+
+    store.currentAnalysis = {
+      id: 'j1',
+      documentId: 'd1',
+      documentFilename: null,
+      status: 'COMPLETED',
+      contentMarkdown: null,
+      contentHtml: null,
+      pagesJson: null,
+      chunksJson: null,
+      hasDocumentJson: true,
+      errorMessage: null,
+      startedAt: null,
+      completedAt: null,
+      createdAt: '2024-01-01',
+    }
+
+    const result = await store.rechunk('j1', { chunker_type: 'hybrid', max_tokens: 256 })
+
+    expect(api.rechunkAnalysis).toHaveBeenCalledWith('j1', {
+      chunker_type: 'hybrid',
+      max_tokens: 256,
+    })
+    expect(result).toEqual(chunks)
+    expect(store.rechunking).toBe(false)
+  })
+
+  it('run passes chunkingOptions to API', async () => {
+    const store = useAnalysisStore()
+    vi.mocked(api.createAnalysis).mockResolvedValue({
+      id: 'j1',
+      documentId: 'd1',
+      documentFilename: null,
+      status: 'PENDING',
+      contentMarkdown: null,
+      contentHtml: null,
+      pagesJson: null,
+      chunksJson: null,
+      hasDocumentJson: false,
+      errorMessage: null,
+      startedAt: null,
+      completedAt: null,
+      createdAt: '2024-01-01',
+    })
+
+    await store.run('d1', null, { chunker_type: 'hierarchical' })
+
+    expect(api.createAnalysis).toHaveBeenCalledWith('d1', null, {
+      chunker_type: 'hierarchical',
+    })
+  })
+})

frontend/src/features/chunking/ui/ChunkPanel.vue

@@ -0,0 +1,336 @@
+<template>
+  <div class="chunk-panel">
+    <!-- Chunking config -->
+    <div class="chunk-config">
+      <div class="config-section">
+        <label class="config-label">{{ t('chunking.settings') }}</label>
+
+        <div class="config-row">
+          <label class="config-label-sm">{{ t('chunking.chunkerType') }}</label>
+          <select class="config-select" v-model="options.chunker_type">
+            <option value="hybrid">Hybrid</option>
+            <option value="hierarchical">Hierarchical</option>
+          </select>
+        </div>
+
+        <div class="config-row">
+          <label class="config-label-sm">{{ t('chunking.maxTokens') }}</label>
+          <input
+            type="number"
+            class="config-input"
+            v-model.number="options.max_tokens"
+            min="64"
+            max="8192"
+            step="64"
+          />
+        </div>
+
+        <div class="config-toggle-row" v-if="options.chunker_type === 'hybrid'">
+          <label class="toggle-label">
+            <input type="checkbox" v-model="options.merge_peers" class="toggle-input" />
+            <span class="toggle-switch" />
+            <span class="toggle-text">{{ t('chunking.mergePeers') }}</span>
+          </label>
+        </div>
+
+        <div class="config-toggle-row" v-if="options.chunker_type === 'hybrid'">
+          <label class="toggle-label">
+            <input type="checkbox" v-model="options.repeat_table_header" class="toggle-input" />
+            <span class="toggle-switch" />
+            <span class="toggle-text">{{ t('chunking.repeatTableHeader') }}</span>
+          </label>
+        </div>
+      </div>
+
+      <button
+        class="chunk-btn primary"
+        :disabled="!canRechunk || analysisStore.rechunking"
+        @click="doRechunk"
+      >
+        <div v-if="analysisStore.rechunking" class="spinner-sm" />
+        {{ analysisStore.rechunking ? t('chunking.chunking') : t('chunking.run') }}
+      </button>
+    </div>
+
+    <!-- Chunks list -->
+    <div class="chunk-results" v-if="analysisStore.currentChunks.length">
+      <div class="chunk-summary">
+        {{ analysisStore.currentChunks.length }} {{ t('chunking.chunks') }}
+      </div>
+      <div class="chunk-list">
+        <div
+          class="chunk-card"
+          v-for="(chunk, idx) in analysisStore.currentChunks"
+          :key="idx"
+        >
+          <div class="chunk-header">
+            <span class="chunk-index">#{{ idx + 1 }}</span>
+            <span class="chunk-tokens" v-if="chunk.tokenCount">
+              {{ chunk.tokenCount }} tokens
+            </span>
+            <span class="chunk-page" v-if="chunk.sourcePage">
+              p.{{ chunk.sourcePage }}
+            </span>
+          </div>
+          <div class="chunk-headings" v-if="chunk.headings.length">
+            <span class="chunk-heading" v-for="h in chunk.headings" :key="h">{{ h }}</span>
+          </div>
+          <div class="chunk-text">{{ chunk.text }}</div>
+        </div>
+      </div>
+    </div>
+
+    <div class="chunk-empty" v-else-if="!analysisStore.rechunking">
+      <p>{{ t('chunking.noChunks') }}</p>
+    </div>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { reactive, computed } from 'vue'
+import { useAnalysisStore } from '../../analysis/store'
+import { useI18n } from '../../../shared/i18n'
+import type { ChunkingOptions } from '../../../shared/types'
+
+const analysisStore = useAnalysisStore()
+const { t } = useI18n()
+
+const options = reactive<Required<ChunkingOptions>>({
+  chunker_type: 'hybrid',
+  max_tokens: 512,
+  merge_peers: true,
+  repeat_table_header: true,
+})
+
+const canRechunk = computed(() => {
+  const analysis = analysisStore.currentAnalysis
+  return analysis?.status === 'COMPLETED' && analysis.hasDocumentJson
+})
+
+async function doRechunk() {
+  if (!analysisStore.currentAnalysis) return
+  await analysisStore.rechunk(analysisStore.currentAnalysis.id, { ...options })
+}
+</script>
+
+<style scoped>
+.chunk-panel {
+  display: flex;
+  flex-direction: column;
+  height: 100%;
+  overflow: hidden;
+}
+
+.chunk-config {
+  padding: 16px;
+  border-bottom: 1px solid var(--border);
+  display: flex;
+  flex-direction: column;
+  gap: 12px;
+}
+
+.config-section {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+}
+
+.config-label {
+  font-size: 11px;
+  font-weight: 600;
+  text-transform: uppercase;
+  letter-spacing: 0.5px;
+  color: var(--text-secondary);
+}
+
+.config-label-sm {
+  font-size: 12px;
+  color: var(--text-secondary);
+}
+
+.config-row {
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+}
+
+.config-select,
+.config-input {
+  background: var(--bg-secondary);
+  border: 1px solid var(--border);
+  border-radius: var(--radius);
+  padding: 6px 10px;
+  font-size: 13px;
+  color: var(--text);
+  width: 100%;
+}
+
+.config-toggle-row {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.toggle-label {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+  cursor: pointer;
+  font-size: 13px;
+  color: var(--text);
+}
+
+.toggle-input {
+  display: none;
+}
+
+.toggle-switch {
+  width: 32px;
+  height: 18px;
+  background: var(--bg-tertiary);
+  border-radius: 9px;
+  position: relative;
+  transition: background 0.2s;
+  flex-shrink: 0;
+}
+
+.toggle-switch::after {
+  content: '';
+  position: absolute;
+  top: 2px;
+  left: 2px;
+  width: 14px;
+  height: 14px;
+  background: white;
+  border-radius: 50%;
+  transition: transform 0.2s;
+}
+
+.toggle-input:checked + .toggle-switch {
+  background: var(--accent);
+}
+
+.toggle-input:checked + .toggle-switch::after {
+  transform: translateX(14px);
+}
+
+.chunk-btn {
+  padding: 8px 16px;
+  border: none;
+  border-radius: var(--radius);
+  cursor: pointer;
+  font-size: 13px;
+  font-weight: 500;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  gap: 6px;
+}
+
+.chunk-btn.primary {
+  background: var(--accent);
+  color: white;
+}
+
+.chunk-btn.primary:disabled {
+  opacity: 0.5;
+  cursor: not-allowed;
+}
+
+.chunk-results {
+  flex: 1;
+  overflow-y: auto;
+  padding: 12px;
+}
+
+.chunk-summary {
+  font-size: 12px;
+  font-weight: 600;
+  color: var(--text-secondary);
+  text-transform: uppercase;
+  letter-spacing: 0.5px;
+  margin-bottom: 8px;
+}
+
+.chunk-list {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+}
+
+.chunk-card {
+  background: var(--bg-secondary);
+  border: 1px solid var(--border);
+  border-radius: var(--radius);
+  padding: 10px 12px;
+}
+
+.chunk-header {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+  margin-bottom: 6px;
+}
+
+.chunk-index {
+  font-size: 11px;
+  font-weight: 700;
+  color: var(--accent);
+}
+
+.chunk-tokens,
+.chunk-page {
+  font-size: 11px;
+  color: var(--text-secondary);
+  background: var(--bg-tertiary);
+  padding: 1px 6px;
+  border-radius: 4px;
+}
+
+.chunk-headings {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 4px;
+  margin-bottom: 6px;
+}
+
+.chunk-heading {
+  font-size: 11px;
+  color: var(--accent);
+  background: var(--accent-bg, rgba(99, 102, 241, 0.1));
+  padding: 1px 6px;
+  border-radius: 4px;
+}
+
+.chunk-text {
+  font-size: 12px;
+  color: var(--text);
+  line-height: 1.5;
+  white-space: pre-wrap;
+  word-break: break-word;
+  max-height: 120px;
+  overflow-y: auto;
+}
+
+.chunk-empty {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  flex: 1;
+  color: var(--text-secondary);
+  font-size: 13px;
+}
+
+.spinner-sm {
+  width: 14px;
+  height: 14px;
+  border: 2px solid rgba(255, 255, 255, 0.3);
+  border-top-color: white;
+  border-radius: 50%;
+  animation: spin 0.6s linear infinite;
+}
+
+@keyframes spin {
+  to { transform: rotate(360deg); }
+}
+</style>

frontend/src/pages/StudioPage.vue

@@ -37,6 +37,15 @@
           >
             {{ t('studio.verify') }}
           </button>
+          <button
+            v-if="chunkingEnabled"
+            class="toggle-btn"
+            :class="{ active: mode === 'preparer' }"
+            @click="mode = 'preparer'"
+            :disabled="!analysisStore.currentAnalysis"
+          >
+            {{ t('studio.prepare') }}
+          </button>
         </div>
       </div>
       <div class="topbar-actions">
@@ -280,6 +289,11 @@
             @highlight-element="highlightedElementIndex = $event"
           />
         </div>
+
+        <!-- PREPARER MODE (feature-flipped) -->
+        <div v-if="mode === 'preparer' && chunkingEnabled" class="prepare-panel">
+          <ChunkPanel />
+        </div>
       </div>
     </div>
   </div>
@@ -293,6 +307,8 @@ import { useAnalysisStore } from '../features/analysis/store'
 import { DocumentUpload, DocumentList } from '../features/document/index'
 import { ResultTabs } from '../features/analysis/index'
 import BboxOverlay from '../features/analysis/ui/BboxOverlay.vue'
+import { ChunkPanel } from '../features/chunking'
+import { useFeatureFlag } from '../features/feature-flags'
 import { getPreviewUrl } from '../features/document/api'
 import { useI18n } from '../shared/i18n'
 import type { PipelineOptions } from '../shared/types'
@@ -302,6 +318,7 @@ const router = useRouter()
 const documentStore = useDocumentStore()
 const analysisStore = useAnalysisStore()
 const { t } = useI18n()
+const chunkingEnabled = useFeatureFlag('chunking')
 
 const mode = ref('configurer')
 const currentPage = ref(1)

frontend/src/shared/i18n.ts

@@ -95,6 +95,18 @@ const messages: Messages = {
     'history.emptyDocs': 'Aucun document. Importez un document depuis le Studio.',
     'history.open': 'Ouvrir',
 
+    // Chunking
+    'studio.prepare': 'Préparer',
+    'chunking.settings': 'Chunking',
+    'chunking.chunkerType': 'Type de chunker',
+    'chunking.maxTokens': 'Tokens max',
+    'chunking.mergePeers': 'Fusionner les pairs',
+    'chunking.repeatTableHeader': 'Répéter en-têtes tableaux',
+    'chunking.run': 'Chunker',
+    'chunking.chunking': 'Chunking...',
+    'chunking.chunks': 'chunks',
+    'chunking.noChunks': 'Lancez le chunking pour préparer les segments.',
+
     // Settings
     'settings.title': 'Paramètres',
     'settings.apiUrl': 'API URL',
@@ -185,6 +197,17 @@ const messages: Messages = {
     'history.emptyDocs': 'No documents yet. Upload a document from the Studio.',
     'history.open': 'Open',
 
+    'studio.prepare': 'Prepare',
+    'chunking.settings': 'Chunking',
+    'chunking.chunkerType': 'Chunker type',
+    'chunking.maxTokens': 'Max tokens',
+    'chunking.mergePeers': 'Merge peers',
+    'chunking.repeatTableHeader': 'Repeat table headers',
+    'chunking.run': 'Chunk',
+    'chunking.chunking': 'Chunking...',
+    'chunking.chunks': 'chunks',
+    'chunking.noChunks': 'Run chunking to prepare segments.',
+
     'settings.title': 'Settings',
     'settings.apiUrl': 'API URL',
     'settings.version': 'Version',

frontend/src/shared/types.ts

@@ -30,12 +30,28 @@ export interface Analysis {
   contentMarkdown: string | null
   contentHtml: string | null
   pagesJson: string | null
+  chunksJson: string | null
+  hasDocumentJson: boolean
   errorMessage: string | null
   startedAt: string | null
   completedAt: string | null
   createdAt: string
 }
 
+export interface ChunkingOptions {
+  chunker_type?: 'hybrid' | 'hierarchical'
+  max_tokens?: number
+  merge_peers?: boolean
+  repeat_table_header?: boolean
+}
+
+export interface Chunk {
+  text: string
+  headings: string[]
+  sourcePage: number | null
+  tokenCount: number
+}
+
 export interface PageElement {
   type: string
   bbox: [number, number, number, number]