feat: add full text segment id support;

dce41c59 · ccran · e047146d · dce41c59 · dce41c59 · dce41c59
Commit dce41c59 authored Jun 24, 2026 by ccran
Show whitespace changes
Inline Side-by-side

Showing with 66 additions and 29 deletions

core/config.py
+1 -0

main.py
+39 -20

utils/spire_word_util.py
+26 -9

No files found.
--- a/core/config.py
+++ b/core/config.py
@@ -35,6 +35,7 @@ MAX_SINGLE_CHUNK_SIZE = 5000
 MERGE_RULE_PROMPT = False
 META_KEY = "META"
 DEFAULT_RULESET_ID = "通用"
+FULL_TEXT_SEGMENT_ID = -1
 ## 规则集ID列表，需与rules.xlsx中的sheet名称保持一致!!!
 ALL_RULESET_IDS = [
    "通用",

--- a/main.py
+++ b/main.py
@@ -18,6 +18,7 @@ from core.config import (
    doc_support_formats,
    pdf_support_formats,
    MERGE_RULE_PROMPT,
+    FULL_TEXT_SEGMENT_ID,
    use_lufa,
    max_model_len,
    debug_mode,
@@ -266,11 +267,18 @@ def summarize_facts(payload: SegmentSummaryRequest) -> SegmentSummaryResponse:
 class SegmentReviewRequest(BaseModel):
    conversation_id: str
-    segment_id: int
+    segment_id: int = Field(
+        default=FULL_TEXT_SEGMENT_ID,
+        description="分段ID；为 FULL_TEXT_SEGMENT_ID 时使用 segment_text 审查全文",
+    )
+    segment_text: Optional[str] = Field(
+        default=None,
+        description="segment_id 为 FULL_TEXT_SEGMENT_ID 时用于审查的全文文本",
+    )
    party_role: Optional[str] = ""
    ruleset_id: Optional[str] = "通用"
    routed_rule_titles: Optional[List[str]] = None
-    file_ext: str
+    file_ext: Optional[str] = None
    context_memories: Optional[List[Dict]] = None
    route_by: Literal["rule", "summary"] = Field(
        default="rule", description="路由依据：rule=审查规则项，summary=摘要项"
@@ -294,9 +302,27 @@ class SegmentRuleRouterResponse(BaseModel):
    routed_rules: List[Dict]
-@app.post("/segments/review/findings", response_model=SegmentReviewResponse)
+def _resolve_review_segment(payload: SegmentReviewRequest):
-def review_segment(payload: SegmentReviewRequest) -> SegmentReviewResponse:
+    if payload.segment_id == FULL_TEXT_SEGMENT_ID:
-    store = get_cached_memory(payload.conversation_id)
+        segment_text = (payload.segment_text or "").strip()
+        if not segment_text:
+            raise HTTPException(
+                status_code=400,
+                detail=f"segment_text cannot be empty when segment_id is {FULL_TEXT_SEGMENT_ID}",
+            )
+        return FULL_TEXT_SEGMENT_ID, segment_text, None
+    if payload.segment_id < FULL_TEXT_SEGMENT_ID:
+        raise HTTPException(
+            status_code=400,
+            detail=f"segment_id must be {FULL_TEXT_SEGMENT_ID} or greater",
+        )
+    if not payload.file_ext:
+        raise HTTPException(
+            status_code=400,
+            detail=f"file_ext is required when segment_id is not {FULL_TEXT_SEGMENT_ID}",
+        )
    try:
        doc_obj, _ = get_cached_doc_tool(payload.conversation_id, payload.file_ext)
    except Exception as exc:
@@ -312,6 +338,13 @@ def review_segment(payload: SegmentReviewRequest) -> SegmentReviewResponse:
            status_code=404,
            detail=f"Segment text not found for id {payload.segment_id}: {exc}. Please parse document first.",
        )
+    return segment_idx, segment_text, doc_obj
+@app.post("/segments/review/findings", response_model=SegmentReviewResponse)
+def review_segment(payload: SegmentReviewRequest) -> SegmentReviewResponse:
+    store = get_cached_memory(payload.conversation_id)
+    segment_idx, segment_text, _ = _resolve_review_segment(payload)
    ruleset_id = payload.ruleset_id or rules_reference_tool.default_ruleset_id
    rules = rules_reference_tool.run(
@@ -366,21 +399,7 @@ def review_segment(payload: SegmentReviewRequest) -> SegmentReviewResponse:
 @app.post("/segments/review/rule-router", response_model=SegmentRuleRouterResponse)
 def route_segment_rules(payload: SegmentReviewRequest) -> SegmentRuleRouterResponse:
-    try:
+    segment_idx, segment_text, doc_obj = _resolve_review_segment(payload)
-        doc_obj, _ = get_cached_doc_tool(payload.conversation_id, payload.file_ext)
-    except Exception as exc:
-        raise HTTPException(
-            status_code=400, detail=f"Document tool not available: {exc}"
-        )
-    segment_idx = payload.segment_id - 1
-    try:
-        segment_text = doc_obj.get_chunk_item(segment_idx)
-    except Exception as exc:
-        raise HTTPException(
-            status_code=404,
-            detail=f"Segment text not found for id {payload.segment_id}: {exc}. Please parse document first.",
-        )
    ruleset_id = payload.ruleset_id or rules_reference_tool.default_ruleset_id
    rules = rules_reference_tool.run(ruleset_id=ruleset_id).get("rules", [])

--- a/utils/spire_word_util.py
+++ b/utils/spire_word_util.py
@@ -4,7 +4,7 @@ import re
 from thefuzz import fuzz
 from utils.doc_util import DocBase
 from utils.common_util import adjust_single_chunk_size
-from core.config import use_lufa
+from core.config import FULL_TEXT_SEGMENT_ID, use_lufa
 import os
@@ -553,6 +553,22 @@ class SpireWordDoc(DocBase):
            for loc in chunk_locations
        ]
+    def get_all_sub_chunks(self):
+        self._ensure_loaded()
+        sub_chunks = []
+        for chunk_id in range(len(self._chunk_list)):
+            sub_chunks.extend(self.get_sub_chunks(chunk_id))
+        return sub_chunks
+    def _get_comment_sub_chunks(self, chunk_id):
+        self._ensure_loaded()
+        if chunk_id == FULL_TEXT_SEGMENT_ID:
+            return self.get_all_sub_chunks()
+        if isinstance(chunk_id, int) and 0 <= chunk_id < self.get_chunk_num():
+            return self.get_sub_chunks(chunk_id)
+        logger.error(f"invalid chunk_id for comment: {chunk_id}")
+        return []
    def format_comment_author(self, comment):
        return "{}|{}".format(str(comment["id"]), comment["key_points"])
@@ -820,14 +836,15 @@ class SpireWordDoc(DocBase):
        for comment in comments:
            if comment.get("result") != "不合格":
                continue
-            # update chunk_id
+            comment_chunk_id = comment.get("chunk_id")
-            comment_chunk_id = comment.get("chunk_id", -1)
+            # 优先使用 comment 中的 chunk_id；-1 表示全文；缺失或无效时回退到入参 chunk_id。
-            # 优先使用comments里提供的chunk_id，如果没有或无效则使用外部传入的chunk_id，如果都没有则异常处理
+            if isinstance(comment_chunk_id, int) and (
-            sub_chunks = (
+                comment_chunk_id == FULL_TEXT_SEGMENT_ID
-                self.get_sub_chunks(comment_chunk_id)
+                or 0 <= comment_chunk_id < self.get_chunk_num()
-                if comment_chunk_id != -1 and comment_chunk_id < self.get_chunk_num()
+            ):
-                else self.get_sub_chunks(chunk_id)
+                sub_chunks = self._get_comment_sub_chunks(comment_chunk_id)
-            )
+            else:
+                sub_chunks = self._get_comment_sub_chunks(chunk_id)
            author = self.format_comment_author(comment)
            suggest = comment.get("suggest", "")
            original_text = (comment.get("original_text") or "").strip()