feat(server): add workbench API, DSH planner worker and RAG pipeline

server/app/rag/context.py

@@ -0,0 +1,41 @@
+"""Bounded source excerpts with relevant middle/tail evidence and honest omissions."""
+import re
+
+
+def source_excerpt(text: str, query: str = "", limit: int = 2400) -> str:
+    text = str(text or "")
+    if len(text) <= limit:
+        return text
+    marker = "\n[…此处省略，未展示内容不能作为已知事实…]\n"
+    # Keep boundary conditions and conclusion as well as query-related evidence.
+    edge = min(350, limit // 5)
+    spans = [(0, edge), (len(text) - edge, len(text))]
+    terms = re.findall(r"[A-Za-z0-9_]+|[\u3400-\u9fff]{2,}", str(query).lower())
+    terms += [term[i:i + 2] for term in list(terms) if len(term) > 2
+              for i in range(len(term) - 1)]
+    lower = text.lower()
+    candidates = []
+    for term in dict.fromkeys(terms):
+        position = lower.find(term)
+        if position >= edge and position < len(text) - edge:
+            candidates.append((position, len(term)))
+    candidates.sort(key=lambda pair: (-pair[1], pair[0]))
+    remaining = max(0, limit - 2 * edge - 3 * len(marker))
+    for position, _ in candidates:
+        if remaining < 120:
+            break
+        if any(start <= position < end for start, end in spans):
+            continue
+        size = min(remaining, 900)
+        start = max(0, position - size // 3)
+        spans.append((start, min(len(text), start + size)))
+        remaining -= size + len(marker)
+    if len(spans) == 2 and remaining:
+        spans.append((edge, min(len(text) - edge, edge + remaining)))
+    merged = []
+    for start, end in sorted(spans):
+        if merged and start <= merged[-1][1]:
+            merged[-1] = (merged[-1][0], max(end, merged[-1][1]))
+        else:
+            merged.append((start, end))
+    return marker.join(text[start:end] for start, end in merged)