deepresearch-flow 0.5.1__py3-none-any.whl → 0.6.0__py3-none-any.whl

deepresearch_flow/paper/snapshot/text.py

@@ -0,0 +1,154 @@
+from __future__ import annotations
+
+import re
+from typing import Iterable
+
+from markdown_it import MarkdownIt
+
+
+_HTML_TABLE_RE = re.compile(r"<table\b.*?</table>", re.IGNORECASE | re.DOTALL)
+_TAG_RE = re.compile(r"<[^>]+>")
+_WS_RE = re.compile(r"\s+")
+
+
+def _is_cjk_char(ch: str) -> bool:
+    code = ord(ch)
+    return (
+        0x3400 <= code <= 0x4DBF  # CJK Unified Ideographs Extension A
+        or 0x4E00 <= code <= 0x9FFF  # CJK Unified Ideographs
+        or 0xF900 <= code <= 0xFAFF  # CJK Compatibility Ideographs
+        or 0x3040 <= code <= 0x309F  # Hiragana
+        or 0x30A0 <= code <= 0x30FF  # Katakana
+        or 0xAC00 <= code <= 0xD7AF  # Hangul syllables
+    )
+
+
+def insert_cjk_spaces(text: str) -> str:
+    out: list[str] = []
+    prev_cjk = False
+    for ch in text:
+        cur_cjk = _is_cjk_char(ch)
+        if prev_cjk and cur_cjk:
+            out.append(" ")
+        out.append(ch)
+        prev_cjk = cur_cjk
+    return "".join(out)
+
+
+def remove_cjk_spaces(text: str) -> str:
+    if " " not in text:
+        return text
+    chars = list(text)
+    out: list[str] = []
+    for idx, ch in enumerate(chars):
+        if ch == " " and 0 < idx < len(chars) - 1:
+            if _is_cjk_char(chars[idx - 1]) and _is_cjk_char(chars[idx + 1]):
+                continue
+        out.append(ch)
+    return "".join(out)
+
+
+def merge_adjacent_markers(text: str, *, start_marker: str = "[[[", end_marker: str = "]]]") -> str:
+    needle = f"{end_marker}{start_marker}"
+    while needle in text:
+        text = text.replace(needle, "")
+    return text
+
+
+def _md_renderer() -> MarkdownIt:
+    md = MarkdownIt("commonmark", {"html": False, "linkify": False})
+    md.enable("table")
+    return md
+
+
+def markdown_to_plain_text(markdown: str) -> str:
+    if not markdown:
+        return ""
+    text = _HTML_TABLE_RE.sub(" ", markdown)
+    md = _md_renderer()
+    tokens = md.parse(text)
+
+    out: list[str] = []
+    in_table = 0
+    for token in tokens:
+        if token.type == "table_open":
+            in_table += 1
+            continue
+        if token.type == "table_close":
+            in_table = max(0, in_table - 1)
+            continue
+        if in_table:
+            continue
+        if token.type != "inline":
+            continue
+        for child in token.children or []:
+            if child.type in {"text", "code_inline"}:
+                out.append(child.content)
+            elif child.type == "softbreak":
+                out.append("\n")
+            elif child.type == "hardbreak":
+                out.append("\n")
+            elif child.type == "image":
+                if child.content:
+                    out.append(child.content)
+
+    collapsed = _WS_RE.sub(" ", " ".join(out)).strip()
+    collapsed = _TAG_RE.sub(" ", collapsed)
+    return _WS_RE.sub(" ", collapsed).strip()
+
+
+def normalize_query_punctuation(text: str) -> str:
+    if not text:
+        return ""
+    return re.sub(r"[，。、《》、；：！？（）【】「」『』·…—]+", " ", text)
+
+
+def split_mixed_cjk_latin(token: str) -> list[str]:
+    if not token:
+        return []
+    parts: list[str] = []
+    buf: list[str] = []
+    buf_is_cjk: bool | None = None
+    for ch in token:
+        cur_is_cjk = _is_cjk_char(ch)
+        if buf_is_cjk is None or cur_is_cjk == buf_is_cjk:
+            buf.append(ch)
+            buf_is_cjk = cur_is_cjk
+            continue
+        parts.append("".join(buf))
+        buf = [ch]
+        buf_is_cjk = cur_is_cjk
+    if buf:
+        parts.append("".join(buf))
+    return parts
+
+
+def rewrite_search_query(user_query: str) -> str:
+    cleaned = normalize_query_punctuation(user_query)
+    cleaned = _WS_RE.sub(" ", cleaned).strip()
+    if not cleaned:
+        return ""
+
+    out: list[str] = []
+    for raw in cleaned.split(" "):
+        if not raw:
+            continue
+        upper = raw.upper()
+        if upper in {"AND", "OR"}:
+            out.append(upper)
+            continue
+
+        segments = split_mixed_cjk_latin(raw)
+        for seg in segments:
+            if not seg:
+                continue
+            if all(_is_cjk_char(ch) for ch in seg):
+                phrase = insert_cjk_spaces(seg)
+                out.append(f"\"{phrase}\"")
+            else:
+                safe = re.sub(r"[^0-9A-Za-z._+-]+", "", seg)
+                if safe:
+                    out.append(safe.lower())
+
+    return " ".join(out)
+

deepresearch_flow/paper/template_registry.py

@@ -24,6 +24,7 @@ class TemplateBundle:
 class StageDefinition:
     name: str
     fields: list[str]
+    depends_on: list[str] | None = None
 
 
 _TEMPLATES: dict[str, TemplateBundle] = {

deepresearch_flow/paper/templates/deep_read.md.j2

@@ -3,6 +3,10 @@
 {% set is_zh = output_language == "zh" %}
 **{{ "作者 / Authors" if is_zh else "Authors" }}:** {{ paper_authors | join(", ") }}
 
+{% if paper_institutions %}
+**{{ "单位 / Institutions" if is_zh else "Institutions" }}:** {{ paper_institutions | join(", ") }}
+{% endif %}
+
 {% if output_language %}
 **{{ "输出语言 / Output Language" if is_zh else "Output Language" }}:** {{ output_language }}
 {% endif %}

deepresearch_flow/paper/templates/deep_read_phi.md.j2

@@ -3,6 +3,10 @@
 {% set is_zh = output_language == "zh" %}
 **{{ "作者 / Authors" if is_zh else "Authors" }}:** {{ paper_authors | join(", ") }}
 
+{% if paper_institutions %}
+**{{ "单位 / Institutions" if is_zh else "Institutions" }}:** {{ paper_institutions | join(", ") }}
+{% endif %}
+
 {% if output_language %}
 **{{ "输出语言 / Output Language" if is_zh else "Output Language" }}:** {{ output_language }}
 {% endif %}

deepresearch_flow/paper/templates/default_paper.md.j2

@@ -3,6 +3,10 @@
 {% set is_zh = output_language == "zh" %}
 **{{ "作者 / Authors" if is_zh else "Authors" }}:** {{ paper_authors | join(", ") }}
 
+{% if paper_institutions %}
+**{{ "单位 / Institutions" if is_zh else "Institutions" }}:** {{ paper_institutions | join(", ") }}
+{% endif %}
+
 {% if output_language %}
 **{{ "输出语言 / Output Language" if is_zh else "Output Language" }}:** {{ output_language }}
 {% endif %}

deepresearch_flow/paper/templates/eight_questions.md.j2

@@ -4,6 +4,10 @@
 
 **{{ "作者 / Authors" if is_zh else "Authors" }}:** {{ paper_authors | join(", ") }}
 
+{% if paper_institutions %}
+**{{ "单位 / Institutions" if is_zh else "Institutions" }}:** {{ paper_institutions | join(", ") }}
+{% endif %}
+
 {% if output_language %}
 **{{ "输出语言 / Output Language" if is_zh else "Output Language" }}:** {{ output_language }}
 {% endif %}

deepresearch_flow/paper/web/app.py

@@ -91,7 +91,7 @@ def create_app(
         pdf_roots=pdf_roots,
     )
     md = create_md_renderer()
-    static_base_url = static_base_url or os.getenv("PAPER_DB_STATIC_BASE_URL")
+    static_base_url = static_base_url or os.getenv("PAPER_DB_STATIC_BASE") or os.getenv("PAPER_DB_STATIC_BASE_URL")
     static_mode = _normalize_static_mode(static_mode or os.getenv("PAPER_DB_STATIC_MODE"))
     resolved_mode = _resolve_static_mode(static_mode, static_base_url)
     export_dir_value = static_export_dir or os.getenv("PAPER_DB_STATIC_EXPORT_DIR")
@@ -111,7 +111,12 @@ def create_app(
     asset_config = None
     if resolved_mode == "prod":
         if not static_base_url:
-            logger.warning("Static mode set to prod without base URL; falling back to dev asset routes.")
+            logger.warning(
+                "Static mode set to prod without base URL; falling back to dev asset routes "
+                "(static_mode=%s, static_base_url=%s)",
+                static_mode,
+                static_base_url or "<empty>",
+            )
             resolved_mode = "dev"
         else:
             asset_config = build_static_assets(
@@ -149,8 +154,10 @@ def create_app(
         )
     elif pdf_roots:
         logger.warning(
-            "PDF.js viewer assets not found at %s; PDF Viewer mode will be unavailable.",
+            "PDF.js viewer assets not found at %s; PDF Viewer mode will be unavailable "
+            "(pdf_roots=%d).",
             PDFJS_STATIC_DIR,
+            len(pdf_roots),
         )
     if STATIC_DIR.exists():
         routes.append(