Support MinerU HTTP file parse endpoint

@@ -24,10 +24,11 @@ parsing:
     xlsx_mode: "row_text"
     min_chars: 80
   mineru:
-    mode: "cli"
+    mode: "http"
     cli_bin: "mineru"
     output_dir: "data/parsed_docs/mineru_raw"
     http_base_url: "http://172.23.184.9:8002"
+    http_parse_path: "/file_parse"
     api_key: "mineru"
     timeout_seconds: 600
     fallback_to_local: false

@@ -128,12 +128,13 @@ def parse_pdf_with_http(
     if mineru_config.get("api_key"):
         headers["Authorization"] = f"Bearer {mineru_config['api_key']}"
 
-    # The checklist does not define a universal MinerU HTTP contract. This
-    # implementation expects a replaceable service exposing POST /parse and
-    # returning {"markdown": "..."} or {"documents": [{"content": "..."}]}.
+    endpoint = str(mineru_config.get("http_parse_path") or "/file_parse")
+    if not endpoint.startswith("/"):
+        endpoint = "/" + endpoint
+
     with target.open("rb") as file:
         response = requests.post(
-            f"{base_url}/parse",
+            f"{base_url}{endpoint}",
             files={"file": (target.name, file, "application/pdf")},
             headers=headers,
             timeout=int(mineru_config.get("timeout_seconds", 600)),
@@ -142,13 +143,13 @@ def parse_pdf_with_http(
         raise MinerUParseError(f"MinerU HTTP failed with {response.status_code}: {response.text[:500]}")
 
     payload = response.json()
-    contents: list[str] = []
-    if isinstance(payload.get("documents"), list):
-        contents = [compact_text(item.get("content")) for item in payload["documents"]]
-    elif payload.get("markdown"):
-        contents = [compact_text(payload["markdown"])]
-    else:
-        raise MinerUParseError("MinerU HTTP response must include `markdown` or `documents`")
+    contents = extract_mineru_contents(payload)
+    if not contents:
+        raw_path = _write_unrecognized_mineru_payload(target, payload, mineru_config)
+        raise MinerUParseError(
+            "MinerU HTTP response did not include recognizable text content. "
+            f"Saved raw response to {raw_path}"
+        )
 
     docs: list[ParsedDocument] = []
     for index, content in enumerate(contents, start=1):
@@ -160,7 +161,57 @@ def parse_pdf_with_http(
                 source_file=target.name,
                 file_type="pdf",
                 content=content,
-                metadata={"parser": "mineru:http"},
+                metadata={"parser": "mineru:http", "mineru_endpoint": endpoint},
             )
         )
     return docs
+
+
+def extract_mineru_contents(payload: Any) -> list[str]:
+    contents: list[str] = []
+    _collect_text_values(payload, contents)
+    return [content for content in dict.fromkeys(contents) if content]
+
+
+def _collect_text_values(value: Any, contents: list[str]) -> None:
+    if isinstance(value, str):
+        text = compact_text(value)
+        if len(text) >= 20:
+            contents.append(text)
+        return
+
+    if isinstance(value, list):
+        for item in value:
+            _collect_text_values(item, contents)
+        return
+
+    if not isinstance(value, dict):
+        return
+
+    for key in (
+        "markdown",
+        "md",
+        "content",
+        "text",
+        "plain_text",
+        "page_content",
+        "document",
+    ):
+        if key in value:
+            _collect_text_values(value[key], contents)
+
+    for key in ("documents", "pages", "chunks", "data", "result", "results"):
+        if key in value:
+            _collect_text_values(value[key], contents)
+
+
+def _write_unrecognized_mineru_payload(
+    pdf_path: Path,
+    payload: dict[str, Any],
+    mineru_config: dict[str, Any],
+) -> Path:
+    output_root = Path(mineru_config.get("output_dir", "data/parsed_docs/mineru_raw"))
+    output_root.mkdir(parents=True, exist_ok=True)
+    raw_path = output_root / f"{pdf_path.stem}.response.json"
+    write_json(raw_path, payload)
+    return raw_path