deepresearch-flow 0.2.0__py3-none-any.whl → 0.3.0__py3-none-any.whl

deepresearch_flow/cli.py

@@ -4,6 +4,7 @@ import click
 
 from deepresearch_flow.paper.cli import paper
 from deepresearch_flow.recognize.cli import recognize
+from deepresearch_flow.translator.cli import translator
 
 
 @click.group()
@@ -13,6 +14,7 @@ def cli() -> None:
 
 cli.add_command(paper)
 cli.add_command(recognize)
+cli.add_command(translator)
 
 
 def main() -> None:

deepresearch_flow/paper/config.py

@@ -40,6 +40,7 @@ class ProviderConfig:
     location: str | None
     credentials_path: str | None
     anthropic_version: str | None
+    max_tokens: int | None
     structured_mode: str
     extra_headers: dict[str, str]
     system_prompt: str | None
@@ -102,6 +103,15 @@ def _as_str(value: Any, default: str | None = None) -> str | None:
     return str(value)
 
 
+def _ensure_http_scheme(base_url: str, *, default_scheme: str = "http://") -> str:
+    normalized = base_url.strip()
+    if normalized.startswith(("http://", "https://")):
+        scheme, rest = normalized.split("://", 1)
+        rest = rest.lstrip("/")
+        return f"{scheme}://{rest}" if rest else f"{scheme}://"
+    return f"{default_scheme}{normalized.lstrip('/')}"
+
+
 def load_config(path: str) -> PaperConfig:
     config_path = Path(path)
     if not config_path.exists():
@@ -158,6 +168,8 @@ def load_config(path: str) -> PaperConfig:
                 raise ValueError(f"Provider '{name}' requires base_url")
         elif provider_type == "azure_openai" and endpoint:
             base_url = endpoint
+        if provider_type == "ollama" and base_url:
+            base_url = _ensure_http_scheme(base_url)
 
         api_keys = _as_list(provider.get("api_keys"))
         if not api_keys:
@@ -188,6 +200,8 @@ def load_config(path: str) -> PaperConfig:
         location = _as_str(provider.get("location"), None)
         credentials_path = _as_str(provider.get("credentials_path"), None)
         anthropic_version = _as_str(provider.get("anthropic_version"), None)
+        max_tokens = provider.get("max_tokens")
+        max_tokens_value = int(max_tokens) if max_tokens is not None else None
 
         if provider_type == "azure_openai":
             if not base_url:
@@ -221,6 +235,7 @@ def load_config(path: str) -> PaperConfig:
                 location=location,
                 credentials_path=credentials_path,
                 anthropic_version=anthropic_version,
+                max_tokens=max_tokens_value,
                 structured_mode=structured_mode,
                 extra_headers=extra_headers,
                 system_prompt=_as_str(provider.get("system_prompt"), None),

deepresearch_flow/paper/db.py

@@ -584,6 +584,13 @@ def register_db_commands(db_group: click.Group) -> None:
         default=(),
         help="Optional markdown root directory (repeatable) for source viewing",
     )
+    @click.option(
+        "--md-translated-root",
+        "md_translated_roots",
+        multiple=True,
+        default=(),
+        help="Optional markdown root directory (repeatable) for translated viewing",
+    )
     @click.option(
         "--pdf-root",
         "pdf_roots",
@@ -606,6 +613,7 @@ def register_db_commands(db_group: click.Group) -> None:
         input_paths: tuple[str, ...],
         bibtex_path: str | None,
         md_roots: tuple[str, ...],
+        md_translated_roots: tuple[str, ...],
         pdf_roots: tuple[str, ...],
         cache_dir: str | None,
         no_cache: bool,
@@ -623,6 +631,7 @@ def register_db_commands(db_group: click.Group) -> None:
                 fallback_language=fallback_language,
                 bibtex_path=Path(bibtex_path) if bibtex_path else None,
                 md_roots=[Path(root) for root in md_roots],
+                md_translated_roots=[Path(root) for root in md_translated_roots],
                 pdf_roots=[Path(root) for root in pdf_roots],
                 cache_dir=Path(cache_dir) if cache_dir else None,
                 use_cache=not no_cache,

deepresearch_flow/paper/llm.py

@@ -32,6 +32,7 @@ async def call_provider(
     timeout: float,
     structured_mode: str,
     client: httpx.AsyncClient,
+    max_tokens: int | None = None,
 ) -> str:
     headers = dict(provider.extra_headers)
     if api_key and provider.type == "openai_compatible":
@@ -98,6 +99,7 @@ async def call_provider(
             model=model,
             messages=messages,
             anthropic_version=provider.anthropic_version,
+            max_tokens=max_tokens or 2048,
         )
 
     if provider.type == "openai_compatible":