refactor: centralize llm defaults

src/scribae/brief.py

@@ -14,7 +14,7 @@
 from pydantic_ai.settings import ModelSettings
 
 from .io_utils import NoteDetails, load_note
-from .llm import LLM_OUTPUT_RETRIES, OpenAISettings, make_model
+from .llm import LLM_OUTPUT_RETRIES, LLM_TIMEOUT_SECONDS, OpenAISettings, make_model
 from .project import ProjectConfig
 from .prompts import SYSTEM_PROMPT, PromptBundle, build_prompt_bundle
 
@@ -33,9 +33,6 @@
     "save_prompt_artifacts",
 ]
 
-LLM_TIMEOUT_SECONDS = 300.0
-
-
 class BriefingError(Exception):
     """Raised when a brief cannot be generated."""
 

src/scribae/idea.py

@@ -14,12 +14,14 @@
 from pydantic_ai.settings import ModelSettings
 
 from .io_utils import NoteDetails, load_note
-from .llm import DEFAULT_MODEL_NAME, LLM_OUTPUT_RETRIES, OpenAISettings, make_model
+from .llm import (
+    LLM_OUTPUT_RETRIES,
+    LLM_TIMEOUT_SECONDS,
+    OpenAISettings,
+    make_model,
+)
 from .project import ProjectConfig
 
-DEFAULT_IDEA_MODEL = DEFAULT_MODEL_NAME
-IDEA_TIMEOUT_SECONDS = 300.0
-
 
 class IdeaError(Exception):
     """Raised when ideas cannot be generated."""
@@ -138,7 +140,7 @@ def generate_ideas(
     reporter: Reporter = None,
     settings: OpenAISettings | None = None,
     agent: Agent[None, IdeaList] | None = None,
-    timeout_seconds: float = IDEA_TIMEOUT_SECONDS,
+    timeout_seconds: float = LLM_TIMEOUT_SECONDS,
 ) -> IdeaList:
     """Run the LLM call and return validated ideas."""
 
@@ -290,7 +292,6 @@ def _report(reporter: Reporter, message: str) -> None:
 
 
 __all__ = [
-    "DEFAULT_IDEA_MODEL",
     "Idea",
     "IdeaContext",
     "IdeaError",

src/scribae/idea_cli.py

@@ -4,7 +4,8 @@
 
 import typer
 
-from .idea import DEFAULT_IDEA_MODEL, IdeaError, generate_ideas, prepare_context, render_json, save_prompt_artifacts
+from .idea import IdeaError, generate_ideas, prepare_context, render_json, save_prompt_artifacts
+from .llm import DEFAULT_MODEL_NAME
 from .project import default_project, load_project
 
 
@@ -50,7 +51,7 @@ def idea_command(
         help="Project name used to load projects/<name>.yaml for prompt context.",
     ),
     model: str = typer.Option(  # noqa: B008
-        DEFAULT_IDEA_MODEL,
+        DEFAULT_MODEL_NAME,
         "--model",
         "-m",
         help="OpenAI/Ollama model name to request.",

src/scribae/llm.py

@@ -7,10 +7,11 @@
 from pydantic_ai.providers.openai import OpenAIProvider
 from pydantic_ai.settings import ModelSettings
 
-DEFAULT_MODEL_NAME = "mistral-nemo"
+DEFAULT_MODEL_NAME = "ministral-3:8b"
 DEFAULT_BASE_URL = "http://localhost:11434/v1"
 DEFAULT_API_KEY = "no-key"
 LLM_OUTPUT_RETRIES = 2
+LLM_TIMEOUT_SECONDS = 300.0
 
 
 @dataclass(frozen=True)
@@ -48,5 +49,6 @@ def make_model(model_name: str, *, model_settings: ModelSettings,
     "DEFAULT_BASE_URL",
     "DEFAULT_API_KEY",
     "LLM_OUTPUT_RETRIES",
+    "LLM_TIMEOUT_SECONDS",
     "make_model",
 ]

src/scribae/meta.py

@@ -16,12 +16,9 @@
 from pydantic_ai.settings import ModelSettings
 
 from .brief import SeoBrief
-from .llm import LLM_OUTPUT_RETRIES, OpenAISettings, make_model
+from .llm import LLM_OUTPUT_RETRIES, LLM_TIMEOUT_SECONDS, OpenAISettings, make_model
 from .project import ProjectConfig
 
-DEFAULT_META_MODEL = "mistral-nemo"
-META_TIMEOUT_SECONDS = 300.0
-
 Reporter = Callable[[str], None] | None
 
 
@@ -278,7 +275,7 @@ def generate_metadata(
     reporter: Reporter = None,
     agent: Agent[None, ArticleMeta] | None = None,
     prompts: PromptBundle | None = None,
-    timeout_seconds: float = META_TIMEOUT_SECONDS,
+    timeout_seconds: float = LLM_TIMEOUT_SECONDS,
     force_llm_on_missing: bool = True,
 ) -> ArticleMeta:
     """Generate final article metadata, calling the LLM when needed."""
@@ -670,7 +667,6 @@ def _report(reporter: Reporter, message: str) -> None:
 __all__ = [
     "ArticleMeta",
     "BodyDocument",
-    "DEFAULT_META_MODEL",
     "MetaBriefError",
     "MetaContext",
     "MetaError",

src/scribae/meta_cli.py

@@ -4,8 +4,8 @@
 
 import typer
 
+from .llm import DEFAULT_MODEL_NAME
 from .meta import (
-    DEFAULT_META_MODEL,
     ArticleMeta,
     MetaBriefError,
     MetaError,
@@ -46,7 +46,7 @@ def meta_command(
         help="Project name used to load projects/<name>.yaml for prompt context.",
     ),
     model: str = typer.Option(  # noqa: B008
-        DEFAULT_META_MODEL,
+        DEFAULT_MODEL_NAME,
         "--model",
         "-m",
         help="OpenAI/Ollama model name to request.",

src/scribae/translate/postedit.py

@@ -10,7 +10,7 @@
 from pydantic_ai import Agent, NativeOutput, UnexpectedModelBehavior
 from pydantic_ai.settings import ModelSettings
 
-from scribae.llm import LLM_OUTPUT_RETRIES, OpenAISettings, make_model
+from scribae.llm import DEFAULT_MODEL_NAME, LLM_OUTPUT_RETRIES, OpenAISettings, make_model
 
 from .markdown_segmenter import ProtectedText
 
@@ -43,7 +43,7 @@ def __init__(
         self,
         agent: Agent[None, str] | None = None,
         *,
-        model_name: str = "mistral-nemo",
+        model_name: str = DEFAULT_MODEL_NAME,
         temperature: float = 0.2,
         create_agent: bool = True,
         max_chars: int | None = 4_000,

src/scribae/translate_cli.py

@@ -8,6 +8,7 @@
 import typer
 import yaml
 
+from scribae.llm import DEFAULT_MODEL_NAME
 from scribae.project import load_project
 from scribae.translate import (
     LLMPostEditor,
@@ -109,7 +110,7 @@ def translate(
         help="Additional regex patterns to protect.",
     ),
     postedit_model: str = typer.Option(  # noqa: B008
-        "mistral-nemo",
+        DEFAULT_MODEL_NAME,
         "--postedit-model",
         help="Model name for post-edit LLM pass.",
     ),

src/scribae/write_cli.py

@@ -4,9 +4,9 @@
 
 import typer
 
+from .llm import DEFAULT_MODEL_NAME
 from .project import default_project, load_project
 from .writer import (
-    DEFAULT_WRITE_MODEL,
     EvidenceMode,
     WritingError,
     WritingValidationError,
@@ -37,7 +37,7 @@ def write_command(
         help="Project name used to load projects/<name>.yaml for prompt context.",
     ),
     model: str = typer.Option(  # noqa: B008
-        DEFAULT_WRITE_MODEL,
+        DEFAULT_MODEL_NAME,
         "--model",
         "-m",
         help="OpenAI/Ollama model name to request.",

src/scribae/writer.py

@@ -14,14 +14,11 @@
 
 from .brief import SeoBrief
 from .io_utils import NoteDetails, load_note
-from .llm import make_model
+from .llm import LLM_TIMEOUT_SECONDS, make_model
 from .project import ProjectConfig
 from .prompts_writer import SYSTEM_PROMPT, build_user_prompt
 from .snippets import SnippetSelection, build_snippet_block
 
-DEFAULT_WRITE_MODEL = "mistral-nemo"
-WRITER_TIMEOUT_SECONDS = 300.0
-
 Reporter = Callable[[str], None] | None
 
 
@@ -253,9 +250,9 @@ async def _call() -> str:
         return str(output).strip()
 
     try:
-        return asyncio.run(asyncio.wait_for(_call(), WRITER_TIMEOUT_SECONDS))
+        return asyncio.run(asyncio.wait_for(_call(), LLM_TIMEOUT_SECONDS))
     except TimeoutError as exc:
-        raise WritingLLMError(f"LLM request timed out after {int(WRITER_TIMEOUT_SECONDS)} seconds.") from exc
+        raise WritingLLMError(f"LLM request timed out after {int(LLM_TIMEOUT_SECONDS)} seconds.") from exc
     except KeyboardInterrupt:
         raise
     except Exception as exc:  # pragma: no cover - surfaced to CLI
@@ -312,7 +309,6 @@ def _report(reporter: Reporter, message: str) -> None:
 
 
 __all__ = [
-    "DEFAULT_WRITE_MODEL",
     "EvidenceMode",
     "WritingContext",
     "WritingError",