fix model resolution

src/fast_agent/core/direct_factory.py

@@ -19,7 +19,7 @@
 from fast_agent.core import Core
 from fast_agent.core.exceptions import AgentConfigError, ModelConfigError
 from fast_agent.core.logging.logger import get_logger
-from fast_agent.core.model_resolution import resolve_model_spec
+from fast_agent.core.model_resolution import HARDCODED_DEFAULT_MODEL, resolve_model_spec
 from fast_agent.core.validation import get_dependencies_groups
 from fast_agent.event_progress import ProgressAction
 from fast_agent.interfaces import (
@@ -82,8 +82,6 @@ async def __call__(
     ) -> AgentDict: ...
 
 
-HARDCODED_DEFAULT_MODEL = "gpt-5-mini.low"
-
 
 def get_model_factory(
     context,

src/fast_agent/core/model_resolution.py

@@ -5,6 +5,8 @@
 import os
 from typing import Any
 
+HARDCODED_DEFAULT_MODEL = "gpt-5-mini.low"
+
 
 def resolve_model_spec(
     context: Any,

src/fast_agent/mcp/mcp_aggregator.py

@@ -29,7 +29,7 @@
 from fast_agent.context_dependent import ContextDependent
 from fast_agent.core.exceptions import ServerSessionTerminatedError
 from fast_agent.core.logging.logger import get_logger
-from fast_agent.core.model_resolution import resolve_model_spec
+from fast_agent.core.model_resolution import HARDCODED_DEFAULT_MODEL, resolve_model_spec
 from fast_agent.event_progress import ProgressAction
 from fast_agent.mcp.common import SEP, create_namespaced_name, is_namespaced_name
 from fast_agent.mcp.gen_client import gen_client
@@ -357,7 +357,7 @@ def session_factory(read_stream, write_stream, read_timeout, **kwargs):
                     self.context,
                     model=self.config.model,
                     cli_model=cli_model_override,
-                    fallback_to_hardcoded=False,
+                    hardcoded_default=HARDCODED_DEFAULT_MODEL,
                 )
                 if model_source:
                     logger.info(

src/fast_agent/mcp/sampling.py

@@ -17,6 +17,7 @@
 from fast_agent.agents.agent_types import AgentConfig
 from fast_agent.agents.llm_agent import LlmAgent
 from fast_agent.core.logging.logger import get_logger
+from fast_agent.core.model_resolution import HARDCODED_DEFAULT_MODEL, resolve_model_spec
 from fast_agent.interfaces import FastAgentLLMProtocol
 from fast_agent.llm.sampling_converter import SamplingConverter
 from fast_agent.mcp.helpers.server_config_helpers import get_server_config
@@ -106,6 +107,7 @@ async def sample(
 
     model: str | None = None
     api_key: str | None = None
+    app_context: Any | None = None
     try:
         # Extract model from server config using type-safe helper
         server_config = get_server_config(context)
@@ -144,11 +146,20 @@ async def sample(
                 # Fall back to system default model
                 if model is None:
                     try:
-                        if app_context and app_context.config and app_context.config.default_model:
-                            model = app_context.config.default_model
-                            logger.debug(f"Using system default model for sampling: {model}")
+                        cli_model_override = None
+                        if app_context and app_context.config:
+                            cli_model_override = getattr(
+                                app_context.config, "cli_model_override", None
+                            )
+                        model, model_source = resolve_model_spec(
+                            app_context,
+                            cli_model=cli_model_override,
+                            hardcoded_default=HARDCODED_DEFAULT_MODEL,
+                        )
+                        if model:
+                            logger.debug(f"Using {model_source} model for sampling: {model}")
                     except Exception as e:
-                        logger.debug(f"Could not get system default model: {e}")
+                        logger.debug(f"Could not resolve default model for sampling: {e}")
 
         if model is None:
             raise ValueError(