add model to mini agent and eval context

Author: haoranpb

src/bcbench/agent/mini/agent.py

@@ -52,6 +52,7 @@ def parse_action(self, response: dict) -> dict:
 def run_mini_agent(
     entry: DatasetEntry,
     repo_path: Path,
+    model: str,
     enable_bc_tools: bool = False,
     include_project_paths: bool = False,
     container_name: str | None = None,
@@ -84,7 +85,7 @@ def run_mini_agent(
     BCAgent = _create_bc_agent_class()
 
     agent = BCAgent(
-        LitellmModel(model_name="azure/gpt-4.1"),
+        LitellmModel(model_name=model),
         BCEnvironment(
             container_name=container_name,
             repo_path=str(repo_path),

src/bcbench/commands/evaluate.py

@@ -2,6 +2,7 @@
 
 import shutil
 from pathlib import Path
+from typing import Literal
 
 import typer
 from typing_extensions import Annotated
@@ -25,6 +26,7 @@ def evaluate_mini(
     container_name: ContainerName,
     username: ContainerUsername,
     password: ContainerPassword,
+    model: Annotated[Literal["azure/gpt-4.1"], typer.Option(help="Azure AI Foundry Model to use for mini-bc-agent")] = "azure/gpt-4.1",
     dataset_path: DatasetPath = _config.paths.dataset_path,
     repo_path: RepoPath = _config.paths.nav_repo_path,
     output_dir: OutputDir = _config.paths.evaluation_results_path,
@@ -66,6 +68,7 @@ def evaluate_mini(
         container_name=container_name,
         username=username,
         password=password,
+        model=model,
         agent_name="mini-bc-agent",
         agent_options={
             "enable_bc_tools": enable_bc_tools,
@@ -79,6 +82,7 @@ def evaluate_mini(
         lambda ctx: run_mini_agent(
             entry=ctx.entry,
             repo_path=ctx.repo_path,
+            model=ctx.model,
             include_project_paths=include_project_paths,
             enable_bc_tools=ctx.get_agent_option("enable_bc_tools", False),
             container_name=ctx.container_name,
@@ -136,6 +140,7 @@ def evaluate_copilot(
         container_name=container_name,
         username=username,
         password=password,
+        model=model,
         agent_name="GitHub Copilot CLI",
     )
 
@@ -144,7 +149,7 @@ def evaluate_copilot(
         lambda ctx: run_copilot_agent(
             entry=ctx.entry,
             repo_path=ctx.repo_path,
-            model=model,
+            model=ctx.model,
             include_project_paths=include_project_paths,
             output_dir=ctx.result_dir,
         ),

src/bcbench/commands/run.py

@@ -1,6 +1,7 @@
 """CLI commands for running agents."""
 
 from pathlib import Path
+from typing import Literal
 
 import typer
 from typing_extensions import Annotated
@@ -33,6 +34,7 @@ def run_mini(
     container_name: ContainerName,
     username: ContainerUsername,
     password: ContainerPassword,
+    model: Annotated[Literal["azure/gpt-4.1"], typer.Option(help="Azure AI Foundry Model to use for mini-bc-agent")] = "azure/gpt-4.1",
     dataset_path: DatasetPath = _config.paths.dataset_path,
     repo_path: RepoPath = _config.paths.nav_repo_path,
     enable_bc_tools: Annotated[
@@ -59,6 +61,7 @@ def run_mini(
     run_mini_agent(
         entry=entry,
         repo_path=repo_path,
+        model=model,
         enable_bc_tools=enable_bc_tools,
         container_name=container_name,
         username=username,

src/bcbench/evaluate/evaluation_context.py

@@ -27,10 +27,11 @@ class EvaluationContext:
     # BC Container configuration
     container_name: str
     password: str
-    username: str = "admin"
+    username: str
 
     # Agent metadata
-    agent_name: str = "Agent"
+    agent_name: str
+    model: str
 
     # Agent-specific options (stored as dict for flexibility)
     agent_options: dict[str, Any] | None = None