Merge branch 'benchmark_simpleqa' of https://github.com/vamplabAI/sgr-deep-research into benchmark_simpleqa

Author: maksimov maksim

docs/examples/simple_shine_cli.py

@@ -0,0 +1,103 @@
+"""Very basic demo for processing research queries and clarification requests
+from agent.
+
+Usage:
+    pip install rich openai
+    python -m docs.examples.simple_shine_cli
+"""
+
+import json
+
+from openai import OpenAI
+from rich.console import Console
+from rich.prompt import Prompt
+
+console = Console()
+client = OpenAI(base_url="http://localhost:8010/v1", api_key="dummy")
+
+
+def safe_get_delta(chunk):
+    if not hasattr(chunk, "choices") or not chunk.choices:
+        return None
+    first_choice = chunk.choices[0]
+    if first_choice is None or not hasattr(first_choice, "delta"):
+        return None
+    return first_choice.delta
+
+
+def stream_response_until_tool_call_or_end(model, messages):
+    """Real-time streaming."""
+    response = client.chat.completions.create(
+        model=model,
+        messages=messages,
+        stream=True,
+        temperature=0,
+    )
+
+    agent_id = None
+    full_content = ""
+    clarification_questions = None
+
+    for chunk in response:
+        if hasattr(chunk, "model") and chunk.model and chunk.model.startswith("sgr_agent_"):
+            agent_id = chunk.model
+
+        delta = safe_get_delta(chunk)
+        if delta is None:
+            continue
+
+        if hasattr(delta, "tool_calls") and delta.tool_calls:
+            for tool_call in delta.tool_calls:
+                if tool_call.function and tool_call.function.name == "clarificationtool":
+                    try:
+                        args = json.loads(tool_call.function.arguments)
+                        clarification_questions = args.get("questions", [])
+                    except Exception as e:
+                        console.print(f"[red]Error parsing clarification: {e}[/red]")
+            # stop streaming after tool calling detect
+            return full_content, clarification_questions, agent_id
+
+        if hasattr(delta, "content") and delta.content:
+            text = delta.content
+            full_content += text
+            console.print(text, end="", style="white")
+
+    return full_content, None, agent_id
+
+
+console.print("\n[bold green]Research Assistant v1.0[/bold green]", style="bold white")
+initial_request = Prompt.ask("[bold yellow]Enter your research request[/bold yellow]")
+console.print(f"\nStarting research: [bold]{initial_request}[/bold]")
+
+current_model = "sgr_agent"
+messages = [{"role": "user", "content": initial_request}]
+agent_id = None
+
+while True:
+    console.print()
+
+    full_content, clarification_questions, returned_agent_id = stream_response_until_tool_call_or_end(
+        model=current_model, messages=messages
+    )
+
+    if returned_agent_id:
+        agent_id = returned_agent_id
+        current_model = agent_id
+    if clarification_questions is not None:
+        console.print()
+
+        console.print("\n[bold red]Clarification needed:[/bold red]")
+        for i, question in enumerate(clarification_questions, 1):
+            console.print(f"[bold]{i}.[/bold] {question}", style="yellow")
+
+        clarification = Prompt.ask("[bold grey]Enter your clarification[/bold grey]")
+        console.print(f"\n[bold green]Providing clarification:[/bold green] [italic]{clarification}[/italic]")
+
+        messages.append({"role": "user", "content": clarification})
+        continue
+
+    else:
+        console.print()
+        break
+
+console.print("\n[bold green] Report will be prepared in appropriate directory![/bold green]")

sgr_deep_research/core/agents/tools_agent.py

@@ -68,7 +68,7 @@ async def _prepare_tools(self) -> list[ChatCompletionFunctionToolParam]:
             tools -= {
                 WebSearchTool,
             }
-        return [pydantic_function_tool(tool, name=tool.tool_name) for tool in tools]
+        return [pydantic_function_tool(tool, name=tool.tool_name, description="") for tool in tools]
 
     async def _reasoning_phase(self) -> None:
         """No explicit reasoning phase, reasoning is done internally by LLM."""