fix(agents): update agents to beeai sdk changes (#1067)

Signed-off-by: Radek Ježek <radek.jezek@ibm.com>

Author: jezekra1

agents/community/aider/Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.11-slim-bookworm
+FROM python:3.12-slim-bookworm
 ARG RELEASE_VERSION="main"
 COPY --from=ghcr.io/astral-sh/uv:0.8.9 /uv /bin/
 WORKDIR /agents/community/aider

agents/community/aider/pyproject.toml

@@ -6,9 +6,9 @@ authors = [
     { name = "IBM Corp." },
 ]
 readme = "README.md"
-requires-python = ">=3.11,<3.12"
+requires-python = ">=3.12,<3.13"
 dependencies = [
-    "aider-chat>=0.18.1",
+    "aider-chat>=0.86.1",
     "pydantic-settings>=2.7.1",
     "pydantic>=2.10.6",
     "google-generativeai>=0.8.4",
@@ -34,3 +34,7 @@ line-length = 120
 
 [tool.uv.sources]
 beeai-sdk = { path = "../../../apps/beeai-sdk", editable = true }
+
+[tool.pyright]
+venvPath = "."
+venv = ".venv"

agents/community/aider/uv.lock

@@ -1,3 +1,7 @@
 # Copyright 2025 © BeeAI a Series of LF Projects, LLC
 # SPDX-License-Identifier: Apache-2.0
+from gpt_researcher_agent.env_patch import patch_os_environ
 
+# gpt researcher was never intended to be used in a server, we need to apply some ugly workarounds
+# This needs to be done first before importing anything else
+patch_os_environ()

agents/community/gpt-researcher/gpt_researcher_agent/__init__.py

@@ -8,8 +8,17 @@
 from a2a.types import AgentSkill, Message
 from gpt_researcher import GPTResearcher
 
+from gpt_researcher_agent.env_patch import with_local_env
 
-from beeai_sdk.a2a.extensions import TrajectoryExtensionServer, TrajectoryExtensionSpec, AgentDetail
+from beeai_sdk.a2a.extensions import (
+    TrajectoryExtensionServer,
+    TrajectoryExtensionSpec,
+    AgentDetail,
+    LLMServiceExtensionServer,
+    LLMServiceExtensionSpec,
+    EmbeddingServiceExtensionServer,
+    EmbeddingServiceExtensionSpec,
+)
 from beeai_sdk.a2a.types import RunYield
 from beeai_sdk.server import Server
 from beeai_sdk.server.context import RunContext
@@ -18,21 +27,12 @@
 
 
 @server.agent(
-    name="GPT Researcher",
+    name="GPT Researcher 2",
     documentation_url=(
         f"https://github.com/i-am-bee/beeai-platform/blob/{os.getenv('RELEASE_VERSION', 'main')}"
         "/agents/community/gpt-researcher"
     ),
-    detail=AgentDetail(
-        interaction_mode="single-turn",
-        user_greeting="What topic do you want to research?",
-        use_cases=[
-            "**Comprehensive Research** – Generates detailed reports using information from multiple sources.",
-            "**Bias Reduction** – Cross-references data from various platforms to minimize misinformation and bias.",
-            "**High Performance** – Utilizes parallelized processes for efficient and swift report generation.",
-            "**Customizable** – Offers customization options to tailor research for specific domains or tasks.",
-        ],
-    ),
+    detail=AgentDetail(interaction_mode="single-turn", user_greeting="What topic do you want to research?"),
     skills=[
         AgentSkill(
             id="deep_research",
@@ -59,37 +59,58 @@
     ],
 )
 async def gpt_researcher(
-    message: Message, context: RunContext, trajectory: Annotated[TrajectoryExtensionServer, TrajectoryExtensionSpec()]
+    message: Message,
+    context: RunContext,
+    trajectory: Annotated[TrajectoryExtensionServer, TrajectoryExtensionSpec()],
+    llm_ext: Annotated[LLMServiceExtensionServer, LLMServiceExtensionSpec.single_demand()],
+    embedding_ext: Annotated[EmbeddingServiceExtensionServer, EmbeddingServiceExtensionSpec.single_demand()],
 ) -> AsyncGenerator[RunYield, None]:
     """
     The agent conducts in-depth local and web research using a language model to generate comprehensive reports with
     citations, aimed at delivering factual, unbiased information.
     """
-    os.environ["RETRIEVER"] = "duckduckgo"
-    os.environ["OPENAI_BASE_URL"] = os.getenv("LLM_API_BASE", "http://localhost:11434/v1")
-    os.environ["OPENAI_API_KEY"] = os.getenv("LLM_API_KEY", "dummy")
-    model = os.getenv("LLM_MODEL", "llama3.1")
-    os.environ["LLM_MODEL"] = model
-
-    class CustomLogsHandler:
-        async def send_json(self, data: dict[str, Any]) -> None:
-            if "output" not in data:
-                return
-            match data.get("type"):
-                case "logs":
-                    await context.yield_async(
-                        trajectory.trajectory_metadata(title="log", content=f"{data['output']}\n")
-                    )
-                case "report":
-                    await context.yield_async(data["output"])
-
-    if not message.parts or not (query := message.parts[-1].root.text):
-        yield "Please enter a topic or query."
-        return
-
-    researcher = GPTResearcher(query=query, report_type="research_report", websocket=CustomLogsHandler())
-    await researcher.conduct_research()
-    await researcher.write_report()
+    # Set up local environment for this request
+
+    llm_conf, embedding_conf = None, None
+    if llm_ext and llm_ext.data:
+        [llm_conf] = llm_ext.data.llm_fulfillments.values()
+
+    if embedding_ext and embedding_ext.data:
+        [embedding_conf] = embedding_ext.data.embedding_fulfillments.values()
+
+    model = llm_conf.api_model if llm_conf else os.getenv("LLM_MODEL", "dummy")
+    embedding_model = embedding_conf.api_model if embedding_conf else os.getenv("EMBEDDING_MODEL", "dummy")
+
+    env = {
+        "RETRIEVER": "duckduckgo",
+        "OPENAI_BASE_URL": llm_conf.api_base if llm_conf else os.getenv("LLM_API_BASE", "http://localhost:11434/v1"),
+        "OPENAI_API_KEY": llm_conf.api_key if llm_conf else os.getenv("LLM_API_KEY", "dummy"),
+        "LLM_MODEL": model,
+        "EMBEDDING": f"openai:{embedding_model}",
+        "FAST_LLM": f"openai:{model}",
+        "SMART_LLM": f"openai:{model}",
+    }
+    with with_local_env(env):
+
+        class CustomLogsHandler:
+            async def send_json(self, data: dict[str, Any]) -> None:
+                if "output" not in data:
+                    return
+                match data.get("type"):
+                    case "logs":
+                        await context.yield_async(
+                            trajectory.trajectory_metadata(title="log", content=f"{data['output']}\n")
+                        )
+                    case "report":
+                        await context.yield_async(data["output"])
+
+        if not message.parts or not (query := message.parts[-1].root.text):
+            yield "Please enter a topic or query."
+            return
+
+        researcher = GPTResearcher(query=query, report_type="research_report", websocket=CustomLogsHandler())
+        await researcher.conduct_research()
+        await researcher.write_report()
 
 
 def run():

agents/community/gpt-researcher/gpt_researcher_agent/agent.py

@@ -0,0 +1,92 @@
+# Copyright 2025 © BeeAI a Series of LF Projects, LLC
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+import contextvars
+from collections import UserDict
+from contextlib import contextmanager
+from typing import Dict
+
+
+# Store original environ and functions
+_original_environ = os.environ
+_original_getenv = os.getenv
+
+# Context variable for environment overrides
+_env_context: contextvars.ContextVar[Dict[str, str]] = contextvars.ContextVar("env_context", default={})
+
+
+def context_aware_getenv(key: str, default=None):
+    """Context-aware replacement for os.getenv"""
+    context_env = _env_context.get({})
+    if key in context_env:
+        return context_env[key]
+    return _original_getenv(key, default)
+
+
+class ContextAwareEnviron(UserDict):
+    """Context-aware environment dictionary that checks context variables first"""
+
+    def __init__(self, original_environ):
+        # Don't call super().__init__() - we'll handle data ourselves
+        self._original = original_environ
+
+    @property
+    def data(self):
+        """Return merged environment (context overrides original)"""
+        result = self._original.copy()
+        context_env = _env_context.get({})
+        result.update(context_env)
+        return result
+
+    def __getitem__(self, key):
+        context_env = _env_context.get({})
+        if key in context_env:
+            return context_env[key]
+        return self._original[key]
+
+    def __setitem__(self, key, value):
+        context_env = _env_context.get({}).copy()
+        context_env[key] = value
+        _env_context.set(context_env)
+
+    def __delitem__(self, key):
+        context_env = _env_context.get({}).copy()
+        if key in context_env:
+            del context_env[key]
+            _env_context.set(context_env)
+        elif key in self._original:
+            del self._original[key]
+
+    def __contains__(self, key):
+        context_env = _env_context.get({})
+        return key in context_env or key in self._original
+
+
+def set_context_env(env_vars: Dict[str, str]):
+    """Set environment variables for current context"""
+    current_env = _env_context.get({}).copy()
+    current_env.update(env_vars)
+    _env_context.set(current_env)
+
+
+@contextmanager
+def with_local_env(env_vars: Dict[str, str]):
+    """Context manager that sets up local environment context for this request"""
+    # Store the current context
+    old_context = _env_context.get({})
+
+    # Set up new context with provided env vars
+    set_context_env(env_vars)
+
+    try:
+        yield
+    finally:
+        # Restore the previous context
+        _env_context.set(old_context)
+
+
+def patch_os_environ():
+    """Apply the context-aware patches to os.environ and os.getenv"""
+    os.environ = ContextAwareEnviron(_original_environ)
+    os.getenv = context_aware_getenv

agents/community/gpt-researcher/gpt_researcher_agent/env_patch.py

@@ -29,3 +29,7 @@ line-length = 120
 
 [tool.uv.sources]
 beeai-sdk = { path = "../../../apps/beeai-sdk", editable = true }
+
+[tool.pyright]
+venvPath = "."
+venv = ".venv"