Update to latest agent framework

Author: pamelafox

.github/skills/pdf-to-markdown/SKILL.md

@@ -0,0 +1,51 @@
+---
+name: pdf-to-markdown
+description: Converts PDF files to Markdown using Microsoft's markitdown package. Use this skill when the user asks to convert a PDF to Markdown, extract text from a PDF, or read/parse PDF content.
+argument-hint: "[path to file]"
+---
+
+# PDF to Markdown Conversion
+
+This skill converts PDF files to Markdown format using [Microsoft's markitdown](https://github.com/microsoft/markitdown) package.
+
+## When to use
+
+- User asks to convert a PDF to Markdown
+- User wants to extract text content from a PDF
+- User needs to read or parse a PDF document
+- User asks to summarize or analyze a PDF file
+
+## How to use
+
+Use `uvx` to run markitdown directly. Pick the dependency group matching the file type:
+
+| File type | Dependency group |
+|-----------|-----------------|
+| PDF       | `pdf`           |
+| PowerPoint | `pptx`         |
+| Word      | `docx`          |
+| Excel (.xlsx) | `xlsx`      |
+| Excel (.xls)  | `xls`       |
+
+```bash
+uvx 'markitdown[pdf]' <path-to-file> -o output.md
+```
+
+Or install all optional dependencies at once:
+
+```bash
+uvx 'markitdown[all]' <path-to-file> -o output.md
+```
+
+## Examples
+
+```bash
+uvx 'markitdown[pdf]' report.pdf -o report.md
+uvx 'markitdown[pptx]' slides.pptx -o slides.md
+uvx 'markitdown[docx]' document.docx -o document.md
+```
+
+## Output
+
+- If you were asked to save the output to a specific file, save it to the requested file using `-o`.
+- If no output file was specified, use the source filename with a `.md` suffix.

examples/agent_basic.py

@@ -1,7 +1,7 @@
 import asyncio
 import os
 
-from agent_framework import ChatAgent
+from agent_framework import Agent
 from agent_framework.openai import OpenAIChatClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
@@ -31,7 +31,7 @@
         api_key=os.environ["OPENAI_API_KEY"], model_id=os.environ.get("OPENAI_MODEL", "gpt-5-mini")
     )
 
-agent = ChatAgent(chat_client=client, instructions="You're an informational agent. Answer questions cheerfully.")
+agent = Agent(client=client, instructions="You're an informational agent. Answer questions cheerfully.")
 
 
 async def main():

examples/agent_mcp_local.py

@@ -3,7 +3,7 @@
 import os
 from datetime import datetime
 
-from agent_framework import ChatAgent, MCPStreamableHTTPTool
+from agent_framework import Agent, MCPStreamableHTTPTool
 from agent_framework.openai import OpenAIChatClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
@@ -47,8 +47,8 @@ async def main() -> None:
     """Run an agent connected to a local MCP server for expense logging."""
     async with (
         MCPStreamableHTTPTool(name="Expenses MCP Server", url=MCP_SERVER_URL) as mcp_server,
-        ChatAgent(
-            chat_client=client,
+        Agent(
+            client=client,
             instructions=(
                 "You help users with tasks using the available tools. "
                 f"Today's date is {datetime.now().strftime('%Y-%m-%d')}."

examples/agent_mcp_remote.py

@@ -2,7 +2,7 @@
 import logging
 import os
 
-from agent_framework import ChatAgent, MCPStreamableHTTPTool
+from agent_framework import Agent, MCPStreamableHTTPTool
 from agent_framework.openai import OpenAIChatClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
@@ -48,8 +48,8 @@ async def main() -> None:
             name="Microsoft Learn MCP",
             url="https://learn.microsoft.com/api/mcp",
         ) as mcp_server,
-        ChatAgent(
-            chat_client=client,
+        Agent(
+            client=client,
             instructions=(
                 "You help with Microsoft documentation questions. "
                 "Use the available tools to search for relevant docs."

examples/agent_memory_mem0.py

@@ -3,11 +3,10 @@
 import os
 import random
 import uuid
-from collections.abc import MutableSequence
-from typing import Annotated, Any
+from typing import Annotated
 
-from agent_framework import ChatAgent, ChatMessage, Context, tool
-from agent_framework.mem0 import Mem0Provider
+from agent_framework import Agent, tool
+from agent_framework.mem0 import Mem0ContextProvider
 from agent_framework.openai import OpenAIChatClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
@@ -49,7 +48,7 @@
 
 # NOTE: approval_mode="never_require" is for sample brevity.
 # Use "always_require" in production.
-@tool(approval_mode="never_require")
+@tool
 def get_weather(
     city: Annotated[str, Field(description="The city to get the weather for.")],
 ) -> str:
@@ -59,53 +58,12 @@ def get_weather(
     return f"The weather in {city} is {conditions[random.randint(0, 3)]} with a high of {random.randint(10, 30)}°C."
 
 
-class Mem0OSSProvider(Mem0Provider):
-    """Mem0Provider subclass that fixes search for Mem0 OSS (AsyncMemory).
-
-    The base Mem0Provider passes user_id/agent_id inside a 'filters' dict,
-    which works for the Mem0 Platform client but not for AsyncMemory (OSS),
-    which expects them as direct keyword arguments.
-    """
-
-    async def invoking(self, messages: ChatMessage | MutableSequence[ChatMessage], **kwargs: Any) -> Context:
-        self._validate_filters()
-        messages_list = [messages] if isinstance(messages, ChatMessage) else list(messages)
-        input_text = "\n".join(msg.text for msg in messages_list if msg and msg.text and msg.text.strip())
-
-        if not input_text.strip():
-            return Context(messages=None)
-
-        # Pass user_id/agent_id as keyword args (Mem0 OSS API)
-        search_kwargs: dict[str, Any] = {"query": input_text}
-        if self.user_id:
-            search_kwargs["user_id"] = self.user_id
-        if self.agent_id:
-            search_kwargs["agent_id"] = self.agent_id
-
-        search_response = await self.mem0_client.search(**search_kwargs)
-
-        if isinstance(search_response, list):
-            memories = search_response
-        elif isinstance(search_response, dict) and "results" in search_response:
-            memories = search_response["results"]
-        else:
-            memories = [search_response]
-
-        line_separated_memories = "\n".join(memory.get("memory", "") for memory in memories)
-
-        return Context(
-            messages=[ChatMessage(role="user", text=f"{self.context_prompt}\n{line_separated_memories}")]
-            if line_separated_memories
-            else None
-        )
-
-
 async def main() -> None:
     """Demonstrate an agent with Mem0 OSS for long-term memory.
 
-    Unlike RedisProvider (which stores raw messages), Mem0 uses an LLM to
+    Unlike RedisContextProvider (which stores raw messages), Mem0 uses an LLM to
     extract and store distilled facts from conversations. When the agent
-    starts a new thread, Mem0 injects relevant memories as context.
+    starts a new session, Mem0 injects relevant memories as context.
 
     Mem0 OSS needs an LLM and embedder for memory extraction. This example
     uses Azure OpenAI when API_HOST=azure, otherwise falls back to OPENAI_API_KEY.
@@ -161,16 +119,16 @@ async def main() -> None:
 
     mem0_client = await AsyncMemory.from_config(mem0_config)
 
-    provider = Mem0OSSProvider(user_id=user_id, mem0_client=mem0_client)
+    provider = Mem0ContextProvider(source_id="mem0_memory", user_id=user_id, mem0_client=mem0_client)
 
-    agent = ChatAgent(
-        chat_client=client,
+    agent = Agent(
+        client=client,
         instructions=(
             "You are a helpful weather assistant. Personalize replies using provided context. "
             "Before answering, always check for stored context."
         ),
         tools=[get_weather],
-        context_provider=provider,
+        context_providers=[provider],
     )
 
     # Step 1: Teach the agent user preferences
@@ -179,12 +137,12 @@ async def main() -> None:
     response = await agent.run("Remember that my favorite city is Tokyo and I prefer Celsius.")
     print(f"[green]Agent:[/green] {response.text}")
 
-    # Step 2: Start a new thread — Mem0 should inject remembered facts
-    print("\n[dim]--- Step 2: New thread — recalling preferences ---[/dim]")
+    # Step 2: Start a new session — Mem0 should inject remembered facts
+    print("\n[dim]--- Step 2: New session — recalling preferences ---[/dim]")
     print("[blue]User:[/blue] What's my favorite city?")
     response = await agent.run("What's my favorite city?")
     print(f"[green]Agent:[/green] {response.text}")
-    print("[dim]Note: Mem0 extracted and stored facts, then injected them into the new thread.[/dim]")
+    print("[dim]Note: Mem0 extracted and stored facts, then injected them into the new session.[/dim]")
 
     # Step 3: Use a tool, demonstrating memory with tool outputs
     print("\n[dim]--- Step 3: Tool use with memory ---[/dim]")

examples/agent_memory_redis.py

@@ -5,9 +5,9 @@
 import uuid
 from typing import Annotated
 
-from agent_framework import ChatAgent, tool
+from agent_framework import Agent, tool
 from agent_framework.openai import OpenAIChatClient
-from agent_framework.redis import RedisProvider
+from agent_framework.redis import RedisContextProvider
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
 from pydantic import Field
@@ -48,7 +48,7 @@
 
 # NOTE: approval_mode="never_require" is for sample brevity.
 # Use "always_require" in production.
-@tool(approval_mode="never_require")
+@tool
 def get_weather(
     city: Annotated[str, Field(description="The city to get the weather for.")],
 ) -> str:
@@ -59,9 +59,9 @@ def get_weather(
 
 
 async def example_agent_with_memory() -> None:
-    """Demonstrate an agent with Redis-backed long-term memory via RedisProvider.
+    """Demonstrate an agent with Redis-backed long-term memory via RedisContextProvider.
 
-    The RedisProvider stores conversational context in Redis and retrieves it
+    The RedisContextProvider stores conversational context in Redis and retrieves it
     using full-text search (BM25), or hybrid search (BM25 + vector similarity)
     when an embedding model is configured.
 
@@ -71,10 +71,11 @@ async def example_agent_with_memory() -> None:
 
     user_id = str(uuid.uuid4())
 
-    # RedisProvider supports hybrid search (full-text + vector) when a vectorizer is configured.
+    # RedisContextProvider supports hybrid search (full-text + vector) when a vectorizer is configured.
     # However, there is currently a version mismatch between agent-framework-redis and redisvl
     # (the HybridQuery API changed), so this example uses text-only search for now.
-    memory_provider = RedisProvider(
+    memory_provider = RedisContextProvider(
+        source_id="redis_memory",
         redis_url=REDIS_URL,
         index_name="agent_memory_demo",
         prefix="memory_demo",
@@ -84,14 +85,14 @@ async def example_agent_with_memory() -> None:
         overwrite_index=True,
     )
 
-    agent = ChatAgent(
-        chat_client=client,
+    agent = Agent(
+        client=client,
         instructions=(
             "You are a helpful weather assistant. Personalize replies using provided context. "
             "Before answering, always check for stored context."
         ),
         tools=[get_weather],
-        context_provider=memory_provider,
+        context_providers=[memory_provider],
     )
 
     # Step 1: Teach the agent a user preference