The AI Toolchain — issue -314, October 7, 2025

Agno v2.1.2 adds an A2A interface for AgentOS and a field-labelled CSV reader for structured knowledge ingestion.

$ git clone --branch v2.1.2 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v2.1.2

• ›Adds A2A interface to AgentOS, enabling Agents, Teams, and Workflows to be exposed and run in Agent-to-Agent (A2A) compatible format.
• ›Adds a field-labelled CSV reader that preserves field–value relationships when ingesting CSV knowledge data.
• ›Supports passing user_id in the forwarded_props field when initiating an AG-UI run.
• ›Extends AgentOS API routers to accept additional file types, including more audio formats.
• ›Supports local binary paths (e.g. ./script) as MCP server commands in MCPTools and MultiMCPTools classes.

langchain-core 1.0.0a8 adds PDF tool message support, include_id for OpenAI message conversion, AWS Bedrock document blocks, and several new OpenAI tool types.

$ git clone --branch langchain-core==1.0.0a8 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==1.0.0a8

from langchain_core.messages.utils import convert_to_openai_messages

messages = [HumanMessage(content='Hello', id='msg-1'), AIMessage(content='Hi!', id='msg-2')]
openai_msgs = convert_to_openai_messages(messages, include_id=True)

from langchain_core.utils import sanitize_for_postgres

clean_text = sanitize_for_postgres(llm_output)
vectorstore.add_texts([clean_text])

• ›Adds optional include_id parameter to convert_to_openai_messages function to control whether message IDs are included in the output.
• ›Adds id field to Document objects passed to the filter callback in InMemoryVectorStore similarity search.
• ›Adds web_search to the list of recognized OpenAI built-in tools in core.
• ›Adds image_generation tool to the list of known OpenAI tools.
• ›Adds sanitize_for_postgres utility to strip PostgreSQL NUL bytes that cause DataError.

$ git clone --branch langchain-openai==1.0.0a4 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-openai==1.0.0a4

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o")  # stream_usage enabled by default with default base URL
for chunk in llm.stream("Explain CVE triage in three sentences"):
    if chunk.usage_metadata:
        print(chunk.usage_metadata)

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o")
llm_with_tools = llm.bind_tools(tools=[my_tool], parallel_tool_calls=False)
result = llm_with_tools.invoke("Run the recon steps in order")
print(result.tool_calls)

$ git clone --branch langchain-anthropic==1.0.0a3 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-anthropic==1.0.0a3

from langchain_anthropic import ChatAnthropic

llm = ChatAnthropic(model="claude-3-5-sonnet-20241022", parallel_tool_calls=True)
result = llm.bind_tools([search_tool, calculator_tool]).invoke("What is the weather in Paris and 42 * 7?")

from langchain_anthropic import ChatAnthropic
from langchain_core.messages import SystemMessage, HumanMessage

llm = ChatAnthropic(model="claude-3-5-sonnet-20241022")
messages = [
    SystemMessage(content="You are a helpful assistant.", additional_kwargs={"cache_control": {"type": "ephemeral"}}),
    HumanMessage(content="Summarize the attached document."),
]
result = llm.invoke(messages)

GitHub Copilot CLI v0.0.336 adds proxy support via HTTP_PROXY/HTTPS_PROXY and a persistent screen-reader preference prompt.

$ git clone --branch v0.0.336 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:
$ git checkout v0.0.336

• ›Enables proxy support via HTTPS_PROXY and HTTP_PROXY environment variables regardless of Node version.
• ›Adds a prompt for users who launch with --screen-reader to persistently save the screen-reader preference.
• ›Significantly reduces token consumption, round trips per problem, and time to result.

The open source coding agent.

OpenCode v0.14.6 adds MCP server status monitoring, connection timeouts, and new todo event types.

$ git clone --branch v0.14.6 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.14.6

$ curl http://localhost:<port>/mcp

• ›New GET /mcp endpoint to check real-time connection status of MCP servers.
• ›Adds 5-second timeout to MCP client verification, automatically removing unresponsive clients during initialization.
• ›Adds new todo-related event types to the event system.

$ git clone --branch v0.14.4 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.14.4

An open-source AI agent that brings the power of Gemini directly into your terminal.

gemini-cli v0.8.0 adds IAP/Cloud Run MCP support, model router, screen reader layout, tool output truncation, and more.

$ git clone --branch v0.8.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.8.0

• ›Enables model router by default and exposes it in the settings dialog.
• ›Adds service account impersonation provider to MCPServers to support IAP-protected Cloud Run endpoints.
• ›Enables tool output truncation by default to keep context windows manageable.
• ›Adds content-based retries for JSON generation when the model returns malformed output.
• ›Adds AbortSignal support for retry logic and tool execution, enabling clean cancellation.

$ git clone --branch v0.9.0-preview.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.9.0-preview.0

$ gemini memory list

$ gemini --allowed-tools read_file,shell_command -p "Summarize the security findings in REPORT.md"

Triton v2.61.0 adds static key authentication and repository scoping for OpenAI Frontend APIs.

$ git clone --branch v2.61.0 https://github.com/triton-inference-server/server.git
# already have the repo? check out this version:
$ git checkout v2.61.0

• ›Adds static key authentication support for OpenAI Frontend APIs.

Phoenix Evals 2.5.0 adds a regex evaluator and broader LLM provider support.

$ git clone --branch arize-phoenix-evals-v2.5.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v2.5.0

• ›Adds a regex evaluator for pattern-based output validation without requiring an LLM judge.
• ›Supports more LLM providers through the LLM integration layer.
• ›Improves binding ergonomics for constructing evaluator inputs.