Backfill issue -319 — The AI Toolchain

EDITIONS tail grep head diff uniq

The daily firehose — everything the toolchain shipped today, already filtered.

// HOW THIS ISSUE IS MADE

We read every release from the 174 tools on our watchlist at the source — GitHub and GitLab release notes, vendor release pages and changelogs, project blogs and feeds, vendor press releases, and the source code behind the tag. Bug-fix-only releases and non-product newsroom noise are dropped; what's left is summarized down to the new capability, how to try it, and any screenshots or videos the release itself published. Every entry links to the sources it was built from.

VIEW

ISSUE VIEW full issue

Do you prefer this view?

$ tct list   # 10 tools matched

AI & LLM Tooling

◆ AI Agent Frameworks

Agno (formerly Phidata)

v2.1.0 NOTES

Agno v2.1.0 adds JWT middleware, pre/post hooks, guardrails (PII/prompt-injection/OpenAI Moderation), and Requesty LLM gateway support.

└──▷ GET THIS VERSION

$ git clone --branch v2.1.0 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v2.1.0

›Adds built-in JWT Middleware to AgentOS for validating tokens and extracting claims into AgentOS endpoints.
›Adds support for any custom FastAPI-compatible middleware on the AgentOS.
›Adds pre_hooks and post_hooks configuration on agents and teams for input/output validation or transformation at the start or end of a run.
›Adds built-in Guardrails as pre-hooks, including prompt injection checks, PII detection (SSN, credit cards, phone numbers, emails), OpenAI Moderation, and support for custom guardrails.
›Adds base_app parameter to AgentOS for configuring a custom FastAPI app (replacing the now-deprecated fastapi_app).

+8 moreshow less

›Adds on_route_conflict parameter to AgentOS for controlling behavior when routes conflict on base_app (replacing deprecated replace_routes).
›Adds enable_mcp_server parameter to AgentOS for converting it into an MCP server (replacing deprecated enable_mcp).
›Adds id parameter to AgentOS as the OS identifier (replacing deprecated os_id).
›Adds support for the Requesty LLM gateway provider for affordable LLM access with advanced governance.
›Adds batch embeddings support to speed up embedding generation and reduce API calls to embedding providers.
›Enables concurrent (parallel) execution of async generator tools during streaming.
›Enables concurrent member execution during async team streaming via Team.arun(..., stream=True); note that event order is no longer guaranteed.
›Uses the user's unique phone number as the default session ID for WhatsApp interface sessions, enabling conversation history.

└──▷ BREAKING ON UPGRADE

!Async team member events during Team.arun(..., stream=True) are now received concurrently — the order of member events is no longer guaranteed. Code depending on a specific event order from team members will behave differently.
!The use_batch parameter has been removed from the PGVectorDB class.

Was this useful?

LangChain

langchain-openai==0.3.34 3 RELEASES · 2025-10-01 NOTES STABLE

langchain-openai 0.3.34 adds OpenAI SDK 2.0 support, PDF inputs in ToolMessages, and max_tokens for AzureChatOpenAI.

└──▷ GET THIS VERSION

$ git clone --branch langchain-openai==0.3.34 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-openai==0.3.34

└──▷ USE IT

Cap output length on an Azure-hosted model to control cost and latency.

python

from langchain_openai import AzureChatOpenAI

llm = AzureChatOpenAI(
    azure_deployment="gpt-4o",
    api_version="2024-02-01",
    max_tokens=512,
)
response = llm.invoke("Summarize this document.")
print(response.content)

›Adds max_tokens parameter to AzureChatOpenAI for controlling output token limits.
›Supports OpenAI SDK 2.0 in the langchain-openai integration.
›Supports PDF inputs in ToolMessage objects, enabling multimodal tool responses.
›Allows overriding ls_model_name from kwargs at invocation time.

2 more releases in this issue · 2025-10-01

langchain-tests==0.3.22 NOTES STABLE

langchain-tests 0.3.22 adds PDF input support in ToolMessages and a new property to skip get_by_ids() tests on unsupporting vector stores.

└──▷ GET THIS VERSION

$ git clone --branch langchain-tests==0.3.22 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-tests==0.3.22

›Adds a property to StandardVectorStoreTests to skip relevant tests when a vector store does not support get_by_ids(), preventing false failures in integrations that omit that method.
›Supports PDF inputs in ToolMessage objects, enabling standard tests to cover tool responses that return PDF content.

langchain-core==0.3.77 NOTES STABLE

langchain-core 0.3.77 adds PDF input support in ToolMessages, custom Mermaid URL override, and ls_model_name kwarg override.

└──▷ GET THIS VERSION

$ git clone --branch langchain-core==0.3.77 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==0.3.77

›Allows overriding ls_model_name from kwargs when constructing model calls, enabling per-call model name customization in tracing.
›Adds support for a custom Mermaid URL via allow custom Mermaid URL, letting teams point graph rendering at a self-hosted or alternate Mermaid service.
›Supports PDF inputs in ToolMessage objects, enabling tools to return PDF content directly in the message payload.

Was this useful?

PydanticAI

v1.0.12 NOTES

PydanticAI v1.0.12 adds Anthropic built-in memory tool support, OpenAI document URL/binary content for text/JSON/XML/YAML, and evals cost metrics.

└──▷ GET THIS VERSION

$ git clone --branch v1.0.12 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.0.12

›Adds retry args to pydantic_evals.Dataset.evaluate_sync for configurable retry behavior in evaluation runs.
›Adds cost metric to pydantic-evals output, giving visibility into token spend per evaluation.
›Supports Anthropic's built-in memory tool, enabling agents to persist and recall information across turns via the provider-native mechanism.
›Supports text, JSON, XML, and YAML DocumentUrl and BinaryContent on OpenAI, expanding the range of document types agents can process.
›Prefers structuredContent in MCP tool results when present, enabling richer structured data from MCP tool calls.

+4 moreshow less

›Exposes .messages and .toolsets types in the top-level pydantic_ai namespace to improve IDE auto-import discovery.
›Broadens the type of common_tools to work with agents of any deps type, removing a previous type-narrowing restriction.
›Handles Gemini responses with more than one candidate without raising an error.
›Handles Ollama responses that omit finish_reason and adds documentation for Ollama Cloud.

Was this useful?

◆ AI Coding Agents

Continue

@continuedev/[email protected] 3 RELEASES · 2025-10-01 NOTES STABLE

Continue 1.21.0 adds AWS credential onboarding, remote agent tunneling, environment.json support, and a new cn status tool.

└──▷ GET THIS VERSION

$ git clone --branch @continuedev/[email protected] https://github.com/continuedev/continue.git
# already have the repo? check out this version:
$ git checkout @continuedev/[email protected]

└──▷ USE IT

Define environment variables for a Continue workspace so cn serve picks them up automatically.

json

{
  "MY_API_KEY": "abc123",
  "ANOTHER_VAR": "value"
}

›Adds --id option to the CLI to connect to an existing remote agent via tunnel.
›Adds .continue/environment.json file support for environment variable configuration, also handled by cn serve.
›Adds status tool for cn to inspect agent state.
›Enhances the /init command to create a custom slash command.
›Adds AWS credentials onboarding flow to the CLI.

+5 moreshow less

›Supports Claude Code MCP JSON config format, with YAML-to-JSON and JSON-to-YAML conversion for MCP configs.
›MCP loading errors are now surfaced in config loading errors.
›Adds new Grok model options.
›Adds UI for global rule creation.
›Dangerous command warnings are now shown only once per session in the CLI.

2 more releases in this issue · 2025-10-01

@continuedev/[email protected] NOTES STABLE

Continue 1.24.0 adds a cn status command, UI for global rule creation, and MCP loading errors surfaced in config diagnostics.

└──▷ GET THIS VERSION

$ git clone --branch @continuedev/[email protected] https://github.com/continuedev/continue.git
# already have the repo? check out this version:
$ git checkout @continuedev/[email protected]

└──▷ TRY IT

Quickly inspect whether the cn agent has tool access configured correctly before running a pipeline.

$ cn status

›Adds status subcommand to the cn CLI for inspecting tool state.
›Surfaces MCP loading errors inside config loading error reporting, making misconfigured MCP servers visible during diagnostics.
›Adds UI for creating global rules directly from the settings page.
›Adds tool call argument preprocessing to normalize inputs before dispatch.
›Hides empty resources and prompts sections from the UI when none are configured.

v1.2.7-vscode NOTES STABLE

Continue v1.2.7 adds CLI auto-updates, Lemonade/CometAPI providers, AGENTS.md support, AWS Bedrock tool streaming, and fuzzy file search.

└──▷ GET THIS VERSION

$ git clone --branch v1.2.7-vscode https://github.com/continuedev/continue.git
# already have the repo? check out this version:
$ git checkout v1.2.7-vscode

└──▷ TRY IT

Serve a Continue session with a specific identifier so remote clients can discover it by name.

$ cn serve --id my-session

List all active remote Continue sessions to see which named sessions are available.

$ cn ls

›Adds --id flag to cn serve for naming/identifying a served session.
›Adds cn ls command to display remote sessions.
›Adds grep and glob ignore support for VS Code and JetBrains context search.
›Adds invokable option to markdown rule creation, enabling rules to be triggered as slash commands.
›Supports ~ when resolving uses blocks in config.

+21 moreshow less

›Adds headless mode that surfaces errors and exits with code 1 on failure.
›Adds Lemonade as a new model provider integration.
›Adds CometAPI as a new model provider.
›Adds Mercury Apply endpoint support to Inception LLM provider.
›Adds AWS Bedrock provider tool streaming support.
›Adds Anthropic prompt caching support to the OpenRouter LLM implementation.
›Adds CLI auto-update capability.
›Supports AGENTS.md for agent-level instructions.
›Revamps file search with fuzzy matching support.
›Exposes legacy /share command to the VS Code UI.
›Adds mark unread capability for chat sessions.
›Adds MCP server connection status display with richer status information.
›Adds keyboard shortcuts display for accept/reject diff buttons.
›Adds in-UI installation guides for MCP commands.
›Adds next-edit character-level diff capability.
›Adds search-and-replace error codes for clearer failure reporting.
›Adds grok model updates with new model entries.
›Adds fullscreen command back to the UI.
›Splits large blocks of chat history rendering to reduce UI flickering.
›Adds more robust context length validation before LLM API calls.
›Moves tool icon names into tool definitions for consistent rendering.

Was this useful?

GitHub Copilot CLI

v0.0.331 NOTES

GitHub Copilot CLI v0.0.331 filters inaccessible models from /model list and adds a scrollbar to the @ file picker.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.331 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:
$ git checkout v0.0.331

›The /model list now filters out models the user cannot access (due to Copilot plan or geographic region), preventing model_not_supported errors from appearing at runtime.
›Adds a scrollbar to the @ file mentioning picker and increases the active buffer size to 10 items.
›Up/down arrow keys now correctly navigate / and @ menu options while the agent is running.

Was this useful?

SST OpenCode

v0.13.6 NOTES

The open source coding agent.

OpenCode v0.13.6 adds directory support to @ mentions and overhauled file search.

└──▷ GET THIS VERSION

$ git clone --branch v0.13.6 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.13.6

›Supports @ mentioning directories in addition to files for faster context inclusion.
›Overhauled file search for improved discovery and usability.

Was this useful?

All Hands AI OpenHands

0.58.0 NOTES

OpenHands: AI-Driven Development

OpenHands 0.58.0 adds configurable SHTTP MCP server timeouts and claude-sonnet-4-5 model support.

└──▷ GET THIS VERSION

$ git clone --branch 0.58.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:
$ git checkout 0.58.0

›Adds configurable timeout settings for SHTTP MCP servers, preventing tool calls from hanging indefinitely.
›Adds support for the claude-sonnet-4-5 model.
›Converts the settings panel to a vertical sidebar layout.

Was this useful?

Google gemini-cli

v0.9.0-nightly.20251001.14dbda91 3 RELEASES · 2025-10-01 NOTES STABLE

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

gemini-cli v0.9.0-nightly adds regex search/replace editing, subagent support, and OpenTelemetry GenAI metrics.

└──▷ GET THIS VERSION

$ git clone --branch v0.9.0-nightly.20251001.14dbda91 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.9.0-nightly.20251001.14dbda91

└──▷ USE IT

Enable subagents in your gemini-cli config to allow the CLI to delegate tasks to registered sub-agents.

json

# In your gemini-cli settings (e.g., settings.json)
{ "enableSubagents": true }

›Adds regex search/replace support to the Smart Edit tool for more precise in-file editing.
›Adds enableSubagents configuration option to enable and register subagents.
›Introduces a Declarative Agent Framework for defining agents declaratively.
›Adds OpenTelemetry GenAI semantic convention metrics for standardized observability of LLM usage.
›Adds Smart Edit Strategy Logging to surface which edit strategy was selected.

└──▷ BREAKING ON UPGRADE

!The model router is now disabled by default (--routing / model router behavior changes for existing setups relying on automatic model routing).

2 more releases in this issue · 2025-10-01

v0.8.0-preview.1 NOTES STABLE

gemini-cli v0.8.0-preview.1 adds model router, IAP support, accessibility improvements, tool output truncation, and smarter context compression.

└──▷ GET THIS VERSION

$ git clone --branch v0.8.0-preview.1 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.8.0-preview.1

›Enables model router by default and exposes it in the settings dialog.
›Adds service account impersonation provider to MCP servers to support Identity-Aware Proxy (IAP) on Cloud Run.
›Implements centralized screen reader layout for accessibility.
›Enables tool output truncation by default to manage context window usage.
›Adds content-based retries for JSON generation to handle malformed model responses.

+17 moreshow less

›Uses lastPromptTokenCount to more accurately determine when context compression is needed.
›Adds a setting to display Gemini status or model thoughts in the terminal title and taskbar icon.
›Adds AbortSignal support for retry logic and tool execution, enabling proper cancellation.
›Adds name auto-complete to the /extensions update command.
›Adds an enable/disable indicator to the extensions list.
›Re-requests user consent when updating extensions that require it.
›Shows a confirmation dialog before applying extension updates.
›Throws an error for invalid extension names to surface misconfiguration early.
›Escapes ANSI control codes from model output before rendering to the terminal.
›Adds character-level edit metrics to telemetry.
›Sandbox images are now pushed to Docker Hub instead of GHCR.
›Supports standard GitHub release archive format for distribution.
›Uses dynamic client registration endpoint from config (when available) instead of re-performing OAuth discovery.
›Relaxes JSON schema validation for broader MCP tool compatibility.
›Improves compression algorithm aggressiveness to reduce context size.
›Removes border from user messages and colors > prompt for improved UI readability.
›Reduces margin on narrow screens and reflows footer contents for better layout.

v0.7.0 NOTES STABLE

gemini-cli v0.7.0 adds auto-updating git extensions, policy engine config, global telemetry/debug/proxy flags, and custom commands in headless mode.

└──▷ GET THIS VERSION

$ git clone --branch v0.7.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.7.0

└──▷ TRY IT

Restrict which tools Gemini may call using a comma-separated list in one flag.

$ gemini --allowed-tools read_file,run_shell_command 'Review the Dockerfile and explain any risks'

›Adds auto-updating support for git-sourced extensions, keeping installed extensions current without manual intervention.
›Enables policy engine configuration directly from existing settings.
›Promotes --telemetry, --debug, and --proxy flags to global scope, making them available across all subcommands.
›Supports custom commands in non-interactive/headless mode, enabling scripted workflows.
›Emits a warning when memory usage exceeds 7 GB.

+6 moreshow less

›Adds extensions release flow support for simpler extension publishing.
›Includes MCP Server name in ToolCallEvent OpenTelemetry logging for richer observability.
›Adds CORS and host validation hardening to the VS Code IDE companion server.
›Supports rendering function calls and responses when sharing chat to markdown.
›Supports comma-separated values for --allowed-tools.
›Adds activity detector with user interaction tracking to telemetry.

└──▷ BREAKING ON UPGRADE

!The shell setting usePty is renamed to enableInteractiveShell; existing configs referencing usePty will stop working.

Was this useful?

Other / Uncategorized

◆ AI OBSERVABILITY

Arize Phoenix

arize-phoenix-evals-v2.3.0 NOTES

Phoenix Evals v2.3.0 adds dot-delimited f-string key support in eval Templates.

└──▷ GET THIS VERSION

$ git clone --branch arize-phoenix-evals-v2.3.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v2.3.0

›Allows dot-delimited f-string keys in eval Templates, enabling nested attribute references inside template strings.

Was this useful?

Langfuse

v3.114.0 NOTES

Langfuse v3.114.0 adds Keycloak auth support and natural language filters in the trace UI.

└──▷ GET THIS VERSION

$ git clone --branch v3.114.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.114.0

›Adds natural language filters to the trace UI, letting practitioners query traces in plain English.
›Adds support for Keycloak scope and ID token configuration for OIDC auth.
›Redesigns dataset item detail view to show item I/O and past runs in two separate tabs.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · October 1, 2025
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -319, October 1, 2025