The AI Toolchain — issue -285, November 5, 2025

Agno v2.2.7 adds run_context access in tools/hooks, RedisVL vector DB support, vLLM embeddings, and MCP tool name prefixing.

$ git clone --branch v2.2.7 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v2.2.7

from agno.agent import RunContext

def my_tool(query: str, run_context: RunContext) -> str:
    previous = run_context.state.get("last_query", "none")
    run_context.state["last_query"] = query
    return f"Previous query was: {previous}"

• ›Adds tool_name_prefix parameter to the MCPTools class to namespace all tool names for a given MCP server, preventing collisions when multiple MCP servers are in use.
• ›Introduces run_context as an injectable parameter in tools, hooks, tool hooks, dependency functions, and instructions functions for unified access to run-level state.
• ›Adds RedisVL as a supported VectorDB backend for Knowledge, enabling Redis-backed vector search.
• ›Adds vLLM embedder support, allowing local or remote vLLM models to be used for embeddings in Knowledge pipelines.

OpenAI Agents SDK v0.5.0 adds SIP protocol support for RealtimeRunner, per-request usage tracking, and Dapr session storage.

$ git clone --branch v0.5.0 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.5.0

• ›Adds SIP protocol connection support to RealtimeRunner, enabling realtime agents to handle SIP-based voice calls.
• ›Adds a list of per-request usage data to the Usage object, giving finer-grained token consumption tracking across multi-step runs.
• ›Adds Dapr as a session storage option for agent runs.
• ›Adds Python 3.14 to the list of officially supported versions.

PydanticAI v1.11.0 adds runtime instructions to agent.run() and partial_output access in output validators.

$ git clone --branch v1.11.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.11.0

result = await agent.run(
    'Summarize this document',
    instructions='Always respond in formal English and limit output to 3 sentences.'
)

from pydantic_ai import RunContext

@agent.output_validator
async def check_output(ctx: RunContext, value: MyOutput) -> MyOutput:
    if ctx.partial_output is not None:
        # inspect intermediate state before full validation
        print('Partial so far:', ctx.partial_output)
    return value

• ›Adds instructions parameter to agent.run(), allowing additional instructions to be injected at call time without reconfiguring the agent.
• ›Adds partial_output field to RunContext supplied to output validators, exposing the partially-constructed output during validation.

Glamourous agentic coding for all

Crush v0.15.0 adds an agentic fetch tool for reasoning over URL content plus keyboard shortcut UX improvements.

$ git clone --branch v0.15.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.15.0

• ›Adds an agentic fetch tool that lets subagents pull content from a URL, reason over it, and iterate until they find what they need.
• ›Press / in an empty textarea to open the command palette (in addition to the existing ctrl+p shortcut).
• ›Use @ to open the file picker in the chat input.

The open source coding agent.

OpenCode v1.0.27 adds file change counts to session summaries.

$ git clone --branch v1.0.27 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.27

• ›Adds file count to session summary, displaying the number of files changed alongside additions and deletions.

$ git clone --branch v1.0.26 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.26

$ opencode run --attach

$ git clone --branch v1.0.25 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.25

$ /export

$ Summarize this code and then ask @reviewer to check it for security issues

$ git clone --branch v1.0.24 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.24

OpenHands: AI-Driven Development

OpenHands 0.61.0 makes the terminal tab read-only and removes the Jupyter tab from the UI.

$ git clone --branch 0.61.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:
$ git checkout 0.61.0

• ›Terminal tab is now read-only, preventing direct user input into the agent's terminal session.
• ›Jupyter tab has been removed from the UI.

Qwen Code v0.1.4 adds configurable OpenAI logging directory support.

$ git clone --branch v0.1.4 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.1.4

• ›Adds support for custom OpenAI logging directory configuration, letting users control where OpenAI API interaction logs are stored.

Zed v0.211.4 adds Windows Arm64 builds, git: branch diff, ACP agent installs via extensions, and inline debug variables for JS/TS/TSX.

$ git clone --branch v0.211.4 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.211.4

"edit_predictions": {
  "codestral": {
    "api_url": "https://your-codestral-proxy.example.com",
    "model": "codestral-latest",
    "max_tokens": 150
  }
}

{
  "relative_line_numbers": "wrapped"
}

• ›Adds edit_predictions.codestral.api_url config key (plus model and max_tokens) to point Edit Prediction at a custom Codestral endpoint.
• ›Changes relative_line_numbers from a boolean to an enum (enabled, disabled, or wrapped) to support counting wrapped lines as relative lines and displaying line numbers for wrapped segments.
• ›Adds prevent_sharing_in_public_channels project setting to prevent projects from being shared in public Zed channels.
• ›Adds git: branch diff command palette action to diff the current branch against main.
• ›Adds editor: go to next reference and editor: go to previous reference editor actions.

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

gemini-cli gains hook I/O contracts, alternate buffer UI, thinking budget cap, WriteTodos tool, and subagent timeout/recovery.

$ git clone --branch v0.13.0-nightly.20251105.da3da198 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.13.0-nightly.20251105.da3da198

• ›Adds hook type decoupling and translation, enabling hooks to be defined independently of their execution context.
• ›Adds hook input/output contracts, formalizing the data shapes passed to and returned from hooks.
• ›Adds hook execution planning and matching, giving the runtime a structured way to select and schedule which hooks fire.
• ›Enables alternate terminal buffer by default, keeping the shell history clean after gemini-cli exits.
• ›Grants subagents a recovery turn when they hit TIMEOUT, MAX_TURNS, or NO_TOOL_CALL failures.

$ git clone --branch v0.13.0-preview.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.13.0-preview.0

Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

Ollama v0.12.10 adds embedding model support in ollama run, tool call IDs in the chat API, and Vulkan GPU improvements.

$ git clone --branch v0.12.10 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.12.10

$ ollama run embeddinggemma "Hello world"

$ echo "Hello world" | ollama run embeddinggemma

• ›Enables ollama run to generate vector embeddings by running embedding models directly from the CLI, including via stdin piping.
• ›Returns tool call IDs from the /api/chat API endpoint.
• ›Enables flash attention for Vulkan GPU backends.
• ›Adds Vulkan memory detection for Intel GPUs using DXGI+PDH.
• ›Shows login instructions when switching to a cloud model in interactive mode.

Arize Phoenix client v1.22.0 adds example binding to evaluators, prompt metadata, dataset split querying, and experiment resumption.

$ git clone --branch arize-phoenix-client-v1.22.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-client-v1.22.0

• ›Adds ability to query dataset examples by splits, enabling filtered retrieval of training/test/validation subsets.
• ›Adds metadata support for prompts, allowing structured metadata to be attached to prompt objects.
• ›Enables binding examples directly to evaluators for more tightly coupled evaluation workflows.
• ›Adds resume capability for experiments and evaluations, allowing interrupted runs to continue from where they left off.

$ git clone --branch arize-phoenix-v12.10.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v12.10.0

auth:
  secret:
    valueFrom:
      secretKeyRef:
        name: phoenix-auth-secret
        key: secret-key

additionalEnv:
  - name: OPENAI_API_KEY
    valueFrom:
      secretKeyRef:
        name: openai-secret
        key: api-key

Chroma 1.3.3 adds a BM25 sparse embedding function to the Python client.

$ git clone --branch 1.3.3 https://github.com/chroma-core/chroma.git
# already have the repo? check out this version:
$ git checkout 1.3.3

• ›Adds chroma_bm25 embedding function to the Python library, enabling sparse BM25-based retrieval alongside existing dense embedding functions.

Weaviate v1.34.0 adds SPFresh vector index, Flat Index rotational quantization, server-side dynamic batching, and Contextual AI modules.

$ git clone --branch v1.34.0 https://github.com/weaviate/weaviate.git
# already have the repo? check out this version:
$ git checkout v1.34.0

• ›Adds spfresh as a new vector index type, supporting custom configs, HNSW centroid indexing, uncompressed vectors, a disk queue, and dynamic posting-size calculation.
• ›Adds rotational quantization (1-bit and 8-bit RQ) support to the Flat Index, with usage metrics for Flat RQ index.
• ›Adds server-side dynamic batching (beta) to reduce client-side batching complexity.
• ›Adds Contextual AI Generative and Reranker module integration.
• ›Adds support for Amazon Nova Multimodal Embeddings model in the embeddings module.