The AI Toolchain — issue -271, November 19, 2025

Cohere Reasoning/Vision/Translate model support and file URL/path loading for agents in openai-adapters 1.31.0

$ 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]

• ›Extends auth header override to support x-api-key in addition to standard authorization headers.
• ›Adds Cohere Reasoning, Vision, and Translate model support via the Cohere adapter.
• ›Supports file paths and file URLs for agent file loading.
• ›Enhances autocomplete cache with async initialization and improved eviction logic.

$ git clone --branch v1.2.11-vscode https://github.com/continuedev/continue.git
# already have the repo? check out this version:
$ git checkout v1.2.11-vscode

$ cn ls

$ 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]

$ 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]

$ 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]

Glamourous agentic coding for all

Crush v0.18.4 adds support for Kimi's coding subscription plan as a model provider.

$ git clone --branch v0.18.4 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.18.4

• ›Adds support for Kimi Coding (kimi.com coding subscription) as a model provider.

$ git clone --branch v0.18.3 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.18.3

an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM

Goose v1.15.0 adds Grok provider support, Supabase MCP, recipe security scanning, and ACP session management.

$ git clone --branch v1.15.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.15.0

• ›Adds Grok as a supported AI provider.
• ›Adds Supabase MCP server to the extension registry.
• ›Adds security scanning of recipes at save time to catch malicious or unsafe content.
• ›Adds SessionManager integration for ACP (Agent Communication Protocol) sessions.
• ›Adds MCP UI proxy with improved message handling for extension interactions.

Lightweight coding agent that runs in your terminal

Codex CLI gains GPT-5.1-Codex-Max, native Windows Agent mode, compaction support, and expanded 10K tool output tokens.

$ git clone --branch rust-v0.59.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.59.0

# In ~/.codex/config.toml
tool_output_token_limit = 10000

• ›Adds GPT-5.1-Codex-Max model for higher reliability and long-running, project-scale agentic coding workflows.
• ›Introduces native Windows Agent mode with an experimental sandbox that limits filesystem and network access, allowing Codex to read/write files and run commands with fewer approvals.
• ›Adds first-class Native Compaction support to improve performance across extended coding sessions, enabled by default.
• ›Expands tool output token limits to 10,000 tokens, configurable via tool_output_token_limit in config.toml.
• ›Adds LM Studio OSS model support.

Qwen Code v0.2.2 adds Trae editor support, ACP slash commands, configurable OpenAPI window, and terminal attention notifications.

$ git clone --branch v0.2.2 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.2.2

• ›Adds slash command support for ACP (Agent Communication Protocol) integration.
• ›Adds support for the Trae editor.
• ›Makes the OpenAPI context window configurable.
• ›Adds support for alternative cached_tokens format in OpenAI-compatible conversations.
• ›Enhances QwenLogger to capture OS platform and version metadata alongside log entries.

$ git clone --branch v0.2.2-nightly.20251119.97bf48b1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.2.2-nightly.20251119.97bf48b1

Zed v0.213.3 adds Gemini 3 Pro, sticky scroll, multi-word snippet prefixes, default_model for agent profiles, and conda venv detection.

$ git clone --branch v0.213.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.213.3

{
  "sticky_scroll": {
    "enabled": true
  }
}

{
  "terminal": {
    "detect_venv": {
      "on": {
        "conda_manager": "mamba"
      }
    }
  }
}

• ›Adds "sticky_scroll": {"enabled": true} setting to pin scope context to the top of the editor.
• ›Adds default_model key to custom agent profile settings to configure a per-profile default model.
• ›Adds project_panel.auto_open.on_create, project_panel.auto_open.on_paste, and project_panel.auto_open.on_drop settings for granular control over when files auto-open in the project panel.
• ›Adds terminal.detect_venv.on.conda_manager setting to configure the conda manager, with options auto (default), conda, mamba, or micromamba.
• ›Adds gdb_path and gdb_args options to gdb debug adapter configuration.

langchain-core 1.0.6 adds proxy support for Mermaid PNG diagram rendering.

$ git clone --branch langchain-core==1.0.6 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==1.0.6

• ›Adds proxy support for Mermaid PNG rendering, enabling diagram generation through an HTTP proxy in restricted network environments.
• ›Supports tool runtime injection when a custom args schema is provided.

PydanticAI v1.20.0 adds Gemini 3 Pro support, metadata fields on model messages, TTL for Anthropic cache, and enhanced Gemini JSON Schema features.

$ git clone --branch v1.20.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.20.0

from pydantic_ai.messages import ModelRequest

# metadata flows through the request/response cycle
request = ModelRequest(parts=[...], metadata={'trace_id': 'abc-123', 'env': 'prod'})
print(request.metadata)  # {'trace_id': 'abc-123', 'env': 'prod'}

• ›Adds ModelRequest.metadata and ModelResponse.metadata fields for attaching arbitrary metadata to model messages.
• ›Adds ttl field to CachePoint and Anthropic caching model settings, enabling control over cache entry lifetime.
• ›Adds support for Gemini 3 Pro via GoogleModel.
• ›Supports Gemini enhanced JSON Schema features when using GoogleModel.
• ›Makes RunContext.usage available in Temporal workflow contexts.

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

Ollama v0.13.0 adds DeepSeek-OCR, Cogito-V2.1, a new bench tool, and DeepSeek-V3.1 architecture support.

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

$ ollama run deepseek-ocr "/path/to/image\n<|grounding|>Convert the document to markdown."

$ go run ./cmd/bench -model gpt-oss:20b

• ›Adds DeepSeek-OCR model support for document OCR, layout-grounded extraction, figure parsing, and markdown conversion from images.
• ›Adds Cogito-V2.1 instruction-tuned model, currently the top open-weight LLM from a US company.
• ›New bench CLI tool for benchmarking local model performance (built from the Ollama repo with Go).
• ›Supports DeepSeek-V3.1 architecture natively in Ollama's engine.

oobabooga text-gen v3.18 adds --cpu-moe flag for VRAM reduction and ROCm portable builds for AMD GPUs on Linux.

$ git clone --branch v3.18 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout v3.18

$ python server.py --cpu-moe

• ›Adds --cpu-moe flag for the llama.cpp backend to offload MoE model experts to CPU, reducing VRAM usage on large mixture-of-experts models.
• ›Adds ROCm portable builds for AMD GPUs on Linux, enabling no-install GGUF inference on AMD hardware.

Phoenix v12.16.0 adds OpenAI 5.1 model support, reasoning for OpenAI and Gemini, and connection pagination on trace spans.

$ git clone --branch arize-phoenix-v12.16.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v12.16.0

• ›Implements connection pagination on trace spans, enabling traversal of large trace datasets.
• ›Adds support for the OpenAI 5.1 model family.
• ›Adds reasoning support for OpenAI and Gemini models, and adds gemini-3 to the supported model list.

LanceDB v0.22.4-beta.2 adds num_attempts to merge-insert results, async to_pydantic, and async namespace connections.

$ git clone --branch v0.22.4-beta.2 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.22.4-beta.2

• ›Adds num_attempts field to merge-insert results, exposing the number of attempts made during a merge-insert operation.
• ›Supports to_pydantic in async Python contexts, enabling Pydantic model conversion in async workflows.
• ›Supports async namespace connections, allowing namespace-scoped operations in async code.

$ git clone --branch python-v0.25.4-beta.2 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.25.4-beta.2

results = await table.search(query_vector).to_pydantic(MyModel)