The AI Toolchain — issue -021, July 29, 2026

AutoGPT Platform v0.6.70 adds configurable transcription endpoints, clipboard image paste, Expert entity with hire/install API, and 'Most Recently Ran' library sorting.

$ git clone --branch autogpt-platform-beta-v0.6.70 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.70

• ›Adds hire/install API and roster seeds for a new Expert entity, enabling programmatic management of expert agents.
• ›Adds configurable transcription endpoints in CoPilot, allowing operators to point speech-to-text at custom or self-hosted services.
• ›Adds 'Most Recently Ran' sorting option in the Library for faster access to active agents.
• ›Enables pasting images directly from the clipboard into CoPilot chat.
• ›Surfaces skipped webhook presets when a trigger is swapped, making preset conflicts visible during trigger changes.

OpenAI Agents SDK v0.19.1 adds native host path support in sandbox path grants.

$ git clone --branch v0.19.1 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.19.1

• ›Adds support for native host paths in sandbox path grants, enabling direct host filesystem access without path translation.

PydanticAI v2.20.0 adds Claude Opus 5 and OpenAI Responses API reasoning.context support for GPT-5 families.

$ git clone --branch v2.20.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v2.20.0

from pydantic_ai import Agent

agent = Agent('anthropic:claude-opus-5')
result = agent.run_sync('Summarize the OWASP Top 10 for 2025.')
print(result.output)

• ›Adds reasoning.context support for the OpenAI Responses API, defaulting to all_turns, for the gpt-5.4, gpt-5.5, and gpt-5.6 model families.
• ›Adds claude-opus-5 model support via the Anthropic provider.

Autonomous coding agent as an SDK, IDE extension, or CLI assistant.

Cline desktop v0.0.7 adds system tray status, session favorites, subagent visibility, and free-tier model support.

$ git clone --branch desktop-v0.0.7 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout desktop-v0.0.7

• ›New system tray icon displays live app status and the count of currently running agent sessions.
• ›Session history is now paginated in ten-session pages, loading older sessions on demand as you scroll.
• ›Sessions can be favorited and are ordered by most recent activity, with consistent status dot colors across views.
• ›Subagent and teammate runs now appear inline within their parent session, showing status and results.
• ›Free Cline models are now available and labeled '(free)' in model pickers, with a message including reset time when the free-tier limit is reached.

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

Goose v1.45.0 adds air-gapped docs via GOOSE_DOCS_ROOT, disableable built-in skills, scheduler flags, compaction templates, and latest Gemini/Opus5 model support.

$ git clone --branch v1.45.0 https://github.com/aaif-goose/goose.git
# already have the repo? check out this version:
$ git checkout v1.45.0

$ export GOOSE_DOCS_ROOT=/mnt/internal-docs/goose
goose session start

$ goose serve --enable-scheduler

• ›Adds GOOSE_DOCS_ROOT environment variable to point Goose at a local documentation mirror for air-gapped deployments.
• ›Adds --enable-scheduler flag to both the serve and acp subcommands to opt in to scheduled recipe execution (disabled by default for ACP).
• ›Adds tab completion and provider-switching support to the model command.
• ›Supports latest Gemini models and Opus5 (including adaptive thinking).
• ›Introduces stable agent event message identity for consistent event tracking across the agent loop.

$ git clone --branch v1.45.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.45.0

$ export GOOSE_DOCS_ROOT=/mnt/internal/goose-docs
goose session start

Lightweight coding agent that runs in your terminal

Codex CLI adds managed in-app updates so the tool can keep itself current without manual reinstallation.

$ git clone --branch rust-v0.146.0-alpha.9.1 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.146.0-alpha.9.1

• ›Adds managed in-app updates, allowing Codex CLI to update itself automatically from within the application.

$ git clone --branch rust-v0.147.0-alpha.1 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.147.0-alpha.1

$ git clone --branch rust-v0.146.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.146.0

$ /new my-audit-session

OpenHands: AI-Driven Development

OpenHands v1.7.0 adds persistent agent memory toggle, LLM selector always visible, and secret overwrite from edit form.

$ git clone --branch v1.7.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:
$ git checkout v1.7.0

• ›Adds a persistent agent memory toggle in the UI, letting users enable or disable memory retention across agent sessions.
• ›LLM selector is now always visible in the UI; agent-profile switching has been moved to the tools menu.
• ›Allows overwriting an existing secret value directly from the secret edit form.

$ git clone --branch v1.7.0 https://github.com/OpenHands/OpenHands.git
# already have the repo? check out this version:
$ git checkout v1.7.0

AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

Pi v0.83.0 adds credential export commands, headless OpenRouter sign-in, and Claude Opus 5 on GitHub Copilot.

$ git clone --branch v0.83.0 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.83.0

$ pi auth print-bearer-token | xargs -I{} curl -H 'Authorization: Bearer {}' https://api.example.com/endpoint

• ›Adds pi auth print-api-key and pi auth print-bearer-token commands to export configured credentials to external clients, with automatic OAuth refresh and minimum-validity enforcement.
• ›Adds support for manual redirect URL and authorization-code entry during pi auth login on OpenRouter, enabling headless and SSH-remote sign-in when the loopback callback is unavailable.
• ›Adds the 'pending' stop reason for partial streaming messages in the custom provider stream pattern.
• ›Adds per-request fetch injection for supported text and image provider transports.
• ›Adds Claude Opus 5 support via GitHub Copilot with adaptive thinking and a 1M context window.

Zed v1.13.1 adds include_ignored file-finder param, title_bar.show_worktree_name setting, run-status gutter indicators, and editable image-viewer zoom.

$ git clone --branch v1.13.1 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v1.13.1

["file_finder::Toggle", { "include_ignored": true }]

• ›Adds include_ignored parameter to file_finder::Toggle, enabling keybindings that open the File Finder with ignored files pre-included.
• ›Adds title_bar.show_worktree_name setting (defaults to true); set to false to hide the worktree-name picker in the title bar.
• ›Adds run-status indicators to runnable gutter controls showing the last run result, plus a Clear Run Status context-menu action for completed runs.
• ›Adds 'Copy this Agent response', 'Scroll to User Message', and 'Scroll to top' buttons to every Agent Panel response.
• ›Adds support for editing image-viewer zoom levels directly from the toolbar.

Triton v2.71.0 adds multi-GPU TensorRT inference, PyTorch 2 batching, NV embedding layers, and tighter OpenAI frontend security.

$ git clone --branch v2.71.0 https://github.com/triton-inference-server/server.git
# already have the repo? check out this version:
$ git checkout v2.71.0

• ›Redacts restriction header values from the OpenAI frontend startup log and enforces inference restrictions in the SageMaker MME invoke handler.
• ›Adds a streaming tool-call parse buffer limit to the OpenAI-compatible frontend to cap memory usage during streaming tool calls.
• ›TensorRT backend gains multi-device (multi-GPU) inference support.
• ›PyTorch backend enables PyTorch 2 batching and adds support for loading NV embedding layers.
• ›Adds support for serving HSTU generative-recommender models via the PyTorch AOTI serving path.

llama.cpp b10174 adds NextN/MTP speculative decoding support for GLM-5.2 (GLM_DSA) with --spec-type draft-mtp and --mtp/--no-mtp convert flags.

$ git clone --branch b10174 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b10174

$ python convert_hf_to_gguf.py --no-mtp path/to/GLM-5.2

• ›Adds --spec-type draft-mtp support for GLM_DSA (GLM-5.2) models, enabling NextN/MTP speculative decoding with a dense MLA draft head, sigmoid-gated MoE with shared expert, and NVFP4 _s scale tensor support.
• ›Adds --mtp and --no-mtp export flags to the converter for GlmMoeDsaForCausalLM (GLM-5.2): --no-mtp drops the NextN block (blk.78) and its nextn_predict_layers KV; --mtp keeps only the NextN block plus shared embeddings, norm, and lm_head.

Langfuse v4 brings full-text search, monitors/alerts, filter search bar, and faster Observations/Metrics API v2 to self-hosted deployments.

$ git clone --branch v4.0.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v4.0.0

• ›Adds Observations API v2 with significantly faster performance for self-hosted deployments.
• ›Adds Metrics API v2 with significantly faster performance for self-hosted deployments.
• ›Adds full-text search across inputs, outputs, and metadata for self-hosted deployments.
• ›Adds a filter search bar for narrowing trace and observation views.
• ›Adds monitors and alerts for metric-based observability.

Milvus 3.0.0 adds FAISS passthrough, TEXT long-text fields, SINDI sparse index, StructArray search, faceted search, composable reranking, and Woodpecker standalone deployment.

$ git clone --branch v3.0.0 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v3.0.0

{
  "index_type": "FAISS",
  "metric_type": "L2",
  "params": {
    "faiss_index_name": "OPQ16,IVF64,PQ16x4"
  }
}

common:
  storage:
    useLoonFFI: true
dataCoord:
  targetVecIndexVersion: 10
  targetScalarIndexVersion: 4

• ›Adds FAISS index type accepting arbitrary Faiss index-factory strings via the faiss_index_name parameter (e.g., IVF64,Flat, HNSW16,Flat, OPQ16,IVF64,PQ16x4), with search parameters passed through so Faiss recipes reproduce directly on Milvus.
• ›Adds milvus-table external format that treats Milvus Snapshot metadata and Storage V3 manifests as an external source, enabling a collection snapshot to be served as an external table for shared, manifest-backed access across batch and serving systems.
• ›Introduces TEXT fields for long text with storage-side length limits removed, supporting text_match, phrase_match, and BM25; values under 64 KB stay inline, larger values go to partition-level LOB files in Vortex format storing only (file_id, offset) references.
• ›Adds native XGBoost scoring for L0 reranking using UBJ models registered as FileResources, plus Hugging Face Inference Providers for server-managed text embedding and sentence-similarity reranking via the Function Chain API.
• ›Overhauled sparse vector index with SINDI, Block-Max WAND, and Block-Max MaxScore algorithms, inverted-list compression, and configurable quantization; SINDI becomes the default for sparse IP search and MaxScore for BM25 once new index versions are enabled — compressed BM25 index is ~3x smaller than the 2.6 sparse index and SINDI reaches ~10x the QPS of MaxScore on learned sparse embeddings.

Weaviate v1.38.8 adds cross-property AND matching in BM25, new S3 auth broker credentials, and a runtime-overridable batched Contains gate.

$ git clone --branch v1.38.8 https://github.com/weaviate/weaviate.git
# already have the repo? check out this version:
$ git checkout v1.38.8

• ›Adds feat(backup-s3) new auth broker credentials for the S3 backup module, enabling alternative credential flows beyond static keys.
• ›Adds a runtime-overridable feature gate to opt into batched ContainsAny/ContainsAll/ContainsNone resolution via the config, inverted layer.
• ›Adds cross-property AND matching support in BM25 search, allowing BM25 queries to require match terms across multiple properties simultaneously.
• ›Exposes the first letters of API keys to admins on namespaced clusters, improving key identification without revealing secrets.
• ›Adds nested object filtering support to the usage module.