Shipped · AI and LLM tooling

What AI and LLM tooling repositories shipped recently

An automatic summary of GitHub releases, pull requests and commits, made by FoxPlug. The projects named did not write it and do not use or endorse FoxPlug; FoxPlug is not affiliated with them. These repositories had too little public activity for a page of their own, so what they shipped is listed here together.

Get updates like this for your repo

browser-use/browser-use

Browser Use · github.com/browser-use/browser-use

Week of October 12, 2026

A failed or canceled browser reconnect now closes its replacement connection and clears its state before the original error is raised, with regression tests. Pull request #6102
Transient failures in browser reads no longer stop the runner, page reads keep valid element references, and diagnostics no longer block ordinary actions. Pull request #6099
Page observations keep their hierarchy, field values and control states, and returned page text is bounded. Pull request #6093
An uncommitted Cloud tab is reported as about:blank so navigation no longer stops the tool runner before it returns the action result. Pull request #6089
The browser-use-sdk dependency now allows versions 3.10 up to but not including 4, so the documented v4 import works. Pull request #6098
Using a cloud browser without an API key now fails early with a single clear message instead of a traceback. Pull request #6112

Week of September 28, 2026

The Cloud skill reference was updated to reflect the new $1 credit for eligible new signups instead of $15. Pull request #5982

Merged pull requests

One CLI: say browser-use everywhere (#6377, merged November 11, 2026 by gregpr07) (source)
1.0.2 (#6376, merged November 10, 2026 by gregpr07) (source)
1.0.1 (#6369, merged November 10, 2026 by gregpr07) (source)
1.0.0 (#6359, merged November 10, 2026 by gregpr07) (source)
README: how the agent works, under the header (#6360, merged November 10, 2026 by gregpr07) (source)
1.0.0rc5: Linux rc test fixes and security review fixes (#6357, merged November 9, 2026 by gregpr07) (source)
browser-use 1.0: REPL engine replaces the 0.x engine (first version) (#6314, merged November 9, 2026 by gregpr07) (source)
telemetry v3: where runs spend time and where models struggle (#6348, merged November 9, 2026 by gregpr07) (source)
fix: streamed compaction items match the non-stream response (#6344, merged November 9, 2026 by gregpr07) (source)
Add Fireworks model example (#6238, merged October 30, 2026 by gregpr07) (source)
Release browser-use 0.13.14 before the Pi migration (#6220, merged October 26, 2026 by MagMueller) (source)
README: rerun a saved task without calling the model on every step (#6183, merged October 22, 2026 by gregpr07) (source)

Releases

1.0.2rc1: release candidate for 1.0.2, a pre-release published November 10, 2026. First change listed: --cli-mcp with Chrome remote debugging off now drives its own Chromium and tells the model how to get your Chrome (toggle at chrome://inspect, restart the server). (source)
browser-use 1.0.1, published November 10, 2026. First change listed: MCP without Chrome: browser-use --cli-mcp on servers, containers and CI boxes now drives its own Chromium, installed on first use. When no browser can run, you get one clear error. (source)
1.0.0 (browser-use 1.0: CDP is all you need), published November 10, 2026. First change listed: An option 1.0 cannot honor raises a TypeError naming it before any browser starts or anything is billed. No silent no-ops. 0.x tuning knobs 1.0 has no use for (usevision=True, flashmode, maxfailures… (source)
1.0.0rc5: release candidate for browser-use 1.0, a pre-release published November 9, 2026. First change listed: An option 1.0 cannot honor raises a TypeError naming it before any browser starts or anything is billed. No silent no-ops. 0.x tuning knobs 1.0 has no use for (usevision=True, flashmode, maxfailures… (source)
1.0.0rc4: release candidate for browser-use 1.0, a pre-release published November 9, 2026. First change listed: An option 1.0 cannot honor raises a TypeError naming it before any browser starts or anything is billed. No silent no-ops. 0.x tuning knobs 1.0 has no use for (usevision=True, flashmode, maxfailures… (source)
1.0.0rc2: release candidate for browser-use 1.0, a pre-release published November 9, 2026. First change listed: An option 1.0 cannot honor raises a TypeError naming it before any browser starts or anything is billed. No silent no-ops. 0.x tuning knobs 1.0 has no use for (usevision=True, flashmode, maxfailures… (source)

From the README

The open-source browser agent. Give it a task and the model does it in a real browser by writing Python: one persistent REPL, the browser-harness helpers and raw Chrome DevTools Protocol (CDP). Runs on any LLM, locally or on Browser Use Cloud. MIT licensed. (source)

Who ships it

MagMueller: 3,811 commits on the default branch (source)
pirate: 1,810 commits on the default branch (source)
gregpr07: 1,146 commits on the default branch (source)
mertunsall: 817 commits on the default branch (source)
sauravpanda: 730 commits on the default branch (source)
laithrw: 344 commits on the default branch (source)

mem0ai/mem0

Mem0 · github.com/mem0ai/mem0

Week of September 21, 2026

Python SDK v2.2.1 fixes add() to report only records the vector store actually inserted, raising an error if none succeed. Release
Node SDK v3.3.0 adds User Profiles with getProfile(), generateProfile(), getProfileSettings(), updateProfileSettings(), sampleProfiles(), and getProfileJob(). Release
Python SDK v2.2.0 adds User Profiles with get_profile(), generate_profile(), get_profile_settings(), update_profile_settings(), sample_profiles(), and get_profile_job(). Release
TurbopufferDB vector store now respects distance_metric in search score calculation instead of always using cosine formula. Pull request #6559
Memory and AsyncMemory context-manager protocol restored to support with...as syntax. Pull request #7354
TypeScript SDK no longer forces pg as a required peer dependency, making it optional for hosted MemoryClient users. Pull request #7450
Python SDK dev tooling (ruff, isort, pre-commit) now wired into hatch environments so setup works on fresh clone. Pull request #6684

Merged pull requests

fix(security): upgrade @azure/msal-node to 7.x via @azure/identity 4.14.1 (CVE-2026-109461) (#7850, merged November 6, 2026 by harshgupta-mem0) (source)
fix(security): resolve 9 Vanta HIGH Dependabot vulnerabilities (#7849, merged November 6, 2026 by harshgupta-mem0) (source)
fix(security): resolve 5 Vanta CRITICAL Dependabot vulnerabilities (#7825, merged November 4, 2026 by harshgupta-mem0) (source)
fix(memory): never re-create a deleted memory when update races delete (#7787, merged October 30, 2026 by harsh-mem0) (source)
fix(ts): never re-create a deleted memory when update races delete (#7788, merged October 30, 2026 by harsh-mem0) (source)
docs(platform): add the pricing FAQ page (#7711, merged October 26, 2026 by harsh-mem0) (source)
docs(platform): rewrite the BYOK page as a feature guide (#7661, merged October 21, 2026 by harsh-mem0) (source)
docs(platform): BYOK: the new model picks and chart in Choose a model (#7666, merged October 17, 2026 by agumpandey) (source)
Add OSS Halloween Happy Hour banner to README (#7637, merged October 14, 2026 by karthik-indla) (source)
fix(client): reject empty entity IDs and filters in add and delete_all (#7581, merged October 12, 2026 by blazingretinol) (source)
docs(skills): refresh mem0 skills against SDK 2.2.1 / TS 3.3.1 (#7548, merged October 7, 2026 by blazingretinol) (source)
fix(vector-store): stop writing service account private key to DEBUG logs in GoogleMatchingEngine (#7506, merged October 7, 2026 by rupak-eng) (source)

Releases

Mem0 Python SDK (v2.2.1), published September 25, 2026. First change listed: Core: add() (Memory and AsyncMemory) no longer reports records the vector store rejected as successful ADD events. Only records that were actually inserted are written to history, entity-linked, and… (source)
Mem0 Python SDK (v2.2.0), published September 23, 2026. First change listed: Client: Add User Profiles to MemoryClient and AsyncMemoryClient: getprofile(), generateprofile(), getprofilesettings(), updateprofilesettings(), sampleprofiles(), and getprofilejob(). A profile is a… (source)
ts-v3.3.0 (Mem0 Node SDK (v3.3.0)), published September 23, 2026. First change listed: Client: Add User Profiles to MemoryClient: getProfile(), generateProfile(), getProfileSettings(), updateProfileSettings(), sampleProfiles(), and getProfileJob(). A profile is a structured… (source)
pi-agent-v0.3.2 (Mem0 Pi Agent Plugin (v0.3.2)), published September 23, 2026. First change listed: Search: The memory policy, mem0memory tool description, and prompt guidelines no longer ask the agent to search before answering anything that may depend on earlier context or to run several searches… (source)
opencode-v0.4.1 (Mem0 OpenCode Plugin (v0.4.1)), published September 23, 2026. First change listed: Search: The searchmemories tool description no longer asks the agent to search proactively or to run several searches for multi-part questions. It now asks for a search before repeating investigation… (source)

From the README

All benchmarks run on the same production-representative model stack. Single-pass retrieval (one call, no agentic loops) at a top200 retrieval budget. Scores reflect Mem0's managed platform, which includes proprietary optimizations not available in the open-source SDK; open-source users should expect directionally similar gains but not identical numbers. (source)

Who ships it

Dev-Khant: 453 commits on the default branch (source)
blazingretinol: 229 commits on the default branch (source)
deshraj: 219 commits on the default branch (source)
taranjeet: 206 commits on the default branch (source)
whysosaket: 173 commits on the default branch (source)
cachho: 104 commits on the default branch (source)

microsoft/markitdown

Microsoft · github.com/microsoft/markitdown

Week of September 28, 2026

Consolidated format-specific regressions from 35 test files into 18 files with one suite per format, documented in tests/README.md. Pull request #2579
Restored JSON regression coverage for late non-ASCII text detection by adding json_late_non_ascii.json test fixture. Pull request #2601
OCR plugin registration tests now discover and load real installed OCR packages through the core plugin loader instead of mocked entry points. Pull request #2588
Conversion regressions now exercise real PDF parsing, PowerPoint charts, and HTTP responses instead of local parser mocks. Pull request #2586
PowerPoint missing-text tests now convert real saved presentations and verify surrounding content survives through library and CLI vectors. Pull request #2594
CLI file-type parsing and formatting tests now run production code paths with valid, normalized, single-type, and invalid inputs. Pull request #2592
MCP input error tests replaced mocked failures with real conversions using unsupported gopher URIs and binary fixtures. Pull request #2597
WikipediaConverter fixed to handle italicised titles by reading from h1#firstHeading when mw-page-title-main span is absent. Pull request #2458
HTTP response reading performance improved by increasing chunk size from 512 bytes to 64 KB. Pull request #1511

Merged pull requests

fix: match ExifTool directories case-insensitively on Windows (#2837, merged November 7, 2026 by Yasser-Ameur) (source)
Override default SpeechRecognition endpoint (#2846, merged November 6, 2026 by afourney) (source)
Clarify hosted service requirements in README (#2845, merged November 6, 2026 by afourney) (source)
feat(doc-intel): expose models, features, query fields, and structured output (#1913, merged October 22, 2026 by chienyuanchang) (source)
Update Black, Pillow, and Windows ARM Cryptography baselines (#2731, merged October 21, 2026 by afourney) (source)
Require certifi>=2026.10.19 for updated CA trust (#2730, merged October 21, 2026 by afourney) (source)
Restore JSON regression coverage for late non-ASCII text (#2601, merged October 4, 2026 by afourney) (source)
Test MCP input errors with real conversions (#2597, merged October 4, 2026 by afourney) (source)
Test PowerPoint missing text with real fixtures and shared vectors (#2594, merged October 3, 2026 by afourney) (source)
Test real CLI file-type parsing and CU result formatting (#2592, merged October 3, 2026 by afourney) (source)
Test real installed OCR plugin discovery (#2588, merged October 3, 2026 by afourney) (source)
Replace local parser mocks with real conversion fixtures (#2586, merged October 3, 2026 by afourney) (source)

Releases

v0.1.9 (Version 0.1.9), published October 21, 2026. First change listed: increase chunk size from 512 bytes to 64 KB for HTTP response reading (source)
v0.1.8 (Version 0.1.8), published September 21, 2026. First change listed: Mitigate UnicodeDecodeError due to wrong ASCII charset guess for long files (source)
v0.1.8b2 (Version 0.1.8b2), a pre-release published September 14, 2026. First change listed: Pin GitHub Actions to full-length commit SHAs (source)
v0.1.8b1 (Version 0.1.8b1), a pre-release published September 4, 2026. First change listed: Pin GitHub Actions to full-length commit SHAs (source)
v0.1.7 (Version 0.1.7), published July 29, 2026. First change listed: fix: avoid O(n^2) value lookups in PPTX chart conversion (source)
v0.1.6 (Version 0.1.6), published May 26, 2026. First change listed: [MS] Add OCR layer service for embedded images and PDF scans (source)

From the README

MarkItDown is a lightweight Python utility for converting various files to Markdown for use with LLMs and related text analysis pipelines. To this end, it is most comparable to textract, but with a focus on preserving important document structure and content as Markdown (including: headings, lists, tables, links, etc.) While the output is often reasonably presentable and human-friendly, it is… (source)

Who ships it

afourney: 141 commits on the default branch (source)
gagb: 70 commits on the default branch (source)
sugatoray: 9 commits on the default branch (source)
PetrAPConsulting: 8 commits on the default branch (source)
Josh-XT: 7 commits on the default branch (source)
l-lumin: 7 commits on the default branch (source)

Other AI and LLM tooling repositories

Last updated 2026-11-10.