For new numerical APIs and kernel paths, use the determinism skill to add/run L0 tests and update API support. For kernel determinism changes, follow the root rules and the coverage table. The shared comparator checks bytes, tensor names, shapes, and dtypes. Clone each result before the next launch so reused storage cannot make replay pass. For backward batch checks, slice fixed upstream gradients with the inputs; do not use a batch-mean loss. GDP expands K/V/beta by num_householder, while Q/G remain…
99410d agoReviewMountCreate a free account to mount
Shared test infrastructure lives in lib/sentry/test/. Root Gemfile.dev defines shared dev dependencies. Root-level bundle exec rake runs the E2E/integration spec suite (not individual gem tests).
9895mo agoReviewMountCreate a free account to mount
The skills previously in this repository now live in the Next.js repo: https://github.com/vercel/next.js/tree/canary/skills Install: npx skills add vercel/next.js See README.md for where each old skill went.
9825mo agoReviewMountCreate a free account to mount
Magic Modules (MMv1) defines the Google Cloud Terraform providers: resource schemas in mmv1/products/, handwritten code and tests under mmv1/third_party/terraform/, generated into the downstream terraform-provider-google and terraform-provider-google-beta repositories. Changes are always…
9754mo agoReviewMountCreate a free account to mount
Wassette is an MCP server that runs tools as WebAssembly components with Wasmtime. This file is the entry point for coding agents and contributors; the linked documentation and skills are the canonical sources for detailed guidance. - Contributing: contribution requirements and release-note policy. - Developer guide: prerequisites, repository layout, build, test, lint, documentation, and local CI workflows. - Architecture: system design and component boundaries. - Release process: release automation and versioning. Focused workflows live under .agents/skills/: Agents that support the…
9723mo agoReviewMountCreate a free account to mount
Agents use an LLM to determine which actions to take and in what order. An action can either be using a tool and observing its output, or returning to the user. For a list of easily loadable tools, see here. Here are the agents available in LangChain. For a tutorial on how to load agents, see here. This agent uses the ReAct framework to determine which tool to use based solely on the tool's description. Any number of tools can…
9664y agoReviewMountCreate a free account to mount
Agents are systems that use a language model to interact with other tools. These can be used to do more grounded question/answering, interact with APIs, or even take actions. These agents can be used to power the next generation of personal assistants - systems that intelligently understand what you mean, and then can take actions to help you accomplish your goal. Agents are a core use of LangChain - so much so that there is a whole module dedicated to…
9664y agoReviewMountCreate a free account to mount
Changesets are created in their own dedicated PRs, separate from feature or bug-fix PRs. A feature PR should contain only the code changes. After it merges, open a follow-up PR that adds the changeset.
9659mo agoReviewMountCreate a free account to mount
A provider commits to a model's weights with cp_commit, then proves with cp_prove that each reply passes an acceptance policy. Built from prover/, both run on one GPU. Clients check proofs on a CPU with cp_verify, the C++ verifier in verifier/. README.md, "For agents", has every route with what it needs and how long it takes. Run everything here from an ordinary account, not under sudo or as root. The programs, scripts and tests need no root rights, and as…
96415d agoReviewMountCreate a free account to mount
Toolkit contains the Sentry CLI, the Sentry MCP server, and shared code. Product-specific instructions live in their packages; start with the shared documentation index. Read before MCP tool changes: Shared…
9608mo agoReviewMountCreate a free account to mount
Mistral-Common is a preprocessing library for Mistral's Large Language Models (LLMs). It encodes requests for Instruct, Transcription or Fill-In-The-Middle (FIM) tasks to tokens and optionally processed images or audios. - Language: Python 3.11 to 3.15 - Package Manager: uv - Testing: pytest - Formatting and Linting: Ruff - Type checker: mypy - CI: GitHub Actions
96012d agoReviewMountCreate a free account to mount
pydantic-ai-harness is the first-party capability library for Pydantic AI. Pydantic AI core owns the primitive runtime: agent loop semantics, normalized messages, model/provider/profile behavior, tool execution semantics, durable execution primitives, and generic capability hooks. Harness owns optional, batteries-included compositions built from those primitives: coding-agent tools, guardrails, memory, context management, repo tools, verification loops, skills, planning, sub-agents, and other reusable agent behaviors. When a change needs new core semantics, stop and propose the Pydantic AI core change instead of reimplementing core behavior…
9544mo agoReviewMountCreate a free account to mount
Read this before touching pydantic-clai2/. The repository-level AGENTS.md still applies (no em-dashes, no Any, pyright strict, keyword-only arguments, 100% branch coverage). This file adds what is specific to the terminal shell. A thin terminal around any Pydantic AI agent. It reads a prompt, runs the agent, streams the answer, and repeats. Everything beyond that is a plugin, including the default coding tools. The three layers, and who owns what: If a change needs the agent loop to behave differently, propose…
9544mo agoReviewMountCreate a free account to mount
- When creating operations, always use the format OpType::create(rewriter, ...) and never use the (deprecated) format rewriter.create<OpType>(...). - When writing unit tests with filecheck, never use CHECK-LABEL; just use CHECK instead. - scripts/templates/templates.py (which requires python-fire and jinja2) provides boilerplate for generating new transforms and passes. After running the script to create a new pass, one still must register the pass in heir-opt.cpp and add the necessary build dependencies. Copied from https://mlir.llvm.org/docs/Tutorials/MlirOpt/ with minor tweaks. - --mlir-timing displays execution times…
9487mo agoReviewMountCreate a free account to mount
Run pnpm type-check after every change to ensure there are no type errors. Use the shadcn CLI (npx shadcn@latest add <component>) to install UI components in apps/web. Never write shadcn components by hand. Never use native JS alert(), confirm(), or prompt(). Use custom UI components instead. All apps and examples with dev servers use portless to avoid hardcoded ports. Portless assigns random ports and exposes each app via .localhost URLs. Naming convention: When adding a new example that runs a…
9467mo agoReviewMountCreate a free account to mount
This is the official Ruby SDK for the Model Context Protocol (MCP), implementing both server and client functionality for JSON-RPC 2.0 based communication between LLM applications and context providers. MCP::Server…
94548d agoReviewMountCreate a free account to mount
When helping with a pull request review, separate context gathering from review comments. Use the repo-local review-nvbench skill when available. The default mode is context gathering: use ci/util/pr_review_context.sh, explain the issue, PR intent, implementation strategy, and suggested file review order, but do not produce review findings until the user explicitly asks for feedback. If no issue or PR context is discoverable, explicitly flag that the review is missing required context before proceeding. Follow docs/pr_review.md for the full review workflow and…
9455mo agoReviewMountCreate a free account to mount
This is a Turborepo monorepo. The CLI application lives in packages/ai-cli/ and shared configuration packages live in packages/. Use Node.js 24+ and npm for all package management and script execution:
94517d agoReviewMountCreate a free account to mount
* Review the CONTRIBUTING.md file for instructions to build and test the software. * Run the .github/Prime-ForCopilot.ps1 script (once) before running any dotnet or msbuild commands. If you see any…
9434mo agoReviewMountCreate a free account to mount
NVIDIAAGENTS.md
