This repository demonstrates building agents using LangGraph, focusing on an email assistant that can: - Triage incoming emails - Draft appropriate responses - Execute actions (calendar scheduling, etc.) - Incorporate human feedback - Learn from past interactions Recommended: Using uv (faster and more reliable) Alternative: Using pip The repository contains several implementations with increasing complexity in src/email_assistant: Each aspect of the agent is explained in dedicated notebooks: - notebooks/langgraph_101.ipynb - LangGraph basics - notebooks/agent.ipynb - Basic agent implementation - notebooks/evaluation.ipynb…
2.2k18mo agoReviewMountCreate a free account to mount
Use tox for testing (not pytest directly): - Test matrix configuration is in tox.ini - Integration tests: uv run tox -e py3.14-{integration}-v{version} - Common tests: uv run tox -e py3.14-common - Run specific test file: TESTPATH=tests/integrations/logging/test_logging.py uv run tox -e py3.14-common - Run single test: TESTPATH=tests/path/to/test_file.py uv run tox -e py3.14-common -- -k "test_name" Mypy and its type stubs live in the opt-in typing dependency group: - Run uv run --group typing mypy sentry_sdk before committing (must pass with zero…
No other project uses it yet
2.2k9mo agoReviewMountCreate a free account to mount
Intelligent Terminal is a Windows Terminal fork that adds first-class AI agent workflows. The inherited Windows Terminal build, architecture, and C++ conventions are documented in .github/copilot-instructions.md; this file contains only the fork-specific context. - WTA (tools/wta/) is the Rust orchestrator. - ACP means Agent Client Protocol. wta-master owns a pool of agent CLI processes keyed by agent identity, execution source, and command; helpers using the same key share one process and multiplex sessions through it. Installed, policy-allowed native host agents…
Steady2 mentions in 2 weeks
No other project uses it yet
2.2k4mo agoReviewMountCreate a free account to mount
One packaged Windows app hosts TerminalApp local tests and real production UI for isolated accessibility checks. It does not replace full-app integration tests. The harness surface fre-settings launches the FRE mode, invokes Next through UI Automation, and scans the resulting settings controls. It is not a separate app or launch argument. Agents mode activates the host window before creating the production view model and navigates its Frame with NavigateToPageArgs. This exercises the page's normal OnNavigatedTo initialization, not a TestHost-only Agents…
2.2k4mo agoReviewMountCreate a free account to mount
These rules apply to tools/wta/. The repo-level AGENTS.md is the canonical source for build, deployment, supported providers, runtime paths, and log names. Do not duplicate those facts here. WTA is a Rust binary with three launch modes: - Master (--master <pipe>): owns a pool of agent CLI subprocesses and their ACP stdio connections, accepts helper named-pipe connections, and routes each ACP SessionId to its owning helper. Helpers with the same master-derived agent key share a process. Installed, policy-allowed native host…
Steady2 mentions in 2 weeks
No other project uses it yet
2.2k4mo agoReviewMountCreate a free account to mount
Skills live in the top-level skills/ directory — repo-development guidance for agents working in this repository. (User-facing skills shipped to merchants in the npm package live in packages/hydrogen/skills/ instead.) Each skill is symlinked into both .agents/skills/ and .claude/skills/ so every agent harness discovers it. To add a new skill: Don't add skills directly to .agents/skills/ or .claude/skills/ — those only hold symlinks. When designing or adjusting APIs for the hydrogen package, closely follow the following principles:
2.2k26d agoReviewMountCreate a free account to mount
This guide helps AI agents review, maintain, and contribute to Cloudflare Worker templates. It documents patterns, common issues, and testing strategies learned from real PR reviews. When reviewing a template PR, verify these items in order: Symptom: Template linter fails with missing description error. Fix: Add a one-line description to package.json: Symptom: Dashboard won't display template details correctly. Fix: Add markers around the content that should appear in Cloudflare Dashboard: Important: The Deploy to Cloudflare button should be at the…
2.2k10mo agoReviewMountCreate a free account to mount
Notes for AI agents working on this template. One enriched content store (Resource[]) projected onto many agent-discovery surfaces. The data is enriched once by Workers AI and cached in KV; every surface (/llms.txt, /index.json, /<slug>.md, /robots.txt, JSON-LD) is just a different rendering of that same store. - surfaces.ts is pure. Render functions take RenderCtx and return strings/objects. No I/O. This keeps surfaces easy to test and add to. - Enrichment must never hard-fail. enrichResource falls back to fallbackEnrichment on any…
2.2k10mo agoReviewMountCreate a free account to mount
A Worker template that tests whether AI models (GPT-5.4, Claude Sonnet 4, Gemini 3 Flash, Llama 4, Mistral) mention your brand when answering relevant queries. All models run through Cloudflare AI Gateway — no external API keys needed.
No other project uses it yet
2.2k10mo agoReviewMountCreate a free account to mount
Context for AI coding agents to help users set up x402-proxy for payment-gated content. A Cloudflare Worker that adds payment gating to any origin using the x402 protocol. Users pay to access protected routes, then get a JWT cookie valid for 1 hour. With Bot Management Filtering: Requires Bot Management for Enterprise to enable bot filtering. With it enabled, x402-proxy can implement "default closed" - humans pass through free, only bots must pay. This is optional and enhances the base…
2.2k10mo agoReviewMountCreate a free account to mount
Scope: Module-specific guidance for the website/ directory. For Blog Posts: See .github/instructions/website.blog.instructions.md For Repo Standards: See .github/copilot-instructions.md Prerequisites: Node.js >= 18.0, npm 9+ First time? Start with npm start, then create a post in blog/YYYY-MM-DD-slug/index.md. Changes hot-reload in browser. See .github/instructions/website.blog.instructions.md for content guidelines and writing style. Quick steps: See .github/instructions/website.blog.instructions.md for required and optional fields.
2.1k12mo agoReviewMountCreate a free account to mount
Alpha: changing or removing behavior needs no deprecation path. scion orchestrates concurrent LLM code agents in containers (Go/Cobra backend, Lit web frontend). Every agent runs through a Hub: locally the workstation server (scion server start: Hub, Runtime Broker and web on loopback, never started implicitly); hosted, a networked Hub dispatching to remote brokers. Architecture: docs-site/src/content/docs/contributing/architecture.md; design specs: .design/hosted/. "Grove" is the legacy name for "project".
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
This guide outlines the organizational philosophy, architecture, and curation standards for the Scion documentation project. It serves as a blueprint for both human and AI contributors to ensure a consistent and navigable documentation experience. Scion documentation is organized on the operational-mode spine defined in the root GLOSSARY.md — the single source of truth for terminology. The site is not organized by persona. Instead, readers self-select by the deployment target they are running: Workstation → Single-node hosted → HA hosted This…
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
Before you ask the user a question, you must always call the sciontool_status tool: sciontool_status("ask_user", "<question>") And then proceed to ask the user. When you are intentionally waiting for something — such as an external process or a scheduled event — signal that you are blocked: sciontool_status("blocked", "<reason>") Once you believe you have completed your task, you must summarize and report back to the user as you normally would, but then be sure to signal completion: sciontool_status("task_completed", "<task title>") Do…
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
You are running inside a scion-managed container. Use sciontool to report your status: Your workspace is mounted at /workspace. This is a git worktree — you have your own branch and can commit freely without affecting other agents. If you see the exact message "System: Please Continue." — ignore it. This is an automated artifact, not a real user message.
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
Before you ask the user a question, you must always execute the script: sciontool status ask_user "<question>" And then proceed to ask the user When you are intentionally waiting for something — such as a child agent you started to complete, or a scheduled event you are expecting — you must signal that you are blocked: sciontool status blocked "<reason>" For example: sciontool status blocked "Waiting for agent deploy-frontend to complete" This prevents the system from falsely marking you as…
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
This document provides instructions for AI agents working on the Scion Web Frontend. Before making changes, review the relevant design documentation: New experimental user-facing features must be launched behind a registered experiment (pkg/experiments/registry.go); see docs-site/src/content/docs/reference/experiments.md. Browser-side flag values are user-editable and only affect presentation; if an experiment changes hub behaviour, register it with the server layer and check it in the hub (requireExperiment / experimentEnabled). Do not add ad-hoc scion:feature: string literals or new booleans to /api/v1/settings/public. The web frontend…
No other project uses it yet
2.1k15d agoReviewMountCreate a free account to mount
%% lat:begin %% # Before starting work Use Lat's knowledge graph to ground each task before changing code. After EVERY task, before responding to the user: Lat.md is the project's structured knowledge graph for architecture, design decisions, behavior, and tests. The graph lives in lat.md/ as cross-linked Markdown describing what the project does and why. Use it to ground work in actual architecture, and keep it a focused snapshot of the current implemented state rather than a journal or changelog.
1 project uses a version of this
2.1k38d agoReviewMountCreate a free account to mount
LangChainAGENTS.md
