https://github.com/coderrob/agentics
GitHub Agentic Workflows
https://github.com/coderrob/agentics
Last synced: 9 days ago
JSON representation
GitHub Agentic Workflows
- Host: GitHub
- URL: https://github.com/coderrob/agentics
- Owner: Coderrob
- License: other
- Created: 2026-04-22T03:26:17.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-29T05:31:09.000Z (2 months ago)
- Last Synced: 2026-04-29T07:23:40.780Z (2 months ago)
- Language: TypeScript
- Size: 285 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
- Codeowners: .github/CODEOWNERS
- Security: SECURITY.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Agentics Monorepo
A TypeScript-first TurboRepo monorepo for building and refining agentic workflows with
GitHub-integrated automation and AI-assisted optimization.
## Project Overview
This repository provides:
- A monorepo architecture for workflow development and refinement
- A Commander.js CLI for refinement operations
- Modular packages for core logic, orchestration, GitHub integration, and AI analysis
- GitHub Actions for deterministic automation and GH-AW markdown for agentic automation
- GitHub workflows, issue templates, and rulesets for governance
## Monorepo Structure
```text
/apps
/cli
/packages
/core
/agentics
/github
/ai
/refinements
/docs
/workflows
/.github
/workflows
/ISSUE_TEMPLATE
/rulesets
```
## Agentic Workflow Philosophy
- Favor direct, deterministic tool invocation over repeated deliberation loops
- Minimize unnecessary reasoning about tools, permissions, and state
- Track optimization outcomes with measurable benchmark metrics
## CLI Usage
```bash
npm install
npm run cli -- refine run --workflow workflows/workflow-factory.md --run-id 1001
npm run cli -- refine analyze --conversation "Reasoning about tool call"
npm run cli -- refine extract --run-id 1001
```
## Refinement Lifecycle
1. Compile workflow (`gh aw compile`)
2. Execute workflow (`gh aw run`)
3. Create `refinements/{run_id}` and download artifacts
4. Extract `prompt.txt`, `conversation.txt`, and `usage.json`
5. Analyze transcripts via AI provider abstraction (default: Ollama)
6. Generate actionable tasks and benchmark improvements
## Workflow Examples
- `.github/workflows/context-cache.yml`
- `.github/workflows/context-cache-effectiveness.yml`
- `workflows/workflow-factory.md`
Reusable GH-AW markdown sources live in `/workflows`. Workflows used only by this repository live in
`.github/workflows` as normal GitHub Actions YAML.
Install reusable GH-AW sources into a target repository by copying them to that repository's
`.github/workflows/*.md`, then running `gh aw compile` there.
## Performance Optimization Philosophy
Optimization targets:
- Token usage reduction
- Tool call count reduction
- Execution time reduction
- Success-rate stability or improvement
## Contributing
1. Install dependencies: `npm install`
2. Run tests: `npm test`
3. Run type checks: `npm run typecheck`
4. Open a pull request using issue templates and required checks