8000 heggria (Heggria) · GitHub
[go: up one dir, main page]

Skip to content
View heggria's full-sized avatar
:bowtie:
:bowtie:

Block or report heggria

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
heggria/README.md
Heggria — Work that outlives the chat. An animated workflow schematic: branch, verify, hold, replay.

I build verifiable agent workflows and tools with a point of view.
Agent infrastructure & product engineering at OpenDesign · Shanghai

Work ↗    /    Writing ↗    /    Collaborate ↗

Selected work

01 — taskflow. Agent workflows. Make the run inspectable.

taskflow · Creator & maintainer
Declarative workflows for coding agents. Verify the graph, inspect what happened, replay the trace, and recompute what changed.

Explore the code ↗ · Read the docs · Try an example

Published beta · MIT · CI · Host support & limits


02 — OpenDesign. At work. Design meets engineering. Conceptual illustration.

OpenDesign · Where I work
An open-source, local-first design application. My work sits between agent infrastructure and full-stack product engineering.

Explore OpenDesign ↗


03 — SelfField. Personal experiment. A mirror, without the labels. Conceptual illustration.

SelfField · Personal experiment
A personal mirror built around evidence and tension, without personality labels.

Open the experiment ↗ · Source

Fresh from the workbench

Recent public releases and writing. Dates show publication, not profile refreshes. Writing is in Chinese.

Releases

Writing

All releases · All writing · RSS

Let's build something

Interested in agent orchestration, incremental computation, or making runs auditable? Bring a reproducible workflow, a host integration fix, or a small example we can ship together.

Start with an issue · Contribution guide · Email

Under the hood — a real run & honest boundaries

The header is an animated schematic. This is a recorded Pi run from my existing profile:

⊗ taskflow self-improve  6/7 · blocked · $0.095
    ✓ discover            agent
  ┌ ✓ write-runner-tests  agent
  ├ ✓ write-store-tests   agent
  ├ ✓ write-agents-tests  agent
  └ ✓ fix-stability       agent
    ✓ verify              gate    BLOCK 3 type errors in test files
    ⊘ report              reduce  skipped · Gate blocked

The verification gate stopped downstream work. The trace explains why.

The published Trusted Effects beta describes declared filesystem effects and ledger-backed explanations. Beta, not GA or an OS sandbox. Adapter availability is separate from verified host enforcement: read the support baseline and security boundaries.

Notes, side projects & why I build this way

Getting something to run once is only the beginning. I want the next person — including future me — to understand what happened and make the next run cheaper.

Evidence outlives confidence. Constraints give a tool its character.

Personal site · Bilibili


Made to be understood. Built to be revisited.
Original SVG studies · light / dark · reduced-motion aware

Pinned Loading

  1. taskflow taskflow Public

    DAG runtime for coding-agent workflows with verification, resume, replay, and effect checks.

    TypeScript 80 11

  2. heggria.github.io heggria.github.io Public

    Personal blog built with Astro, with RSS and full-text search.

    Astro 1

  3. home-compass home-compass Public

    Beijing neighborhood scorecards and a mortgage affordability calculator.

    TypeScript 1

  4. selffield selffield Public

    Browser-local journal for recording experiences, reviewing evidence, and tracking changes over time.

    TypeScript

0