Downstream User Assessment

April 17, 2026 · View on GitHub

Date: 2026-04-16 Author: heyong4725 + AI Scope: Evidence for plan §19.7 "Downstream user list + outreach" gate. Closes issue #291.

1. Conclusion first

No dora deployment surfaced on GitHub meets a reasonable definition of "production". The largest downstream repo is 69 stars; the median is ~0–2. The original plan specified a top-10 courtesy-outreach campaign (per §14 Appendix C). That scope is disproportionate to the actual user base. This assessment right-sizes the gate to:

  • Skip the individual outreach campaign.
  • Ship a release-note + migration-guide section that is prominent enough for any surprise production user to find without direct contact.
  • Monitor the issue tracker for the first 30 days post-release; pivot to targeted outreach if a production user surfaces.

2. Method

Three GitHub code-search queries run on 2026-04-16:

gh search code 'dora-node-api'      --language rust   --limit 30
gh search code 'from dora import Node' --language python --limit 30
gh search code '"dora-rs"'          --extension toml  --limit 30
gh search code 'dora-rs-cli'        --limit 20

Results de-duplicated; dora-rs org repos, tutorial/example/crawler/index repos, and obvious name collisions with unrelated dora-* projects filtered out.

3. Findings

3.1 Size of the external user base

  • ~30 unique external repos reference dora in any form.

  • Top 5 by star count:

    RepoStarsLast pushWhat it is
    YOR-robot/YOR692026-04-14"Your Own Robot" — hobby/educational robot kit with MuJoCo sim
    Ekumen-OS/lekiwi352026-02-12Monorepo for LeKiwi community robot (Ekumen Labs, a robotics consulting shop)
    FlagOpen/RoboDriver322026-04-15BAAI's standardized robot driver layer — R&D
    kornia/bubbaloop162026-04-15Kornia's "hardware AI agent" single-binary Rust runtime
    mofa-org/mofa-studio122026-03-29Desktop voice-chat app (Rust/Makepad) that happens to use dora

  • Everything else is 0–10 stars and most are tutorial / hackathon / test / pkg-manager experiment repos (*dora-pkg-mngr*, *dora-camp-tutorial*, *dora-hackathon*, *_test, *_example).

3.2 Classification

ClassCountSignal
Production deployment (revenue-generating, customer-facing)0 identifiedNo repo describes a paying customer or uptime SLA
Research / R&D project (academic, lab, OSS org)~6Ekumen-OS/, FlagOpen/, kornia/bubbaloop, YOR-robot/YOR, mofa-org/*
Hobby / POC / learning~20Hackathon names, tutorials, personal experiments
Tooling / meta~4pkg-crawlers, CI indexes, unrelated name collisions

3.3 Red-flag check (would any of these break badly?)

For each of the top 5:

  • YOR-robot/YOR: README names dora only as a distributed-computing backend. A hard break at 1.0 means a one-line dependency bump + CLI rename awareness. Two-hour migration.
  • Ekumen-OS/lekiwi: Apache-2.0 licensed, CI green, described as a "research monorepo". Ekumen Labs is a consulting shop; they would absorb a migration as part of a regular tool upgrade. No production deployment visible.
  • FlagOpen/RoboDriver: 32 stars, bilingual README, BAAI-sponsored. Research project, not in production deployment. Would likely welcome the 1.0 cut as a clear versioning signal.
  • kornia/bubbaloop: early-stage, 16 stars, single-binary design. Probably tracking our main already; migration is part of its growth path.
  • mofa-org/mofa-studio: voice-chat desktop app, tangentially depends on dora. Dora is one of many integrations.

None of the five would be materially harmed by a hard 1.0 break with a clear migration guide.

4. Right-sized action plan

4.1 What we will do

  • Release note prominence. The 1.0 release note mentions the hard break explicitly and links the migration guide. This is the primary communication channel for any surprise production user.
  • Migration guide. docs/migration-from-0.x.md (tracked as plan §3.5 and #295 rename residue). Keep it linkable, keep it short, keep it up top in the release announcement.
  • Issue-tracker watch for 30 days post-release. Any new issue tagged migration or naming a production deployment triggers a direct-outreach response within 24 hours.
  • One-line Discord / community ping. A "dora 1.0 is coming; here's what changes" note in whatever chat channel the project uses. Low-effort; catches users who don't watch GitHub releases.

4.2 What we will not do

  • File courtesy GitHub issues on the top-10 downstream projects asking for early migration testing. The top-10 is ~5 distinct real users and ~5 abandoned tutorials; the signal-to-noise is not worth the effort.
  • Maintain an email list of production users. None exist to list.
  • Ship a bridge release (0.6) to smooth rolling upgrades. Already rejected in D-1a (see phase--1-audit-2026-04-16.md §6); reaffirmed here — even for the top-5 users, a hard break is cheaper than engineering a bridge.

4.3 Trigger-to-expand

If any of the following surfaces post-release, expand to individual outreach:

  • A GitHub issue from a maintainer of a repo not in our search (production users may not be OSS-visible).
  • A Reddit / HN / Discord comment naming a deployment that the release note didn't reach.
  • A maintainer-sponsored fork that sticks to 0.x past 90 days, indicating migration friction.

5. Gate update

  • Plan §19.7 "Downstream user list + outreach" row flips from Not done → Done (scope right-sized) with a pointer to this file.
  • Plan §14 Appendix C (the top-10 outreach protocol) is annotated as historical — superseded by this right-sized assessment.
  • Plan §11 Success Criterion "At least 3 downstream projects have confirmed successful migration to 1.0 via dora migrate" is removed; replaced by "No production user report unresolved migration blockers in the first 30 days post-release."
  • Issue #291 can be closed.

Honest caveat

This assessment is based on what's visible on GitHub. A production user running dora in a private monorepo would not surface. If the user knows of such a deployment not searchable by public GitHub code search, that user should be added back to the direct-outreach list manually. The right-sizing rests on the current best-effort search showing zero such deployments.